JP2001016588A - Signal processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a signal processing system that applies compression coding to image data (or audio data) with high efficiency. SOLUTION: The signal processing system is provided with a storage means 2 that stores a coefficient obtained by decomposing image data (or audio data) into a plurality of frequency components, a coefficient read control means 3 that reads the coefficient from the frequency component storage means 2, a read control information generating means 4 that generates read control information C, and a coefficient encoding means 5 that encodes the coefficient. The read control information generating means 4 has a scanning sequence decision means 401 that stores information denoting the scanning sequence of each frequency component, a scan sequence table 402 that stores information of the scanning sequence and a scan sequence information encoding means 403 that encodes the information denoting the scanning sequence.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing technique for efficiently compressing and encoding information such as image data and audio data.

[0002]

2. Description of the Related Art In recent years, signal systems for efficiently compressing and encoding information such as image data and audio data have been actively studied.

[0003] "Signal processing system" refers to the system L
In the case of software that runs on a computer, it has a configuration of SI and other hardware, or software that runs on a computer, or hardware that is built into a computer, or both software and hardware. In the form of a program that runs on
The processing is performed by the PU, and the various storage units are configured by a memory or a hard disk on the computer.

[0004] Further, the "signal processing system" is also a signal processing method itself. The “signal processing system” is also a recording medium on which a program that operates on a computer is recorded.

[0005] These signal processing techniques relating to compression and encoding are important technologies that are indispensable to the development of industry, leading to reduction in costs related to storage and transmission of image data (or audio data). Among them, a signal processing technique for decomposing image data (or audio data) into a plurality of frequency components (hereinafter, referred to as frequency decomposition) and compressing and encoding the same is being actively studied all over the world.

As typical examples of the frequency decomposition technique, discrete cosine transform (hereinafter, DCT), Fourier transform, Karhunen-Loeve transform, Hadamard transform, Haar transform,
There are various orthogonal transforms such as a wavelet transform.

[0007] Image data obtained by signal conversion of image data photographed by a video camera or the like or a photograph or the like by an optical reading means such as a scanner is, for example, temporally (or spatially) adjacent (neighboring). ) Pixel values have characteristics based on natural phenomena, such as similarity. By utilizing such features based on natural phenomena of image data, it is possible to reduce the information amount of image data. For example, when image data in which values of adjacent pixels are similar are frequency-decomposed into frequency components, energy is concentrated in low-frequency components.
Therefore, even if the data of the low-frequency component is preferentially used and the data of the high-frequency component is somewhat reduced, it is possible to perform the compression encoding with little deterioration. In addition, since audio data is a combination of a plurality of sound waves having different frequencies and amplitudes, it is possible to perform compression encoding after performing frequency decomposition on audio data.

As described above, among the signal processing techniques for frequency-decomposing and compressing and encoding image data (or audio data), MPEG and JPEG have already been put into practical use as standard techniques in the world. In MPEG and JPEG, DCT is used as a technique for frequency-decomposing input image data.

Hereinafter, a conventional signal processing system will be described.
As an example, an outline of a compression encoding method used in MPEG and JPEG will be described.

FIG. 81 shows the configuration of a conventional signal processing system (compression encoding side). In the following signal processing system, an input signal will be described as an example as image data. (Sound data can be processed in the same way, but will be omitted in the following description.) In FIG.
Is a frequency decomposing means for decomposing image data as an input signal into a plurality of frequency components, and 10 is a quantizing means for quantizing data after frequency decomposition (hereinafter referred to as coefficient A).
2 is a frequency component storing means for storing the quantized data of the coefficient A (hereinafter, referred to as coefficient B), and 3 is an address signal D given to the frequency component storing means (2) for storage. Coefficient reading control means for reading the coefficient B
Reference numeral 450 denotes a fixed scan order table in which the order in which each coefficient B is read from the frequency component storage means (2) (hereinafter referred to as a scan order) is stored in advance as a fixed value.
Is a coefficient encoding means for compression encoding each coefficient B read from the frequency component storage means (2).

In the following description, as an example, it is assumed that the frequency resolving means (9) performs DCT, the fixed scan order table (450) stores a scan order called zigzag scan, and the coefficient encoding is performed. Means (5
In 1), Huffman coding is performed. The fixed scan order table (450 in FIG. 81) will be described with respect to zigzag scan, but in MPEG, a scan order called an alternate scan is also defined. However, the fact that the alternate scan is also prepared in a fixed manner in advance is the same as the zigzag scan, and the alternate scan is omitted in the following description.

FIGS. 82 and 83 show the flow of processing in a conventional signal processing system (compression encoding side). Hereinafter, the operation of the conventional signal processing system (compression encoding side) will be described. As shown in FIG. 83, one frame of original image data is divided into 8 × 8 pixel original image blocks, and compression encoding is performed for each original image block. The data of the original image block generally includes a luminance signal (Y),
Although they are color difference signals (U, V), only the luminance signal (Y) will be described here. The range of the luminance signal (Y) of the original image is generally “0 to 255”. In the case of MPEG, the value of "0 to 255" is directly used as an input signal for subsequent compression encoding processing. In the case of JPEG, 128 is subtracted from each value in the range of "0 to 255". The value obtained is used as an input signal. Thus, even in the prior art, the value range of the input signal is
Although it may be different, the following description will be given assuming that the value range of the input signal is “0 to 255” as an example.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 82).
01).

<DCT and Quantization> Next, in the frequency decomposition means (9 in FIG. 81), the input original image block is subjected to a DCT operation as shown in Equation 1 (S002 in FIG. 82), and the quantization is performed. In the conversion means (10 in FIG. 81), DC
Each coefficient after T is quantized (S003 in FIG. 82). Here, as an example, all coefficients are quantized with the value of the uniform quantization step being 1.

[0015]

(Equation 1)

In equation (1), f (i, j) is calculated as shown in FIG.
4, the original image block f (i, j), and F (n, m) in Equation 1 is the original image block f (
i, j) is the result of performing a DCT operation on
This is the DCT block F (n, m) in the example of FIG. Note that, as shown in Expression 1, the DCT operation is a real number calculation. DCT block F (n, m) in FIG. 84
Strictly speaking, the result obtained by performing the DCT operation according to Equation 1 is quantized with the value of the quantization step as 1, and expressed as an integer value closest to the real number immediately after the DCT operation.

In the DCT block F (n, m), the coefficient at the position where both the horizontal frequency n and the vertical frequency m are zero is a DC component, and the other coefficients are AC components.
Component. As described above, the frequency decomposition means (9 in FIG. 81) has a function of decomposing an original image block, which is an input signal, into a plurality of frequency components (a DC component and 63 AC components). As described above, image data obtained by a video camera or the like or a photograph or the like obtained by an optical reading means such as a scanner may be such that adjacent (neighboring) pixel values are similar. Such characteristics are based on natural phenomena. Therefore, when image data is frequency-decomposed into frequency components, energy is concentrated on low frequency components.

As described above, each coefficient of the quantized DCT block is stored in the frequency component storage means (2 in FIG. 81).

<Zigzag Scan> Next, according to the scan order of the zigzag scan stored in the fixed scan order table (450 in FIG. 81), as shown in FIG.
The quantized DCT blocks stored in the frequency component storage means (2 in FIG. 81) are read out and arranged one-dimensionally (S004 in FIG. 82).

<Huffman Coding> Next, each coefficient of the one-dimensionally arranged DCT block (one-dimensional data in FIG. 83)
Are subjected to Huffman coding (S005 in FIG. 82) by coefficient coding means (51 in FIG. 81).

Hereinafter, the Huffman coding performed here will be briefly described. In the one-dimensional data, the number of consecutive zero coefficients (hereinafter referred to as run), the absolute value of the non-zero coefficient (hereinafter referred to as amp), and whether the non-zero coefficient is positive or negative (hereinafter referred to as sign)), and r
By outputting one code from the combination of “run, amp, and sign” while comparing it with a Huffman code table corresponding to un, amp, and sign (hereinafter, referred to as a Huffman table), compression encoding is performed. Is

The Huffman table used here is "r
In the combination of “un, amp, and sign”, the more frequently appearing (higher appearing frequency) is represented by the “shorter code”, so that the compression becomes possible. Generally, the smaller the “amp” (or the smaller the run), the higher the appearance frequency, and the Huffman code is configured to be represented by a short code. Table 28 shows an example of the Huffman table.

[0023]

[Table 1]

In Table 28, "S" at the end of each code is used.
"Is for expressing a sign (hereinafter, referred to as signature information). This signature information is" zero "when the non-zero coefficient is positive, and" 1 "when the non-zero coefficient is negative.
". For example, if the one-dimensional data is “0, 0,
In the case of "0, -4", since three zeros continue, the value of run is "3" and the subsequent non-zero coefficient is "-4", so the value of amp is "4" and the non-zero coefficient is negative. Therefore, the signature information (S in Table 28) is “1”.
Therefore, in the example of Table 28, the one-dimensional data “0,0,
"0, -4" is represented by a 4-bit Huffman code "0111".

The above is one original image block (FIG. 83).
Is a series until the compression encoding is performed. This process is performed on all blocks (S006 in FIG. 82),
The original image of the frame (FIG. 83) can be compression encoded.

Next, a description will be given of a process of decompressing (decoding) compression-encoded data.

FIG. 85 shows the configuration of a conventional signal processing system (extension decoding side). In FIG. 85, reference numeral 52 denotes a coefficient decoding means for decoding the Huffman code of the input compressed coded data, and reference numeral 21 denotes a frequency component storage means for decoding by the coefficient decoding means (52). The coefficient of each frequency component obtained by the conversion is stored.
Numeral 5 is a coefficient writing control means for giving an address signal D to the frequency component storage means (21) and controlling storage (writing) of each coefficient, and 450 is a frequency component storage means (2).
1) A fixed scan table in which the order (scan order) for writing coefficients in is set as a fixed value in advance.
The set scan order value is the same as that used in the compression encoding. Reference numeral 720 denotes an inverse quantization means for performing inverse quantization on each coefficient stored in the frequency component storage means (21), and 709 synthesizes each frequency component using the inversely quantized coefficient. It is a frequency synthesis means. Note that, in this example, the frequency synthesizing unit (709)
Performs an inverse DCT operation.

FIGS. 86 and 87 show the flow of processing in the conventional signal processing system (decompression decoding side). Hereinafter, the operation of the conventional signal processing system (decompression decoding side) will be described.

<Huffman Decoding> First, the coefficient decoding means (52 in FIG. 85) decodes the Huffman code of the input compressed and coded data to obtain one-dimensional data as shown in FIG. 87 (FIG. 86). S001). The decryption is as shown in Table 28
The run / amp / sign combination is derived by referring to the Huffman table shown in FIG. 87 to obtain one-dimensional data (FIG. 87).

<Zigzag Scan> Next, according to the scan order of the zigzag scan stored in the fixed scan order table (450 in FIG. 85), one-dimensional data is converted into a two-dimensional DCT block (FIG. 87). (Fig. 87)
The frequency component storage means (FIG. 8)
5-2) (S002 in FIG. 86).

<Inverse Quantization> Next, the DCT block (FIG. 8)
In 7), inverse quantization of each coefficient is performed by the inverse quantization means (720 in FIG. 85) (S003 in FIG. 86). If the quantization is performed using a large quantization step value at the time of compression encoding, the error is large, and the DCT block after the inverse quantization becomes the original (before the quantization before the compression encoding ) DC
It cannot be restored to the exact same state as the T block.

<Inverse DCT> Next, frequency synthesis means (FIG. 8)
5 709), the inverse DCT operation shown in Equation 2 is performed on the inversely quantized DCT block (FIG. 87).
A restored image block (FIG. 87) is obtained (S00 in FIG. 86).
4).

[0033]

(Equation 2)

F (n, m) in Equation 2 is, for example, a DCT block F (n, m) as shown in FIG. 88, and f (i, j) in Equation 2 is, for example, as shown in FIG. For the DCT block F (n, m), the inverse DC
This is a result of performing the T operation, and is a result of the restoration image block f (i,
j). Note that, as shown in Equation 2, the inverse DCT operation is a real number calculation.
Strictly speaking, (i, j) expresses the result obtained by performing the inverse DCT operation by Expression 2 as the nearest integer. This inverse DCT includes a plurality of frequency components (a DC component and
This is to generate a restored image by synthesizing the waveforms of (63 AC components). However, although DCT and inverse DCT are real number calculations, since they are converted into integers, a slight error occurs at this point. Therefore, the restored image block f (i, j) obtained by performing the inverse DCT operation on the DCT block F (n, m) in FIG. 88 is completely the same as the original image block f (i, j) shown in FIG. It is not always restored to the state.

According to the above procedure, one restored image block is obtained. This process is repeated until the processing of all information for one frame is completed (S005 in FIG. 86), so that one frame of restored image data can be obtained.

[0036]

Problems to be solved by the conventional signal processing system (the following first to fifth problems) will be described.

<First Problem> A first problem of the above-mentioned conventional signal processing system is that a fixed scan order table (FIG. 8) is used.
1 (450 in FIG. 82) is not necessarily an effective compression means. It is. less than,
The reason will be described.

FIGS. 89, 90 and 91 show DCT blocks 1 and 2 respectively as examples of quantized DCT blocks.
The DCT block 2 and the DCT block 3 are shown. These three
Each of the DCT blocks includes the same number of non-zero coefficients (1 is 4, 2 is 3, 3 is 3, 4 is 2)
, 5 is 1) is the location (coordinate) of each value
Are different.

Each of D in FIGS. 89, 90 and 91
A process of arranging data in a two-dimensional state one-dimensionally is performed on the CT block according to the scan order of the zigzag scan. At this time, encoding is performed up to the last non-zero coefficient, and a code (hereinafter, this code is referred to as EOB) is given, which means that there is no non-zero coefficient thereafter. That is, as information of each block,
Up to EOB is encoded, and subsequent zero coefficients are not encoded. As a result of performing a zigzag scan on DCT block 1 (FIG. 89) and arranging the data in one dimension, up to the last non-zero coefficient, 0 is zero, 1 is 4, 2 is 3, and 3 is 3, 4, 2 and 5 are 1 in total,
There are thirteen coefficients. Therefore, the number n of each coefficient B
(B) and the probability P (B) are as shown in Table 29, and EOB
Is about 28.6 [bit]. Here, the information amount was calculated according to Equation 3. In this case, N in Expression 3 is 5.

[0040]

[Table 2]

[0041]

(Equation 3)

However, DCT block 2 (FIG. 9)
0), in the DCT block 3 (FIG. 91), this zigzag scan does not function effectively as compression means. Table 3
0 and Table 31 respectively show DCT block 2 (FIG. 9).
0) and the number n (B) of coefficients B and the probability P (B) as a result of performing the zigzag scan on the DCT block 3 (FIG. 91).

[0043]

[Table 3]

[0044]

[Table 4]

DCT block 2 (FIG. 90) and DCT
In block 3 (FIG. 91), the non-zero coefficients are 4 as 1, 2, 3, 3, 4, 2, and 5, as in DCT block 1 (FIG. 89). There are 13 in total. However, by performing the zigzag scan, DCT block 2 (FIG. 90) and DCT block 3
(FIG. 91) differs from DCT block 1 (FIG. 89) in that 15 zero coefficients are sampled. This is compared to DCT block 1 (FIG. 89),
The CT block 2 (FIG. 90) and the DCT block 3 (FIG. 9)
In 1), the non-zero coefficients are skewed upward in the block. DCT block 2 (FIG. 90)
And the information amount of the DCT block 3 (FIG. 91) excluding the EOB is about 56.5 [bit].
The amount of information in the T block 1 (FIG. 89) has greatly exceeded. Sampling the coefficients by zigzag scanning in this manner is equivalent to DCT block 1 (FIG. 8).
This is effective when the non-zero coefficients are concentrated in the upper left part of the block (where the horizontal frequency and the vertical frequency are low) as in 9).
When coefficients are not concentrated on the upper left as in block 3 (FIG. 91), there is a problem that a very large waste occurs.

In addition, for example, in MPEG2, there is an alternate scan in which scanning is performed in another order in addition to the zigzag scan. However, since this is also a fixed value similar to the zigzag scan, it is not necessarily the same as the zigzag scan. It is not always suitable for the block, and waste such as the above example may occur.

<Second Problem> A second problem of the above-described conventional signal processing system is that each coefficient (one-dimensional data in FIG. 83) of the one-dimensionally arranged DCT block is converted into coefficient encoding means (FIG. 81). 51), one non-zero coefficient is set to one non-zero coefficient in the Huffman coding process (S005 in FIG. 82).
Since the sign information of [bit] is used, the code amount is increased.

For example, as with a general television,
If the image size of one frame is composed of 720 × 480 pixels, 5400 DCs per frame
There are T blocks. When it is desired to improve the image quality of the data after compression encoding (that is, when it is desired to reduce the image quality degradation), the value of the quantization step at the time of quantization is set to be small. In that case, each coefficient of the AC component of the quantized DCT block is likely to be non-zero. For example, if all the coefficients are non-zero, then 5400 × 63 = 340200 [bit] = 42.
525 [KB] signature information is required. As described above, there is a problem that many codes are required for the sign information.

<Third Problem> A third problem of the above-mentioned conventional signal processing system is that each coefficient (one-dimensional data in FIG. 83) of the one-dimensionally arranged DCT block is converted into coefficient encoding means (FIG. 81). 51), in the Huffman coding process (S005 in FIG. 82), the Huffman table used generally has a shorter code (a smaller amplitude absolute value (amp) of each coefficient of the AC component of the DCT block). 81, a coefficient having a large amplitude absolute value (amp) is encoded by a number encoding means (FIG. 81).
If the frequency of input to (51) is high, there is a problem that the code length (code length) becomes longer as a whole.

<Fourth Problem> The fourth problem of the above-mentioned conventional signal processing system is that frequency decomposition such as DCT operation is performed as an integer and compression-encoded even though it is a real number operation. Even if the inverse DCT operation is performed on the block on which the DCT operation has been performed, the original image block f (i, j) in FIG. 84 and the restored image block f (i, j) in FIG.
As in the example of j), there is a problem that an error occurs between the original image block and the restored image block.

<Fifth Problem> A fifth problem of the conventional signal processing system is that it cannot cope with an operation error caused by a physical problem of the signal processing system itself. For example, a semiconductor component or the like used inside a signal processing system (hardware or a computer that performs signal processing) may not operate normally depending on a temperature condition in which the component is used. Generally, it is considered that a semiconductor component or the like is commercialized and put on the market after obtaining a sufficient temperature resistance test or the like. Therefore, it is considered that the device normally operates normally even under severe conditions. However, in recent years, there has been an increasing number of opportunities to use various electrical products, such as mobile phones and notebook computers, in outdoor locations where the temperature is high. As described above, the electric appliances themselves have been used under more severe conditions than before. Therefore, it is conceivable that the device does not operate normally instantaneously. Thus, there is a possibility that a failure based on a natural phenomenon occurs. In the event that a normal operation does not occur momentarily while the input signal is being compression-encoded by the above-described conventional signal processing system, the compression-encoded data has an unintended configuration, and is subsequently decompressed. There is a problem that even if decoding (decoding) is performed, it cannot be correctly restored.

The present invention provides the above first to fifth objects.
The purpose is to solve.

[0053]

According to a first aspect of the present invention, at least image data (or audio data) includes a plurality of frequency components. A basic encoding means (1) for receiving a coefficient A (or a coefficient B obtained as a result of quantizing the coefficient A) obtained as an input and outputting compressed encoded data; (1) frequency component storage means (2) for storing at least coefficient A (or coefficient B) therein; and frequency component storage means (2).
Generates an address signal D for the coefficient A (or coefficient B) stored in the frequency component storage means (2).
(3) a coefficient readout control means (3) for reading out (or coefficient B) and at least a coefficient A (or coefficient B) for inputting the coefficient A (or coefficient B) stored in the frequency component storage means (2) Read control information generating means (4) for generating read control information C;
A coefficient encoding unit (5) for encoding the coefficient A (or coefficient B) is provided. Inside the read control information generating unit (4), the frequency component storage unit ( A scan order determining means (401) for determining the scan order of each frequency component in 2); and a scan for storing information on the scan order determined by the scan order determining means (401) (hereinafter, scan order information E). A scan order information encoding unit (403) that encodes the scan order information E. At least the scan order information E is used as the readout control information C, and the coefficient readout control unit (3) , An address signal D is generated based on at least the read control information C.

With the above arrangement, the scan order determining means (401) creates and creates a scan order table suitable for the coefficient A (or coefficient B) of the input signal (image data or audio data) after frequency decomposition. Since each coefficient A (or coefficient B) is read out and sampled in accordance with the scan order information E, encoding suitable for the input signal can be performed as compared with the related art using a fixed scan order table, and the information amount as a whole is improved. Can be reduced. Thus, the first problem can be solved.

According to the second aspect of the present invention, at least the coefficient A obtained by decomposing the image data (or audio data) into a plurality of frequency components (or the coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least the coefficient A (or coefficient B) inside the basic coding means (1). And a coefficient read control for generating an address signal D of the coefficient A (or coefficient B) stored in the frequency component storage means (2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2). Means (3) and at least the complexity information F of the image data (or audio data) as input and read out the coefficient A (or coefficient B) stored in the frequency component storage means (2) And read control information generating means for generating a read control information C (4), the coefficient A
(Or coefficient B) for encoding the frequency components in the frequency component storage means (2) from the complexity information F inside the read control information generation means (4). Grouping means (404) for classifying the information into a plurality of complexity groups, and a complexity group information storage means (40) for storing the complexity group information G classified by the complexity group classification means (404).
5) and a complexity group information encoding means (406) for encoding the complexity group information G 1, wherein at least the complexity group information G is used as the read control information C, and the coefficient read control means (3) , The address signal D is generated based on at least the read control information C.

According to the above configuration, the complexity group classifying means (404) converts the input signal (image data or audio data) or the coefficient A (or coefficient B) obtained by frequency-decomposing the input signal into a plurality of complexity groups. And the coefficient A from the frequency component storage means (2) for each complexity group.
(Or coefficient B), and the coefficient encoding means (5) performs statistical processing on the coefficient A (or coefficient B) to be read, and encodes the coefficient A (or coefficient B). One complexity group includes absolute amplitude values (am
p) can be biased for each complexity group, such as collecting values with large amplitudes and certain amplitude groups with small amplitude absolute values (amp), and the amount of information in each group. Can be reduced.

According to the third aspect of the present invention, a coefficient A obtained by decomposing at least image data (or audio data) into a plurality of frequency components (or a coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least the coefficient A (or coefficient B) inside the basic coding means (1). And a coefficient read control for generating an address signal D of the coefficient A (or coefficient B) stored in the frequency component storage means (2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2). Means (3); absolute value conversion means (6) for obtaining the absolute value of coefficient A (or coefficient B); and coefficient coding means (encoding the absolute value of each coefficient A (or coefficient B) of the frequency component ( 5) and Sign information processing means (7) which receives A (or coefficient B) as an input and processes a sign (hereinafter, sign information) indicating whether the inputted coefficient A (or coefficient B) is positive or negative. In the signature information processing means (7), it is determined whether or not the coefficient A (or coefficient B) is non-zero. If the coefficient is non-zero, the signature information is extracted.
And signature information storage means (70) for storing signature information.
2) and sign information encoding means (703) for performing statistical processing on the sign information and encoding the sign information.

With the above configuration, the non-zero determination sign information extracting means (70) is internally provided in the sign information processing means (7).
In 1), only the sign information of the non-zero coefficient is extracted, and the sign information encoding means (703) performs statistical processing of the sign information and performs compression encoding. As a result, in the related art, 1 [bit] of sine information is required for one non-zero coefficient in each coefficient after frequency decomposition. In contrast to the above, the sign information is collectively subjected to the compression encoding, whereby the code amount required for the sign information can be reduced. This solves the second problem.

According to the fourth aspect of the present invention, at least a coefficient A obtained by decomposing image data (or audio data) into a plurality of frequency components (or a coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least the coefficient A (or coefficient B) inside the basic coding means (1). And a coefficient read control for generating an address signal D of the coefficient A (or coefficient B) stored in the frequency component storage means (2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2). Means (3) for inputting at least the coefficient A (or coefficient B) and reading out the coefficient A (or coefficient B) stored in the frequency component storage means (2)
Read control information generating means (4) for generating read control information C; absolute value converting means (6) for obtaining the absolute value of coefficient A (or coefficient B);
Coefficient encoding means (5) for encoding the absolute value of (or coefficient B), and signature information for processing the signature information of the input coefficient A (or coefficient B) with input of coefficient A (or coefficient B) A processing unit (7) for determining the scanning order of each frequency component in the frequency component storage unit (2) from the input coefficient A (or coefficient B) inside the read control information generation unit (4) Scan order determining means (4)
01), a scan order table (402) storing information on the scan order determined by the scan order determining means (401) (hereinafter, scan order information E), and a scan order information code for encoding the scan order information E 1 Means (4
03), and at least the scan order information E is used as read control information C, and the coefficient read control means (3)
, The address signal D is generated based on at least the read control information C. In the sign information processing means (7), it is determined whether or not the coefficient A (or coefficient B) is non-zero. Non-zero determination sign information extracting means (701) for extracting sign information when the value is zero
And signature information storage means (70) for storing signature information.
2) and sign information encoding means (703) for performing statistical processing on the sign information and encoding the sign information.

With the above configuration, the scan order determining means (401) creates and creates a scan order table suitable for the coefficient A (or coefficient B) of the input signal (image data or audio data) after frequency decomposition. Since each coefficient A (or coefficient B) is read out and sampled in accordance with the scan order information E, the number of zero coefficients to be encoded can be reduced as compared with the related art using a fixed scan order table, and overall The amount of information can be greatly reduced. In the sign information processing means (7), the non-zero determination sign information extracting means (701) extracts only the sign information of the non-zero coefficient, and outputs the sign information encoding means (70).
In 3), the signature information is statistically processed and compression-encoded. As a result, in the related art, 1 [bit] of sine information is required for one non-zero coefficient in each coefficient after frequency decomposition. In contrast to the above, the sign information is collectively subjected to the compression encoding, whereby the code amount required for the sign information can be reduced. Thus, the first problem and the second problem can be simultaneously solved.

According to the fifth aspect of the present invention, at least a coefficient A obtained by decomposing image data (or audio data) into a plurality of frequency components (or a coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least the coefficient A (or coefficient B) inside the basic coding means (1). And a coefficient read control for generating an address signal D of the coefficient A (or coefficient B) stored in the frequency component storage means (2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2). Means (3) and at least the complexity information F of the image data (or audio data) as input and to read out the coefficient A (or coefficient B) stored in the frequency component storage means (2) And read control information generating means for generating a read control information C (4), the coefficient A
Absolute value conversion means for obtaining the absolute value of (or coefficient B) (6)
And coefficient encoding means (5) for encoding the absolute value of each coefficient A (or coefficient B) of the frequency component, and the coefficient A (or coefficient B) which receives the coefficient A (or coefficient B) as an input Sign information processing means (7) for processing the sign information of
And a complexity group classification for classifying each frequency component in the frequency component storage means (2) into a plurality of complexity groups from the input complexity information F inside the read control information generating means (4). Means (404);
Classified by the complexity group classifier (404)
A complexity group information storage unit (405) for storing the complexity group information G; and a complexity group information encoding unit (406) for encoding the complexity group information G. Used as read control information C, the coefficient read control means (3) generates an address signal D based on at least the read control information C, and the signature information processing means (7)
, Determining whether the coefficient A (or coefficient B) is non-zero, and in the case of non-zero, sign information extracting means (701) for extracting signature information; and a signature for storing signature information. Information storage means (702);
It is configured to have a signature information encoding unit (703) that performs statistical processing on the signature information and encodes the signature information.

According to the above configuration, the complexity group classifying means (404) divides an input signal (image data or audio data) or a coefficient A (or coefficient B) obtained by frequency decomposition of the input signal into a plurality of complexity groups. And a coefficient A from the frequency component storage means (2) for each complexity group.
(Or coefficient B), and the coefficient encoding means (5) performs statistical processing on the coefficient A (or coefficient B) to be read, and encodes the coefficient A (or coefficient B). One complexity group includes absolute amplitude values (am
p) can be biased for each complexity group, such as collecting values with large amplitudes and certain amplitude groups with small amplitude absolute values (amp), and the amount of information in each group. Can be reduced. Further, inside the sign information processing means (7), only the sign information of the non-zero coefficient is extracted by the non-zero determination sign information extracting means (701), and the sign information encoding means (703)
The signature information is statistically processed and compression-encoded. As a result, in the related art, 1 [bit] of sine information is required for one non-zero coefficient in each coefficient after frequency decomposition. In contrast to the above, the sign information is collectively subjected to the compression encoding, whereby the code amount required for the sign information can be reduced.

According to a sixth aspect of the present invention, there is provided an adjacent sample difference calculating means (8) for calculating a difference between adjacent sample values of image data (or audio data) (hereinafter referred to as an adjacent sample difference value); Frequency decomposition means (9) for decomposing the difference value into a plurality of frequency components and outputting coefficients A of the respective frequency components; and at least a coefficient A (or a coefficient B which is a result of quantizing the coefficient A) A frequency component storage means (2) having at least a coefficient A (or a coefficient B) inside the basic coding means (1) for outputting compressed coded data; An address signal D for the coefficient A (or coefficient B) stored in the frequency component storage means (2) is generated, and the coefficient readout for reading out the coefficient A (or coefficient B) from the frequency component storage means (2). And control means (3), the coefficient A (or coefficient B) configured to have a coefficient encoding means for encoding (5).

With the above arrangement, the difference between adjacent sample values of image data (or audio data) (hereinafter, adjacent sample difference value) is calculated by the adjacent sample difference calculation means (8), and the frequency decomposition means ( In 9), by performing frequency decomposition on the adjacent sample difference value, the amplitude absolute value of each coefficient after frequency decomposition can be reduced as compared with the related art, and the amount of information can be significantly reduced. by this,
The third problem can be solved.

According to the seventh aspect of the present invention, image data (or audio data) itself or a difference between adjacent sample values of image data (or audio data) (hereinafter, an adjacent sample difference value) is used as the original signal H. Frequency decomposition means (9) which takes the original signal H as an input, decomposes the original signal H into a plurality of frequency components and outputs coefficients A of the respective frequency components, and a coefficient A obtained from the frequency decomposition means (9).
Quantizing means (10) for outputting a coefficient B after quantization, basic encoding means (1) for receiving at least the coefficient B as input and outputting compressed coded data, Compression-encoded data storage means (1
1) and an inverse quantization means (13) for inversely quantizing the coefficient B
Frequency synthesis means (709) for frequency-synthesizing the inversely quantized coefficient B; frequency synthesis result storage means (710) for storing a frequency synthesis result by the frequency synthesis means (709); Difference value calculation means (16) for calculating a difference value from the result (hereinafter, difference value M)
And difference value storage means (17) for storing the difference value M
A difference value encoding means (18) for encoding the difference value M and outputting the difference value encoded data; a difference value encoded data storage means (19) for storing the difference value encoded data;
The compressed coded data stored in the compressed coded data storage means (11) and the difference coded data storage means (1
9) A total code amount calculating means (20) for calculating a total code amount (hereinafter, total code amount) of the difference value encoded data stored in (1), and a total code amount storing means ( 21) a frequency component storage means (2) for storing a coefficient B inside the basic coding means (1);
A coefficient reading control unit (3) for generating an address signal D of the coefficient B stored in the frequency component storage unit (2) and reading out the coefficient B from the frequency component storage unit (2); and a frequency component storage unit (2) And a coefficient encoding means (5) for encoding the coefficient B read out from the storage means. In the total code amount storage means (21), the quantization used in the quantization means (10) and the inverse quantization means (13) is performed. It has a function of storing a total code amount corresponding to information, and obtains a minimum value from a plurality of total code amount values stored in the total code amount storage means (21), so that the total code amount is minimized. A minimum code amount determining means (26) for determining quantization information (hereinafter, minimum quantization information); and a quantization information encoding means (2) for encoding the minimum quantization information and outputting encoded quantization information data.
7), and selects the compressed coded data (hereinafter referred to as selection data R) corresponding to the minimum quantization information and the difference value coded data (hereinafter referred to as selection data T) corresponding to the minimum quantization information. A configuration is provided that includes a selection output unit (28) that outputs data R, selection data T, and coded quantization information data as final compressed coded data.

According to the above-described configuration, the compression encoded data and the difference value encoded data are obtained while changing the quantization information (quantization step and bit shift amount), and the quantization that minimizes the total code amount of these is obtained. The information, the corresponding compressed coded data (the output of the basic coding means (1)) and the difference value coded data (the output of the difference value coding means (18)) are output as final compressed coded data. Thus, the input signal (image data or audio data) can always be compression-coded with a small code amount and without loss. Thus, the fourth problem can be solved.

According to the eighth aspect of the present invention, at least image data (or audio data) itself or data encoding the image data (or audio data) is used as the input signal W, and the input signal W is stored. Input signal storage means (29); coding conversion means (30) for coding (or converting the format) of the input signal W and outputting a coding conversion signal; and coding conversion signal for storing the coding conversion signal. Storage means (31); decoding inverse conversion means (32) for decoding the encoded conversion signal (or inverse format conversion) and outputting a decoded inverse conversion signal; decoding for storing the decoded inverse conversion signal. Inverse conversion signal storage means (33);
An error detecting means (34) for detecting an error in the decoding (or format inversion) processing of the encoded conversion signal in the decoding inverse conversion means (32) and outputting a first error signal indicating the presence or absence of the error; A comparison operation means for comparing the input signal W with the decoded inverse transform signal to obtain a change amount, setting the change amount exceeding a predetermined value as an error state, and outputting a second error signal indicating whether the state is an error state ( 35) outputs a signal indicating whether the first error signal or the second error signal is in an error state, and outputs the first error signal or the second error signal.
Selection output means for selecting and outputting the input signal W when the error signal is in the error state, and selecting and outputting the coded conversion signal when the first error signal and the second error signal are not in the error state (36).

With the above configuration, the selection output means (3
6) In the case where the first error signal or the second error signal is in the error state, the input signal W is selected and output. If the first error signal and the second error signal are not in the error state, the code is output. By selecting and outputting the conversion conversion signal,
Even if an instantaneous operation error occurs due to a physical problem of the signal processing system itself during encoding (or format conversion) of the input signal W, the original signal (input signal W) is output from the output signal.
Can be restored. Thus, the fifth problem can be solved.

According to a ninth aspect of the present invention, in the signal processing system according to the fifth aspect, the read control information generating means (4) is configured to input a coefficient A (or a coefficient B). In the means (4),
Scan order determining means (401) for determining the scan order of each frequency component in the frequency component storage means (2) from the input coefficient A (or coefficient B), and the scan order determining means (401) A scan order table (402) for storing scan order information (hereinafter referred to as scan order information E); and a scan order information encoding unit (403) for encoding the scan order information E. , And the coefficient read control means (3) generates the address signal D based on at least the read control information C.

The above-described configuration is a combination of the first and fifth aspects, and can achieve the effects obtained in the first and fifth aspects simultaneously. For this reason, the amount of information can be reduced more than the configuration of claim 1 alone or claim 5 alone.

According to a tenth aspect of the present invention, in the signal processing system according to the first aspect, the coefficient A (or the coefficient B) is provided inside the scan order determining means (401) with respect to the coefficient A (or coefficient B) for each frequency component. (Or non-zero coefficient)
Zero coefficient counting means (410) for counting the number of
And the number of zero coefficients for each frequency component tabulated by the zero coefficient counting means (410) are compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized. And a priority order determining means (411) for ranking.

According to the above configuration, the coefficient A (or coefficient B) of the frequency component having a small number of zero coefficients (or a large number of non-zero coefficients) is preferentially scanned (read out) and sampled. Compared with the related art using the scan order table, the number of zero coefficients to be encoded can be reduced, and the amount of information can be greatly reduced as a whole.

According to an eleventh aspect of the present invention, in the signal processing system according to the fourth aspect, the coefficient A (or the coefficient B) is provided inside the scan order determining means (401) with respect to the coefficient A (or coefficient B) for each frequency component. (Or non-zero coefficient)
Zero coefficient counting means (410) for counting the number of
And the number of zero coefficients for each frequency component tabulated by the zero coefficient counting means (410) are compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized. And a priority order determining means (411) for ranking.

With the above configuration, in addition to the effect of the fourth aspect alone, the coefficient A (or coefficient B) of the frequency component having a small number of zero coefficients (or a large number of non-zero coefficients) is preferentially scanned. Since sampling is performed by reading (reading), the number of zero coefficients to be encoded can be reduced as compared with the conventional technique using a fixed scan order table, and the amount of information can be greatly reduced as a whole.

According to a twelfth aspect of the present invention, in the signal processing system according to the ninth aspect, the coefficient A (or coefficient B) has a zero coefficient for each frequency component in the scan order determining means (401). (Or non-zero coefficient)
Zero coefficient counting means (410) for counting the number of
And the number of zero coefficients for each frequency component tabulated by the zero coefficient counting means (410) are compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized. And a priority order determining means (411) for ranking.

With the above arrangement, in addition to the effect of the ninth aspect alone, the coefficient A (or coefficient B) of the frequency component having a small number of zero coefficients (or a large number of non-zero coefficients) is preferentially scanned. Since sampling is performed by reading (reading), the number of zero coefficients to be encoded can be reduced as compared with the conventional technique using a fixed scan order table, and the amount of information can be greatly reduced as a whole.

According to a thirteenth aspect of the present invention, in the signal processing system according to the first or the tenth aspect, each of the coefficients A (or the coefficients B) is provided inside the scan order determining means (401) with respect to the coefficient A (or the coefficient B). A coefficient strength calculating means (412) for obtaining a coefficient strength for each component; and a coefficient strength value for each frequency component by comparing at least a coefficient strength value for each frequency component with an input. Has priority order determining means (411) for prioritizing the scanning order of frequency components having a large value of, and the number of zero coefficients for each frequency component is input to the priority order determining means (411). First, the intensity values of the coefficients are compared, and the scanning order of the frequency components having a large coefficient intensity is prioritized in order, and the frequency components having the same magnitude of the coefficient intensity are duplicated. If present, the number of zero coefficients (or non-zero coefficients) is compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is preferentially ranked. I do.

According to the above-described configuration, in addition to the effects of the first aspect alone and the tenth aspect alone, the coefficient A (or coefficient B) of the frequency component having a large coefficient value is preferentially scanned (read and read). 2.) Since sampling is performed, the number of zero coefficients to be encoded can be reduced as compared with the conventional technique using a fixed scan order table, and coefficients having a large intensity value are collected in the first half (in the scan order). ) Can greatly reduce the amount of information as a whole.

According to a fourteenth aspect of the present invention, in the signal processing system according to the fourth or eleventh aspect, the scan order determining means (401) includes, for each of the coefficients A (or coefficients B), A coefficient strength calculating means (412) for obtaining a coefficient strength for each component; and a coefficient strength value for each frequency component by comparing at least a coefficient strength value for each frequency component with an input. Has priority order determining means (411) for prioritizing the scanning order of frequency components having a large value of, and the number of zero coefficients for each frequency component is input to the priority order determining means (411). First, the intensity values of the coefficients are compared, and the scanning order of the frequency components having a large coefficient intensity is prioritized in order, and the frequency components having the same magnitude of the coefficient intensity are duplicated. If present, the number of zero coefficients (or non-zero coefficients) is compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is preferentially ranked. I do.

With the above-described configuration, in addition to the effects of the fourth aspect and the eleventh aspect alone, the coefficient A (or coefficient B) of the frequency component having a large coefficient intensity value is preferentially scanned (read and read). 2.) Since sampling is performed, the number of zero coefficients to be encoded can be reduced as compared with the conventional technique using a fixed scan order table, and coefficients having a large intensity value are collected in the first half (in the scan order). ) Can greatly reduce the amount of information as a whole.

According to a fifteenth aspect of the present invention, in the signal processing system according to the ninth or twelfth aspect, the scan order determination means (401) includes, within the scan order determining means (401), a coefficient A (or a coefficient B) for each frequency. A coefficient strength calculating means (412) for obtaining a coefficient strength for each component; and a coefficient strength value for each frequency component by comparing at least a coefficient strength value for each frequency component with an input. Has priority order determining means (411) for prioritizing the scanning order of frequency components having a large value of, and the number of zero coefficients for each frequency component is input to the priority order determining means (411). First, the intensity values of the coefficients are compared, and the scanning order of the frequency components having a large coefficient intensity value is prioritized, and the frequency components having the same magnitude value of the coefficient are duplicated. If present, the number of zero coefficients (or non-zero coefficients) is compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is preferentially ranked. I do.

With the above configuration, in addition to the effects of the ninth and twelfth aspects alone, the coefficient A (or coefficient B) of the frequency component having a large coefficient intensity value is preferentially scanned (read and read). 2.) Since sampling is performed, the number of zero coefficients to be encoded can be reduced as compared with the conventional technique using a fixed scan order table, and coefficients having a large intensity value are collected in the first half (in the scan order). Can greatly reduce the amount of information as a whole.

According to a sixteenth aspect of the present invention, in the signal processing system according to the first, tenth, or thirteenth aspect, the read control information generating means (4) includes:
The configuration is such that complexity information F 2 of image data (or audio data) is inputted, and each frequency in the frequency component storage means (2) is read from the inputted complexity information F inside the read control information generation means (4). A complexity group classifier (40) for classifying components into a plurality of complexity groups
4) a complexity group information storage means (405) for storing the complexity group information G classified by the complexity group classification means (404); and a complexity group information encoding for encoding the complexity group information G. Means (40
6), and at least the complexity group information G is used as the read control information C, and the coefficient read control means (3) generates the address signal D based on at least the read control information C.

According to the above-described structure, in addition to the effects of the first aspect alone, the tenth aspect alone, and the thirteenth aspect alone, the complexity group classifying means (404) allows the input signal (image data or audio data) or The coefficient A (or coefficient B) after frequency decomposition of the signal is classified into a plurality of complexity groups, and the frequency component storage means (2) for each complexity group
, A coefficient A (or coefficient B) is read out, and the coefficient encoding means (5) performs statistical processing on the read coefficient A (or coefficient B) to obtain a coefficient A (or coefficient B).
Is encoded, a certain complexity group collects values having a large amplitude absolute value (amp), and a certain complexity group collects values having a small amplitude absolute value (amp). Each of the groups can be biased, and the amount of information in each group can be reduced.

The invention according to claim 17 is the invention according to claim 1, 10, 10, 13, or 16.
In the signal processing system described in (1), inside the readout control information generating means (4), it is determined from which value of the coefficient A (or coefficient B) the frequency component is biased to which frequency component, and the information indicating the bias status is determined. Coefficient distribution determining means (407) for outputting coefficient distribution information (hereinafter referred to as coefficient distribution information); coefficient distribution information storage means (408) for storing coefficient distribution information; and coefficient distribution information encoding means (40) for encoding coefficient distribution information.
9), and in the scan order determining means (401),
The scan order is determined for each coefficient distribution information, and the scan order information E for each coefficient distribution information is stored in the scan order table (402).

According to the above-described configuration, in addition to the effects of the first aspect alone, the tenth aspect alone, the 13th aspect alone, and the 16th aspect alone, each of the coefficients A in the optimum scanning order according to the coefficient distribution.
(Or coefficient B) to read and sample
Compared to the conventional technique using a fixed scan order table, encoding suitable for an input signal can be performed, and the amount of information can be reduced overall.

The invention according to claim 18 is the invention according to claim 4, claim 9, or claim 11, or claim 12,
Alternatively, in the signal processing system according to claim 14 or 15, read control information generating means (4).
Coefficient distribution judging means (407) for judging to which frequency component the energy is biased from the value of the coefficient A (or coefficient B), and outputting information (hereinafter, coefficient distribution information) indicating the bias status. ), Coefficient distribution information storage means (408) for storing coefficient distribution information, and coefficient distribution information encoding means (409) for encoding coefficient distribution information. The scan order is determined for each piece of information, and the scan order table (40
In 2), scan order information E for each coefficient distribution information is stored.

According to the above-described structure, only the fourth aspect, the ninth aspect alone, the eleventh aspect alone, the twelfth aspect alone, and the first aspect are provided.
4 alone and in addition to the effects of claim 15 alone, each coefficient A (or coefficient B) is read out and sampled in the optimal scan order according to the coefficient distribution, so that the input is smaller than in the prior art using a fixed scan order table. Encoding suitable for a signal can be performed, and the amount of information can be reduced as a whole.

The present invention as defined in claim 19, according to claim 3, or claim 4, or claim 5, or claim 9, or claim 11, or claim 12, or claim 14,
19. The signal processing system according to claim 15, wherein a unit of the frequency decomposition among the coefficients of each frequency component in the frequency component storage means (2) is provided inside the sine information processing means (7). A coefficient strength rank determining means (704) for measuring the strength of the energy of each coefficient (hereinafter referred to as "strength") and determining the rank of the strength of each coefficient (hereinafter referred to as "coefficient strength rank"); Coefficient information ranking table (705) and the coefficient with the highest coefficient strength ranking are preferentially used to estimate the sign information.
It has a signature information estimating means (706) for generating and outputting new signature information. The non-zero determination signature information extracting means (701) determines whether each coefficient is non-zero. Signature information estimating means (706)
It is configured to execute the processing inside.

With the above-described configuration, the third aspect alone, the fourth aspect alone, the fifth aspect alone, the nineteenth aspect alone, the eleventh aspect solely, the twelfth aspect solely, the fourteenth aspect alone, and the fifteenth aspect solely claim In addition to the effect of 18 alone, sign information is preferentially estimated from the coefficient with the highest coefficient strength rank, and new sign information is generated and encoded from the estimation result.
The encoding efficiency of the sign information is improved.

According to a twentieth aspect of the present invention, in the signal processing system according to the nineteenth aspect, the sign information processing means (7) includes a frequency component storage means (2) inside the signature information estimating means (706). Temporary signature information (hereinafter, temporary signature information) for the coefficients of each frequency component in
Sign information setting means (707) for setting true sign information (hereinafter, true sign information) as necessary; and temporary sign information by sign information setting means (707).
Each coefficient for which true sign information is set as necessary
A sign information-added coefficient storage means (708) for storing the sign information-added coefficient), a frequency component synthesizing means (709) for synthesizing each frequency component of the sign information-added coefficient and outputting a frequency synthesis result with the sign information, Frequency synthesis result storage means (7) for storing a frequency synthesis result with information
10) a first signature information estimation parameter calculating means for calculating a first signature information estimation parameter (hereinafter, a first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. (71
1) a first signature information estimation parameter storage means (712) for storing a first signature information estimation parameter;
From the value of the first signature information estimation parameter corresponding to each temporary signature information, it is estimated which temporary signature information is the true signature information, and a match indicating whether the estimated signature information matches the true signature information or not. First signature information generating means (713) for outputting the mismatch information as new signature information
The signature information setting means (707) has a function of setting true signature information for a coefficient for which estimation of whether or not the temporary signature information is true signature information has been completed. Frequency synthesis result (hereinafter referred to as true sign frequency synthesis result) using a coefficient (hereinafter referred to as a coefficient with true signature information) and frequency synthesis result (hereinafter referred to as a coefficient with temporary signature information) with provisional signature information set. , The sum of the absolute values of the differences from the temporary signature frequency synthesis result) is used as the first change amount, and the frequency synthesis result (hereinafter, referred to as “the true signature information-added coefficient” and the temporary signature information-added coefficient) is mixed.
With respect to the mixed frequency synthesis result), the sum of the absolute value of the difference between the central axis of the waveform and the central axis of the waveform in the region surrounded by the waveform in the waveform of the mixed frequency synthesis result is the first value.
The first sign information estimation parameter calculating means (711) calculates the first change amount or the first similarity as the first signature information estimation parameter, and the first signature information generation means (713) calculates the similarity. When the first signature information estimation parameter corresponding to the temporary signature information is the first change amount, the temporary signature information when the value of the first signature information estimation parameter is small is estimated to be true signature information, and each temporary signature information is estimated. When the first signature information estimation parameter corresponding to the first signature information estimation parameter is the first similarity, the temporary signature information when the value of the first signature information estimation parameter is large is estimated to be the true signature information.

With the above arrangement, in addition to the effect of the nineteenth aspect alone, the frequency component synthesizing means (709) is used to estimate the sign information preferentially from the coefficient having the highest coefficient strength order. Therefore, the function can be realized in common with the conventional decoding means (decoder).

According to a twenty-first aspect of the present invention, in the signal processing system according to the nineteenth aspect, the sign information processing means (7) includes a frequency component storage means (2) inside the signature information estimating means (706). Temporary signature information (hereinafter, temporary signature information) for the coefficient of each frequency component in
Sign information setting means (707) for setting true sign information (hereinafter, true sign information) as necessary; and temporary sign information by sign information setting means (707).
Each coefficient for which true sign information is set as necessary
A sign information-added coefficient storage means (708) for storing the sign information-added coefficient), a frequency component synthesizing means (709) for synthesizing each frequency component of the sign information-added coefficient and outputting a frequency synthesis result with the sign information, Frequency synthesis result storage means (7) for storing a frequency synthesis result with information
10) a first signature information estimation parameter calculating means for calculating a first signature information estimation parameter (hereinafter, a first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. (71
1) a first signature information estimation parameter storage means (712) for storing a first signature information estimation parameter;
A second signature information estimation parameter calculating means (714) for calculating a second signature information estimation parameter (hereinafter, a second signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set; ,
A second signature information estimation parameter storage means (715) for storing the second signature information estimation parameter; and signature information in which each temporary signature information is set, using the first signature information estimation parameter and the second signature information estimation parameter. A third signature information estimation parameter calculating means (716) for calculating a third signature information estimation parameter (hereinafter, a third signature information estimation parameter) corresponding to the frequency-combined result;
A third signature information estimation parameter storage means (717) for storing the third signature information estimation parameter; and a value of the third signature information estimation parameter corresponding to each temporary signature information, to determine which temporary signature information is true signature information. Second signature information generating means (718) for outputting match / mismatch information indicating whether the estimated signature information matches the true signature information as new signature information. The setting means (707) has a function of setting true signature information for a coefficient for which estimation of whether or not the temporary signature information is true signature information has been completed. The result of frequency synthesis based on the coefficient with signature information (hereinafter referred to as the true signature frequency synthesis result) and the coefficient for which temporary signature information is set (hereinafter referred to as the temporary signature information The sum of the absolute value of the difference from the frequency synthesis result (hereinafter referred to as “temporary sign frequency synthesis result”) is used as the first variation, and the frequency synthesis result in a mixed state of the coefficient with the true sign information and the coefficient with the temporary sign information Regarding the mixed frequency synthesis result (hereinafter referred to as “mixed frequency synthesis result”), the sum of the absolute values of the differences between the center axis of the waveform and the center axis of the waveform in the region surrounded by the waveform in the waveform of the mixed frequency synthesis result is the first. As the similarity, the sum of the absolute values of the differences between the waveform and the value range in the region where the waveform of the mixed frequency synthesis result protrudes beyond the predetermined value range is set as the difference, and the first sign information estimation parameter calculating means (711) The first change amount or the first similarity is calculated as a first signature information estimation parameter, and the second signature information estimation parameter calculation means (714) calculates the difference between the first signature information estimation parameter and the second signature information estimation parameter. The first change amount or difference is normalized by multiplying it by a normalization coefficient, and a value obtained by adding the normalized difference to the normalized first change is defined as a second change. The similarity or dissimilarity is normalized by a normalization coefficient, and a value obtained by subtracting the normalized dissimilarity from the normalized first similarity is set as a second similarity. 716)
, The second change amount or the second similarity is calculated as a third signature information estimation parameter, and the second signature information generation means (718) calculates the third signature information estimation parameter corresponding to each temporary signature information into the second signature information estimation parameter. In the case of the change amount, the temporary signature information when the value of the third signature information estimation parameter is small is estimated as true signature information, and the third signature information estimation parameter corresponding to each temporary signature information is the second similarity. In this case, the temporary signature information when the value of the third signature information estimation parameter is large is estimated to be true signature information.

According to the above configuration, in addition to the effect of the nineteenth aspect alone, the frequency component synthesizing unit (709) is used to estimate the sign information preferentially from the coefficient having the highest coefficient strength rank. Therefore, the function can be realized in common with the conventional decoding means (decoder).
In addition, by using the degree of difference, it is possible to reduce erroneous estimation of the signature information when the waveform of the mixed frequency synthesis result exceeds a predetermined value range. Therefore, the probability that the estimation result is correct is increased, and the coding efficiency can be increased.

According to a twenty-second aspect of the present invention, in the signal processing system according to the nineteenth aspect, the sign information processing means (7) includes a frequency component storage means (2) inside the signature information estimating means (706). Temporary signature information (hereinafter, temporary signature information) for the coefficients of each frequency component in
Sign information setting means (707) for setting true sign information (hereinafter, true sign information) as necessary; and temporary sign information by sign information setting means (707).
Each coefficient for which true sign information is set as necessary
A sign information-added coefficient storage means (708) for storing the sign information-added coefficient), a frequency component synthesizing means (709) for synthesizing each frequency component of the sign information-added coefficient and outputting a frequency synthesis result with the sign information, Frequency synthesis result storage means (7) for storing a frequency synthesis result with information
10) a first signature information estimation parameter calculating means for calculating a first signature information estimation parameter (hereinafter, a first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. (71
1) a first signature information estimation parameter storage means (712) for storing a first signature information estimation parameter;
A second signature information estimation parameter calculating means (714) for calculating a second signature information estimation parameter (hereinafter, a second signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set; ,
A second signature information estimation parameter storage means (715) for storing the second signature information estimation parameter; and signature information in which each temporary signature information is set, using the first signature information estimation parameter and the second signature information estimation parameter. A third signature information estimation parameter calculating means (716) for calculating a third signature information estimation parameter (hereinafter, a third signature information estimation parameter) corresponding to the frequency-combined result;
A third signature information estimation parameter storage means (717) for storing the third signature information estimation parameter; and a value of the third signature information estimation parameter corresponding to each provisional signature information, which matches the combination of the true signature information. A third signature information generating unit (719) for adding new signature information to the combination of the temporary signature information and outputting new signature information; and a signature information setting unit (707).
Has a function of setting true signature information for a coefficient for which estimation of whether or not the temporary signature information is true signature information is performed, and a coefficient in which the true signature information is set (hereinafter, a coefficient with true signature information) The difference between the result of frequency synthesis (hereinafter referred to as true sign frequency synthesis result) and the result of frequency synthesis based on the coefficient for which temporary sign information is set (hereinafter referred to as a coefficient with temporary sign information) The sum of the absolute values is defined as a first change amount, and regarding the frequency synthesis result in the mixed state of the coefficient with the true signature information and the coefficient with the temporary signature information (hereinafter, the mixed frequency synthesis result), the waveform of the mixed frequency synthesis result The sum of the absolute value of the difference between the center axis of the waveform and the center axis of the waveform in the region surrounded by the waveform is defined as the first similarity, and the waveform of the mixed frequency synthesis result falls within a predetermined value range. The sum of the absolute value of the difference between the waveform and the value range in the region to be output is defined as the degree of difference, and the first sign information estimation parameter calculating means (711) calculates the first amount of change or the first similarity as the first sign information estimation. A second signature information estimation parameter calculation means (714) calculates the difference as a second signature information estimation parameter, normalizes the first variation or the difference by a normalization coefficient, and normalizes the difference. A value obtained by adding the normalized difference to the already-processed first change is defined as a second change, and the first similarity or the difference is normalized by a normalization coefficient,
The value obtained by subtracting the normalized dissimilarity from the normalized first similarity is defined as the second similarity, and the third sign information estimation parameter calculating means (716) calculates the second variation or the second similarity. The third signature information estimation parameter is calculated as the third signature information estimation parameter. If the third signature information estimation parameter corresponding to each temporary signature information is the second change amount, the third signature information estimation means (719) calculates the third signature information estimation parameter. A combination of temporary signature information having a small value is preferentially set as a code including a large number of identical symbols (hereinafter, a first code), or a combination of temporary signature information having a small value of a third signature information estimation parameter is preferentially given. The first code corresponding to the combination of the temporary signature information that matches the combination of the true signature information is referred to as a short code (hereinafter, a second code).
The code or the second code is output as new signature information, and when the third signature information estimation parameter corresponding to each temporary signature information is the second similarity, the temporary signature having a larger third signature information estimation parameter value is output. The combination of the information is preferentially set as the first code, or the combination of the temporary sign information having a large value of the third sign information estimation parameter is set preferentially as the second code, and matches the combination of the true sign information. The first code or the second code corresponding to the combination of the temporary signature information to be output is output as new signature information.

With the above arrangement, in addition to the effect of the nineteenth aspect, the frequency component synthesizing means (709) is used to preferentially estimate the sign information from the coefficient having the highest coefficient strength order. Therefore, the function can be realized in common with the conventional decoding means (decoder).
In addition, by using the degree of difference, it is possible to reduce erroneous estimation of the signature information when the waveform of the mixed frequency synthesis result exceeds a predetermined value range. Therefore, the probability that the estimation result is correct is increased, and the coding efficiency can be increased. Furthermore, since new signature information is generated from a combination of temporary signature information of a plurality of coefficients, the amount of information can be reduced as compared with processing signature information of a single coefficient.

The invention according to claim 23 is the invention according to claim 20,
Alternatively, in the signal processing system according to claim 21 or 22, an inverse quantization means (720) for performing inverse quantization on the coefficient B or the coefficient with signature information in the signature information estimation means (706). ), And input at least the inversely quantized coefficient with sign information to the frequency component synthesizing means (709) to synthesize each frequency component.

According to the above-described configuration, in addition to the effects of the twentieth aspect, the twenty-first aspect, and the twenty-second aspect alone, the sign information using the sign information-added coefficient inversely quantized by the inverse quantization means (720). Since the estimation is performed, it is hardly affected by an error at the time of quantization, the signature information can be estimated with higher accuracy, and the encoding efficiency can be improved.

The invention according to claim 24 is the invention according to claim 20,
Or claim 21, or claim 22, or claim 2
3. The signal processing system according to item 3, wherein the frequency component synthesizing means (709) calculates with high bit accuracy.

With the above arrangement, in addition to the effects of the twentieth aspect alone, the twenty-first aspect alone, the twenty-second aspect alone, and the thirty-third aspect alone, it is less susceptible to an error at the time of quantization, and the sign information can be obtained with higher accuracy. Can be estimated, and the coding efficiency can be improved.

The invention according to claim 25 is the invention according to claim 1, 2, 3, or 4, or 5, or 6, or 9, or 10. , Or claim 11, or claim 12, or claim 13, or claim 14, or claim 1.
A signal according to claim 5, or claim 16, or claim 17, or claim 18, or claim 19, or claim 20, or claim 21, or claim 22, or claim 23, or claim 24. In the processing system, the coefficient encoding means (5) classifies the inputted coefficients into a plurality of groups, and performs statistical processing and encoding processing for each group.

According to the above-mentioned structure, the present invention can be applied to the first aspect, the second aspect, the third aspect, the fourth aspect, the fifth aspect, the fifth aspect, the sixth aspect, the ninth aspect, the tenth aspect, and the independent aspect. 11 alone, claim 12 alone, claim 13 alone, claim 14 alone, claim 15 alone, claim 16 alone, claim 17 alone, claim 18 alone, claim 19 alone, claim 2
0 alone, claim 21 alone, claim 22 alone, claim 23
In addition to the effect of the group alone, the encoding efficiency can be improved when the magnitude of the coefficient amplitude (or the absolute value of the amplitude) is biased depending on each group.

The invention according to claim 26 is the invention according to claim 3, 4, 5, or 9, or 11, or 12, or 14,
Or claim 15, or claim 18, or claim 1
In the signal processing system according to claim 9, or 20, or 21, or 22, or 23 or 24, the sign information encoding unit (7
03) is characterized by classifying each input coefficient into a plurality of groups, and performing statistical processing and encoding processing for each group.

With the above-described configuration, the third aspect alone, the fourth aspect alone, the fifth aspect alone, the nineteenth aspect alone, the eleventh aspect alone, the twenty-second aspect alone, the fifteenth aspect alone, and the fifth aspect In addition to the effects of 18 alone, claim 19 alone, claim 20 alone, claim 21 alone, claim 22 alone, claim 23 alone, and claim 24 alone, each group has a sign information of zero (or 1). For example, when the value of ()) is unevenly (concentrated), the coding efficiency can be increased.

According to a twenty-seventh aspect of the present invention, in the signal processing system according to the sixth aspect, the frequency decomposition means (9)
And a quantizing means (10) for quantizing the coefficient A, which is the output of, and outputting it as a coefficient B, and storing the coefficient B in the frequency component storage means (2).

With the above configuration, the amount of code can be further reduced by quantizing the coefficient A.

According to a twenty-eighth aspect of the present invention, in the signal processing system according to the sixth or seventh aspect, the basic encoding means (1) has the first, second, or third aspect. Or claim 4, or claim 5, or claim 9, or claim 10, or claim 11, or claim 12, or claim 13, or claim 14,
Or claim 15, or claim 16, or claim 1
7, or claim 18, or claim 19, or claim 20, or claim 21, or claim 22, or claim 23, or claim 24, or claim 25, or claim 26, or claim 27, which has the same configuration as that described above.

According to the above configuration, claim 1, claim 2, or claim 3, or claim 4, or claim 5, or claim 9, or claim 10, or claim 11, or claim 12, or claim 13, or claim 14, or claim 15, or claim 16, or claim 17, or claim 18, or claim 1
9, or 21, or 22, or 23, or 24 or 25, 26, or 27, and further, claim 6. The effect of reducing the code amount by obtaining the adjacent sample difference value described in (1) and the effect of enabling lossless compression with the minimum code amount by the structure of claim 7 are obtained.

[0109]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS <First Embodiment> A first embodiment of the present invention for solving the first problem of the conventional signal processing system will be described below.

The first problem of the conventional signal processing system is that the two-dimensional state data is stored in accordance with the fixed zigzag scan order stored in advance in the fixed scan order table (450 in FIG. 81). One-dimensional arrangement (S004 in FIG. 82) is not necessarily an effective compression means. In the first embodiment of the present invention, an original scan order table suitable for each input signal (in this example, image data) is created, and compression encoding is performed using the original scan order table. This is to solve the first problem of the conventional signal processing system described above.

FIG. 1 shows the configuration of a signal processing system (compression encoding side) according to the first embodiment of the present invention. In FIG. 1, reference numeral 9 denotes frequency decomposition means for decomposing image data as an input signal into a plurality of frequency components, and reference numeral 10 denotes quantization means for quantizing data after frequency decomposition (hereinafter, referred to as coefficient A). Reference numeral 1 denotes a basic encoding unit which receives the quantized data of the coefficient A (hereinafter, referred to as a coefficient B) and outputs compressed encoded data, and 2 denotes the inside of the basic encoding unit (1). And frequency component storage means for storing the quantized data (coefficient B) of the coefficient A. Reference numeral 3 denotes a coefficient read control for giving an address signal D to the frequency component storage means (2) and reading out the stored coefficient B. Means, 4 is the coefficient B
Read control information generating means for generating read control information C for reading coefficient A (or coefficient B) stored in frequency component storage means (2)
5 is a coefficient encoding means for encoding the coefficient B read from the frequency component storage means (2). Inside the read control information generating means (4), 401 is a frequency component from the input coefficient B. Scan order determining means for determining the scan order of each frequency component in the storage means (2)
Reference numeral 402 denotes information on the scan order determined by the scan order determination means (401) (hereinafter, scan order information E).
403 is a scan order information encoding unit that encodes the scan order information E.
In the scan order determining means (401), 4
Numeral 10 denotes a zero coefficient counting means for counting the number of zero coefficients (or non-zero coefficients) for each frequency component with respect to the coefficient A (or coefficient B). Numeral 411 denotes each of the numbers counted by the zero coefficient counting means (410). Priority order determining means for comparing the number of zero coefficients for each frequency component and prioritizing the scan order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients); E is used as read control information C, and the coefficient read control means (3) generates an address signal D based on the read control information C.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT and the coefficient coding means (5) performs Huffman coding.

Hereinafter, the operation of the signal processing system (compression encoding side) according to the first embodiment of the present invention will be described with reference to FIGS.
This will be described with reference to FIG. First, processing for creating an original scan order table suitable for an original image frame to be subjected to compression encoding is performed.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(The original image block) is input (S00 in FIG. 2).
1).

<DCT and Quantization> Next, the frequency decomposition means (9 in FIG. 1) performs a DCT operation as shown in Equation 1 on the input original image block (S0 in FIG. 2).
02, S100 in FIG. 3, and the quantizing means (10 in FIG. 1) quantizes each coefficient (DCT block) after DCT (S003 in FIG. 2, S101 in FIG. 3). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 1) (S004 in FIG. 2, S102 in FIG. 3).
Here, as an example, any coefficient is quantized by applying 1 as a uniform quantization step. This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Counting of Number of Zero Coefficients> Here, the zero coefficient counting means (410 in FIG. 1)
It has a zero coefficient counter corresponding to each component of the T block, and when the coefficient of each component is zero, the zero coefficient counter is incremented and the number of zero coefficients for each component is counted (counted). 2 S005, FIG. 3 S
103). For example, assume that there are two DCT blocks, DCT block 2 and DCT block 3, as shown in FIGS. In these DCT blocks, the number of zeros of the coefficient of each AC component of the coordinates (n, m) is counted. The coefficient of the coordinates (1, 0) in the DCT block 2 is 5 (non-zero), and the number of zeros in the coordinates (1, 0) is 0. The coefficient of the coordinates (1, 0) in the DCT block 3 is 5
(Non-zero), and the number of zeros in the coordinates (1, 0) of the DCT block 2 and the DCT block 3 is 0 in total. Thus, DCT block 2 and DCT block 3
The result of counting the number of zeros at each coordinate (n, m) is as shown in FIG.

The above series of processing (S001 and S0 in FIG. 2)
02, S003, S004, and S005, S100 in FIG.
, S101, S102, and S103) have been completed for all the original image blocks (S006 in FIG. 2).
This processing is repeated until the processing of all original image blocks is completed.

<Creation of Scan Order Table> For example, it is assumed that all original image blocks are only DCT block 2 and DCT block 3 shown in FIG. 4 and FIG. In that case, the result of counting the number of zeros at each coordinate (n, m) (ie, zero coefficient counting means (410) in FIG. 1)
Is as shown in FIG. With respect to the value (FIG. 6) of the zero coefficient counting means (410 in FIG. 1), the priority order determining means (411 in FIG. 1) determines a scanning order in which sampling is performed from the coefficient of a coordinate having a small number of zeros. Then, it is stored in the scan order table (402 in FIG. 1) (S007 in FIG. 2). In this case, the scan order table is as shown in FIG. Although there are a number of frequency components having the same number of zeros, how to determine the priority order in this case is not limited.

Next, a scan order table (40 in FIG. 1)
Scan order information (eg, coordinate values) stored in 2) is encoded by a scan order information encoding unit (403 in FIG. 1) and output as an output signal (S in FIG. 2).
008). In the encoding of the scan order information performed here, a value obtained by simply expressing a coordinate value in a binary number may be output, or a non-zero coefficient hardly appears in a low frequency component in image data or audio data. Alternatively, a method of outputting a variable-length code such that a code becomes shorter as the coordinate value is closer to the low-frequency component may be used.

<Execution and Encoding of Scan> Next, in accordance with the scan order stored in the created scan order table (402 in FIG. 1), the readout control means (3 in FIG. 1) uses the frequency component storage means (3 in FIG. 1). The address signal D shown in 2) of FIG. 1 is generated.
Each coefficient of the quantized DCT block is read up to the last non-zero coefficient (S009 in FIG. 2). As an example, using the scan order table shown in FIG.
FIG. 8 shows a state in which the CT block 3 is scanned.
FIG. As a result, DCT block 2 (FIG. 8)
Has one zero coefficient and a total of 13 other non-zero coefficients, as shown in Table 1. DCT block 3 (FIG. 9)
There are two zero coefficients and a total of 13 other non-zero coefficients, as shown in Table 2. These are encoded by coefficient encoding means (5 in FIG. 1) (S010 in FIG. 2).
This process is performed for all DCT blocks (S009, S010, S011 in FIG. 2). At this time, a conventional Huffman code may be applied as the encoding method, or the encoding may be performed after an optimal code is derived after performing statistical processing.

[0121]

[Table 5]

[0122]

[Table 6]

Here, when the amount of information (excluding EOB) is calculated according to Equation 3, the DCT block 2 (FIG. 8) is approximately 33.8 [bit] and the DCT block 3 (FIG. 9)
Is about 37.1 [bits], and the fixed scan table (450 in FIG. 81) shown in the first problem of the related art is used.
In the example of scanning in the zigzag scan order, the information amount of the DCT block 2 (FIG. 90) and the information amount of the DCT block 3 (FIG. 91) are both about 56.5 [bit]. Therefore, in the present embodiment (the first embodiment of the present invention), the coordinates (n, m)
By counting the number of zero coefficients in, and scanning from the coefficient at the position where the zero coefficient is small, the amount of information can be reduced.

In the present embodiment (first embodiment of the present invention), information (eg, coordinate values) of the scan order is used.
Is encoded by the scan order information encoding means (403 in FIG. 1) and output as an output signal (S00 in FIG. 2).
8) Therefore, the amount of information becomes overhead,
For this overhead, the amount of information that can be reduced from each DCT block only needs to be larger, and there are many DCT blocks in one frame as a whole (for example, 5
(There are 400 DCT blocks), the overhead can be negligibly small on average per block.

<Second Embodiment> Next, a second embodiment of the present invention for solving the first problem of the conventional signal processing system will be described.

In the first embodiment of the present invention, the “number of zero coefficients” of each coordinate (n, m) of the DCT block after frequency decomposition and quantization is counted, and the coefficient of a coordinate having a small zero coefficient is counted. , The amount of information is reduced by scanning from. However, in the second embodiment of the present invention described below, the frequency decomposition and the quantization of the DCT block after each quantization (n, m) of " By obtaining the coefficient intensity and scanning from the coefficient at the coordinate having the larger intensity value, it is possible to reduce the amount of information more than in the first embodiment of the present invention.

The term “coefficient intensity” used in the present embodiment means the intensity of the energy of the frequency component. For example, the absolute value of the coefficient of the coordinates (n, m) (or the coefficient 2) ), The sum of the absolute values of the coefficients of the same coordinates (n, m) in a plurality of DCT blocks (or the square value of the coefficients), or the same coordinates (n, m) in a plurality of DCT blocks. m) is the number of absolute values of the coefficient (or the square value of the coefficient). In the following, as an example, the strength of the coefficient will be described as the sum of the absolute values of the coefficients at the same coordinates (n, m) in a plurality of DCT blocks.

FIG. 10 shows the configuration of a signal processing system (compression encoding side) according to the second embodiment of the present invention. FIG.
, 9 is a frequency decomposition means, 10 is a quantization means, 1 is a basic coding means, and 2 is a frequency component storage means in the basic coding means (1).
Is a coefficient readout control means, 4 is a readout control information generating means, and 401 is a scan order determining means inside the readout control information generating means (4).
2 is a scan order table. 403 is a scan order information encoding unit. The scan order information E is used as read control information C. The coefficient read control unit (3) uses the scan order information E based on the read control information C. A configuration for generating an address signal D is employed. The above is the first of the present invention described above.
The same operation as in the embodiment of FIG. The difference from the first embodiment of the present invention is that, within the scan order determining means (401), for the coefficient A (or coefficient B), the coefficient strength calculating means (412) for obtaining the strength of the coefficient for each frequency component. ) And priority order determining means (411), wherein the priority order determining means (411) receives at least the value of the coefficient strength for each frequency component and inputs the value of the coefficient strength for each frequency component. In comparison, priority is given to prioritizing the scanning order of frequency components having a large coefficient intensity value, and the coefficient encoding means (5) divides the coefficients into a plurality of groups and performs encoding processing for each group. The point is to do.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (compression encoding side) according to the second embodiment of the present invention will be described below with reference to FIG. First, processing for creating an original scan table suitable for an original image frame to be subjected to compression encoding is performed.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 11).
01).

<DCT and Quantization> Next, in the frequency decomposition means (9 in FIG. 10), the input original image block is subjected to a DCT operation as shown in Equation 1 (S002 in FIG. 11), and the quantization is performed. In the conversion means (10 in FIG. 10), D
Each coefficient (DCT block) after CT is quantized (FIG. 1).
1 (S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 10) (FIG. 11).
S004). Here, as an example, all coefficients are
Uniform quantization is performed by applying 1 as a quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 10) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S005 in FIG. 11). As described above, in the present embodiment (the second embodiment of the present invention), as an example, the intensity of the coefficient is
The sum of absolute values of coefficients at the same coordinates (n, m) in a plurality of DCT blocks. For example, as shown in FIGS. 4 and 5, two DCT blocks of DCT block 2 and DCT block 3 are used.
Assume that there is a T block. These DCT blocks are the same as those used in the description of the first embodiment of the present invention. In these DCT blocks, the coordinates (n,
The intensity of the coefficient of each AC component of m) is calculated. (That is, the sum of the absolute values of the coefficients at the same coordinates (n, m) is obtained.)
The absolute value of the coefficient of the coordinates (1, 0) in the DCT block 2 (FIG. 4) is 5, and the absolute value of the coefficient of the coordinates (1, 0) in the DCT block 3 (FIG. 5) is 5, DC
The sum of the absolute values of the coefficients of the coordinates (1, 0) of the T block 2 (FIG. 4) and the DCT block 3 (FIG. 5) is 10. DCT block 2 thus obtained (FIG. 4)
FIG. 12 shows the intensity values (sums of the absolute values of the coefficients) at the coordinates (n, m) of the DCT block 3 (FIG. 5).

The above series of processing (S001 and S001 in FIG. 11)
002, S003, S004, and S005) are completed for all original image blocks (S0 in FIG. 11).
06), until the processing of all the original image blocks is completed.

<Creation of Scan Order Table> For example, it is assumed that all the original image blocks are only the DCT block 2 and the DCT block 3 shown in FIGS. In that case, the intensity of the coefficient at each coordinate (n, m) is as shown in FIG. With respect to the coefficient strength value (FIG. 12), the priority order determining means (411 in FIG. 10) scans the coefficient in such a manner that sampling is performed preferentially from a coefficient having a larger strength value (sum of absolute values of coefficients). Is determined and stored in the scan order table (402 in FIG. 10) (S00 in FIG. 11).
7). In this case, the scan order table is as shown in FIG.

Here, the scan order table (4 in FIG. 10)
02) stored in the scan order information encoding unit (40 in FIG. 10).
According to 3), it is encoded and output as an output signal (S008 in FIG. 11). In the encoding of the scan order information performed here, a value obtained by simply expressing a coordinate value in a binary number may be output, or a non-zero coefficient hardly appears in a low frequency component in image data or audio data. Alternatively, a method of outputting a variable-length code such that a code becomes shorter as the coordinate value is closer to the low-frequency component may be used.

<Execution and Encoding of Scan> Next, in accordance with the scan order stored in the created scan order table (402 in FIG. 10), read control means (FIG. 1)
In 0) 3), an address signal D of the frequency component storage means (2 in FIG. 10) is generated, and from the frequency component storage means (2 in FIG. 10), each coefficient of the quantized DCT block is stored as the last non-zero. The coefficients are read out (S009 in FIG. 11). As an example, using the scan order table shown in FIG.
The state of scanning block 2 and DCT block 3
These are shown in FIGS. 14 and 15, respectively. As a result, DCT block 2 (FIG. 14) has one zero coefficient and a total of 13 other non-zero coefficients, as shown in Table 1. DCT block 3 (FIG. 15) has two zero coefficients. , And other non-zero coefficients are 13 in total, as shown in Table 2. This Table 1
And the total result shown in Table 2 is the same as the result in the first embodiment of the present invention, but in the present embodiment (the second embodiment of the present invention), the encoding is performed as it is. not,
The coefficient coding means (5 in FIG. 10) divides the coefficients arranged one-dimensionally in the scan order into a plurality of groups, and performs coding for each group (S010 in FIG. 11). D
When the CT block 2 (FIG. 14) is arranged one-dimensionally in the scan order obtained above, "5, 4, 3, 4, 3, 3, 2, 2"
1, 2, 2, 0, 1, 1, 1 ". these,
Grouping into "upper group" and "lower group". For example, if the top six items are defined as the “upper group” and the subsequent groups are defined as the “lower group”,
Upper group "is" 5,4,3,4,3,3 ",
"Lower group" is "2,1,2,2,0,
1,1,1 ". Table 3 shows the tabulated results of the “upper group” and Table 4 shows the tabulated results of the “lower group”.

[0138]

[Table 7]

[0139]

[Table 8]

When the information amount (excluding EOB) is obtained according to Equation 3 for each group, the “upper group” becomes “
Approximately 8.8 [bit] and “lower group” are “approximately 11.2 [bit]. 0 [bit] ". This is the same as the example of FIG. 90 (about 56.5 [bit]) described in the first conventional problem and the example of FIG. 8 (about 33.8 [bit]) described in the first embodiment of the present invention. bit])
The result is better than either of the above. The reason is that when grouping is not performed, the number of states is 5 states (the coefficient values are 5 states of 0, 1, 2, 3, 4, and 5). By grouping into “lower groups”, “higher values” concentrate on “upper groups” and “lower values concentrate” on “lower groups”.
Number of states in each group (in the "upper group", the coefficient values are 3 and 4 and 5; in the "lower group", the coefficient values are 0, 1 and 2)
This is because it is possible to reduce

The above processing (S009, S01 in FIG. 11)
0, S011) is performed for all DCT blocks.
At this time, a conventional Huffman code may be applied as the encoding method, or the encoding may be performed after an optimal code is derived after performing statistical processing.

As described above, in the present embodiment (the second embodiment of the present invention), each coordinate (n,
The information amount can be reduced by calculating the “coefficient strength” of m) and scanning preferentially from the coefficient of the coordinate having the larger strength value.

It is to be noted that the present invention is not limited to the above-described configuration, and may include not only the coefficient strength calculating means (412 in FIG. 10) but also the zero coefficient counting means (41 in FIG. 1) according to the first embodiment of the present invention.
0), and the scanning order may be determined using both “the coefficient strength” and “the number of zero coefficients”. For example, when there are a plurality of frequency components having the same “coefficient strength”, a method may be adopted in which the frequency components having a small “number of zero coefficients” are set to the priority scanning order.

In the present embodiment (the second embodiment of the present invention), information on scan order (for example, coordinate values)
Is encoded by the scan order information encoding means (403 in FIG. 10) and output as an output signal (S in FIG. 11).
008) Therefore, the amount of information becomes an overhead. However, the amount of information that can be reduced from each DCT block only needs to be larger than the overhead, so that there are many DCT blocks in one frame as a whole. In the case (eg, there are 5400 DCT blocks), the overhead can be sufficiently negligible on average per block.

<Third Embodiment> Next, a description will be given of a third embodiment of the present invention for solving the first problem of the above-mentioned conventional signal processing system.

In the first embodiment of the present invention, the number of zero coefficients of each coordinate (n, m) of the DCT block after frequency decomposition and quantization is counted, and the coefficient of a coordinate having a small zero coefficient is counted. In the second embodiment of the present invention, the intensity of the coefficient for each coordinate (n, m) of the DCT block after frequency decomposition and quantization is reduced by preferentially scanning from , And the information amount is reduced by preferentially scanning from the coefficient of the coordinate having the larger intensity value. By adding a coefficient distribution determination process as pre-processing to the above-described first and second embodiments of the present invention, it becomes more practical. In the third embodiment of the present invention, as an example, it will be described that adding a coefficient distribution determination process as preprocessing to the second embodiment of the present invention makes it more practical.

FIG. 16 shows the configuration of a signal processing system (compression encoding side) according to the third embodiment of the present invention. FIG.
, 9 is a frequency decomposition means, 10 is a quantization means, 1 is a basic coding means, and 2 is a frequency component storage means in the basic coding means (1).
Is a coefficient readout control means, 4 is a readout control information generating means, and 401 is a scan order determining means inside the readout control information generating means (4).
2 is a scan order table. 403 is a scan order information encoding unit. The scan order information E is used as read control information C. The coefficient read control unit (3) uses the scan order information E based on the read control information C. An address signal D is generated, and 412 inside the scan order determining means (401) is a coefficient strength calculating means, 411 is a priority order determining means, and the above is basically the same as that of the present invention described above. The function is the same as that of the second embodiment. The difference from the second embodiment of the present invention is that inside the read control information generating means (4), it is determined from the value of the coefficient A (or coefficient B) to which frequency component the energy is biased. ,
Coefficient distribution determining means (407) for outputting information indicating a bias state (hereinafter, coefficient distribution information), coefficient distribution information storing means (408) for storing coefficient distribution information, and coefficient distribution information for encoding coefficient distribution information It has an encoding unit (409), a scan order determining unit (401) determines a scan order for each coefficient distribution information, and stores a scan order information E for each coefficient distribution information in a scan order table (402). This is the configuration.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (compression encoding side) according to the third embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 17).
01).

<DCT and Quantization> Next, the frequency decomposition means (9 in FIG. 16) performs a DCT operation as shown in Equation 1 on the input original image block (S002 in FIG. 17), and DC (10 in FIG. 16)
Each coefficient (DCT block) after T is quantized (FIG. 17).
S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 16) (S004 in FIG. 17). Here, as an example, any coefficient is quantized by applying 1 as a uniform quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system. The above processing (S00 in FIG. 17)
1, S002, S003, and S004) correspond to the second embodiment of the present invention.
This embodiment is the same as the above embodiment, but differs in the following processing.

<Determination of Coefficient Distribution> Here, the coefficient distribution determining means (407 in FIG. 16) determines which frequency component the energy is biased by each DCT block after quantization, and determines the bias state. Information (hereinafter, coefficient distribution information)
And stores the coefficient distribution information in the coefficient distribution information storage means (408 in FIG. 16) (S005 in FIG. 17). In this coefficient distribution determination process, it is determined whether the energy of the DCT block of interest is biased to the area A or the area B (coefficient distribution), for example, as shown in FIG. Things. Specifically, in the DCT block of interest, the sum of the absolute value (or square value) of the coefficient of the area A and the sum of the absolute value (or square value) of the coefficient of the area B are calculated. It is determined that the energy is biased (coefficients are distributed) in a region where the sum of absolute values (or square values) of the coefficients is larger. (However, this is an example, and it is not always necessary to use all the coefficients of the area A and the area B, and the absolute values (or 2
Alternatively, the determination may be made based on the sum of the power values. For example, in the case of the DCT block 2 shown in FIG. 4, the sum of the absolute values of the coefficients in the area A is 25 and the sum of the absolute values of the coefficients in the area B is 3, and this DCT block 2 (FIG. 4) In the case of (1), it is determined that the energy is biased in the region A (coefficients are distributed), and such a DCT block is classified as a coefficient distribution group A. (If the energy is biased in the region B (coefficients are distributed), D
Classify the CT blocks as coefficient distribution group B.
) Grouping is performed in this way. In order to identify the group into which the DCT block of interest is grouped, 1 [bit]
] Is prepared. For example, in the case of the coefficient distribution group A, the coefficient distribution information is set to zero (1 [bit]
]), And in the case of the coefficient distribution group B, the coefficient distribution information is set to 1 (1 [bit]). These pieces of coefficient distribution information are stored in coefficient distribution information storage means (408 in FIG. 16). After performing such grouping of the coefficient distribution groups as preprocessing, the following processing is performed to create a scan order table for each coefficient distribution group. In the following processing, any of the processing of the first embodiment and the second embodiment of the present invention may be performed. However, in the following, as an example, processing similar to that of the second embodiment will be described. (Eg, calculation of intensity). In the above example, the coefficient distribution is determined directly using the coefficients of the DCT block. However, the coefficient distribution of the DCT block is estimated from the contour shape of the input signal (original image data) before frequency decomposition. Such a method may be used. For example, a difference between pixels in the horizontal direction is calculated for an input signal (original image data) before frequency decomposition is performed, and when the difference (absolute value) is large, The DCT block is classified into the coefficient distribution group A, the difference between pixels in the vertical direction is calculated, and when the difference (absolute value) is large, the DCT block is classified into the coefficient distribution group B, As in direct DCT
It is possible to obtain almost the same determination result as when determining the coefficient distribution from the block coefficients.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 16) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S006 in FIG. 17). The difference from the second embodiment of the present invention is that
In order to create a scan order table for each coefficient distribution group, coefficient distribution information is obtained from the coefficient distribution determining means (407 in FIG. 16), the intensity of coefficient distribution group A is calculated, and the coefficient distribution group B is calculated. The strength calculations are performed separately.

The above series of processing (S001 and S001 in FIG. 17)
It is determined whether or not 002, S003, S004, S005, and S006) have been completed for all the original image blocks (S007 in FIG. 17), and are repeated until the processing of all the original image blocks is completed.

<Coding of Coefficient Distribution Information> Next, each DCT is encoded by the coefficient distribution information encoding means (409 in FIG. 16).
Coding of coefficient distribution information corresponding to a block is performed (FIG. 1).
7 S008). If there are 5400 DCT blocks in one frame, the coefficient distribution information stored in the coefficient distribution information storage means (408 in FIG. 16) is
There are 5400 [bits]. This information may be encoded as it is without compression, or compression encoding such as Huffman encoding or arithmetic encoding may be performed as a binary signal. The coefficient distribution information is (in the case of uncompressed) one DC
There is 1 [bit] for the T block, which is an overhead. For example, if the information amount can be reduced by 1 [bit] or more per DCT block, this overhead does not pose a problem. Rather, the image data (or audio data) that is the input signal is not always composed of, for example, only the DCT blocks of the coefficient distribution group A. Is created, and in the subsequent processing, each DCT block is coded after being scanned using a suitable scan order table, so that the amount of information can be reduced.

<Preparation of Scan Order Table> Next, a scan order table is prepared for each coefficient distribution group. The calculation of the coefficient intensities for each coefficient distribution group (in the above example, coefficient distribution group A and coefficient distribution group B) has already been completed (S006 in FIG. 17). The priority order determining means (411 in FIG. 16) determines a scan order (for each coefficient distribution group) such that sampling is performed from a coefficient having a large intensity value (sum of absolute values of coefficients), and a scan order table (FIG. 16). (402 in FIG. 17)
S009). That is, in this example, a scan order table for the coefficient distribution group A and a scan order table for the coefficient distribution group B are created. Here, scan order information (eg, coordinate values) stored in the scan order table (402 in FIG. 16) is encoded by a scan order information encoding unit (403 in FIG. 16) and output as an output signal. (S010 in FIG. 17).
The encoding of the scan order information performed here simply involves the coordinate value of 2
A value expressed in a hexadecimal number may be output, or in image data or audio data, since a non-zero coefficient is unlikely to appear in a low-frequency component, a coordinate value closer to the low-frequency component has a shorter sign. A method such as outputting as a long code may be used.

<Execution and Coding of Scan> Next, a scan order table for the coefficient distribution group A and a coefficient distribution group depending on whether each quantized DCT block is the coefficient distribution group A or the coefficient distribution group B. Select one of the scan order tables for B (S in FIG. 17).
011), an address signal D of the frequency component storage means (2 in FIG. 16) is generated by the read control means (3 in FIG. 16) in accordance with the scan order, and quantized from the frequency component storage means (2 in FIG. 16). Each coefficient of the subsequent DCT block is read up to the last non-zero coefficient (S012 in FIG. 17). Then, the coefficient encoding means (5 in FIG. 16) assigns 1 to the scan order.
The coefficients arranged in a dimension are encoded (S013 in FIG. 17).

The above processing (S011, S01 in FIG. 17)
2, S013, S014) are performed for all DCT blocks. At this time, a conventional Huffman code may be applied as the encoding method, or the encoding may be performed after an optimal code is derived after performing statistical processing.

As described above, in the present embodiment (third embodiment of the present invention), by adopting the above configuration, the image data (or the audio data) which is the input signal is transferred to the coefficient distribution group A. Even in a state where the DCT blocks are not necessarily composed only of the DCT blocks, the coefficient distribution groups are grouped, and a scan order table is created for each group. By performing encoding after scanning using a suitable scan order table, it is possible to prevent each DCT block from excessively sampling a zero coefficient, and to reduce the amount of information.

The grouping of the coefficient distribution groups is not limited to the shapes such as the regions A and B in FIG. 18, but may be divided into various shapes.

<Fourth Embodiment> Next, a fourth embodiment of the present invention for solving the first problem of the conventional signal processing system will be described.

In the first embodiment of the present invention, the “number of zero coefficients” of each coordinate (n, m) of the DCT block after frequency decomposition and quantization is counted, and the coefficient of a coordinate having a small zero coefficient is counted. In the second embodiment of the present invention, the intensity of the coefficient for each coordinate (n, m) of the DCT block after frequency decomposition and quantization is reduced by preferentially scanning from , And the information amount is reduced by preferentially scanning from the coefficient of the coordinate having the larger intensity value. In the third embodiment of the present invention, there has been described an example which becomes more practical by adding a coefficient distribution determination process as preprocessing to the first and second embodiments of the present invention.
By adding a process of classifying into a complexity group as preprocessing to the first, second, and third embodiments,
The amount of information can be reduced more effectively. In the fourth embodiment of the present invention, as an example, it is described that the amount of information can be more effectively reduced by adding a process of classifying into a complexity group as preprocessing to the second embodiment of the present invention. I do.

FIG. 19 shows the configuration of a signal processing system (compression encoding side) according to the fourth embodiment of the present invention. FIG.
, 9 is a frequency decomposition means, 10 is a quantization means, 1 is a basic coding means, and 2 is a frequency component storage means in the basic coding means (1).
Is a coefficient readout control means, 4 is a readout control information generating means, and 401 is a scan order determining means inside the readout control information generating means (4).
2 is a scan order table. 403 is a scan order information encoding unit. The scan order information E is used as read control information C. The coefficient read control unit (3) uses the scan order information E based on the read control information C. An address signal D is generated, and 412 inside the scan order determining means (401) is a coefficient strength calculating means, 411 is a priority order determining means, and the above is basically the same as that of the present invention described above. The function is the same as that of the second embodiment. The difference from the second embodiment of the present invention resides in that a complexity determining means (40) for calculating "complexity" of image data (or audio data) and outputting complexity information F is provided. The control information generating means (4) is configured to input complexity information F of image data (or audio data). Inside the read control information generating means (4), a frequency component is stored from the input complexity information F. A complexity group classifying means (404) for classifying each frequency component in the means (2) into a plurality of complexity groups; and a complexity storing the complexity group information G classified by the complexity group classifying means (404). Group information storage means (405);
A complexity group information encoding unit (406) for encoding the complexity group information G 1; at least the complexity group information G is used as read control information C; The point is that an address signal D is generated based on the information C.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (compression encoding side) according to the fourth embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
20 (original image block) (S0 in FIG. 20)
01).

<DCT and Quantization> Next, in the frequency decomposition means (9 in FIG. 19), the input original image block is subjected to a DCT operation as shown in Expression 1 (S002 in FIG. 20), and the quantization is performed. In the conversion means (10 in FIG. 19), D
Each coefficient (DCT block) after CT is quantized (FIG. 2).
0 S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 19) (FIG. 20).
S004). Here, as an example, all coefficients are
Uniform quantization is performed by applying 1 as a quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system.

The above processing (S001 and S002 in FIG. 20)
, And S003 and S004) are the same as in the second embodiment of the present invention, except for the following processing.

<Classification of Complexity Group> Here, the complexity determining means (40 in FIG. 19) determines the AC component of the quantized DCT block stored in the frequency component storing means (2 in FIG. 19). The sum of the absolute values of the coefficients is obtained, and this value is output as complexity information F. The complexity group classifying means (404 in FIG. 19) classifies each of the quantized DCT blocks into a plurality of complexity groups based on the input complexity information F, and outputs complexity group information G. The complexity group information G is stored in the complexity group information storage means (405 in FIG. 19) (S00 in FIG. 20).
5). In the complexity determining means (40 in FIG. 19), for example, the sum of the absolute values of the coefficients of the AC components of the quantized DCT block stored in the frequency component storing means (2 in FIG. 19) is used as an example. Although the value is set as the complexity information F, the complexity information F is information that means the “complexity” of the original image block (or DCT block) of interest. When the brightness (or color) changes greatly due to the contour (edge) of the subject, etc., it is complicated (the complexity is strong).
If the luminance (or color) of each pixel of the original image block changes little, it is assumed to be monotonous (the complexity is low). Therefore, the complexity information F 1 is, for example, a sum of squares a of each pixel value of the original image block, a variance b of each pixel value of the original image block, and a DCT block (or a quantized DCT block). Sum of squares of coefficients c, DCT
Absolute value d of some coefficients of block (or quantized DCT block) or sum e of absolute values of all coefficients
When classifying the values of the sum of squares a, the variance b, the sum of squares c, the absolute value d, and the sum of absolute values e by size, the class number (hereinafter, the class number) (Number) or the like. In the process of classifying each of the quantized DCT blocks into a plurality of complexity groups based on the complexity information F, a group having a high complexity (hereinafter, referred to as a complex group) and a group having a low complexity (hereinafter, a complex group). , Called a monotone group). At that time, in order to identify which group the DCT block of interest is grouped into, one for each DCT block.
The complexity group information G of [bit] is prepared. For example, in the case of a monotone group, the complexity group information G is set to zero (1 [bit]), and in the case of a complex group,
The complexity group information G is set to 1 (1 [bit]).
The complexity group information G is stored in the complexity group information storage means (405 in FIG. 19). After performing such grouping of complexity groups as preprocessing, the following processing is performed to create a scan order table for each complexity group. In the following processing, any of the processing of the first embodiment and the second embodiment of the present invention may be performed, and the coefficient distribution group of the third embodiment of the present invention may be used. Although it may be used together with the grouping process, in the following, as an example, it is assumed that the same process (eg, calculation of intensity) as in the second embodiment is performed. In the above example, the process of obtaining the complexity information F is performed directly using the coefficients of the DCT block. However, the process of obtaining the pixel value of each pixel value of the original image block of the input signal (image data) before the frequency decomposition is performed.
The sum of squares a, the variance b of each pixel value of the original image block, and D
Sum of squares c of each coefficient of CT block (or quantized DCT block), absolute value d of some coefficients of DCT block (or quantized DCT block), or sum of absolute values of all coefficients e, the sum of squares a, the variance b, the sum of squares c, the absolute value d, and the sum of absolute values e, when classifying the values by size (hereinafter referred to as the class number). Class number).

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 19) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S006 in FIG. 20). The difference from the second embodiment of the present invention is that
In order to create a scan order table for each complexity group, the complexity group information G is obtained from the complexity group classification means (404 in FIG. 19) to calculate the intensity for the complexity group and the intensity for the monotone group. Are performed separately.

The above series of processing (S001 and S001 in FIG. 20)
It is determined whether or not 002, S003, S004, S005, and S006) have been completed for all the original image blocks (S007 in FIG. 20), and are repeated until the processing for all the original image blocks is completed.

<Encoding of Complexity Group Information> Next, the complexity group information encoding means (406 in FIG. 19) encodes the complexity group information G corresponding to each DCT block.
(S008 in FIG. 20). If there are 5400 DCT blocks in one frame,
The complexity group information G stored in the complexity group information storage means (405 in FIG. 19) is 5400 [bi
t] exists. This information may be encoded as it is without compression, or compression encoding such as Huffman encoding or arithmetic encoding may be performed as a binary signal. The complexity group information G is 1 [bit] for one DCT block (in the case of non-compression), and this becomes overhead. However, 1 [b] per DCT block.
It] is not a problem if the amount of information can be reduced by more than [it]. It will be described later that coding is performed after being classified into a plurality of complexity groups, thereby improving the coding efficiency so that overhead does not become a problem.

<Creation of Scan Order Table> Next, a scan order table is created for each complexity group. The calculation of the strength of the coefficient for each complexity group (in the above example, the monotone group and the complex group) has already been completed (S006 in FIG. 20). Priority determination means (4 in FIG. 19)
According to 11), a scan order (for each complexity group) is determined such that sampling is performed from a coefficient having a large intensity value (sum of absolute values of coefficients), and a scan order table (FIG. 1).
9 at 402) (S009 in FIG. 20). That is, in this example, a scan order table for a monotone group and a scan order table for a complex group are created. Next, a scan order table (40 in FIG. 19)
The scan order information (for example, coordinate values) stored in 2) is converted into scan order information encoding means (403 in FIG. 19).
And output as an output signal (FIG. 20).
S010). The encoding of the scan order information performed here is
A value obtained by simply expressing the coordinate value in a binary number may be output, or in image data or audio data, since a non-zero coefficient is unlikely to appear in a low-frequency component, a shorter code value is used for a coordinate value closer to the low-frequency component. For example, a method of outputting as a variable length code such that

<Execution and Encoding of Scan> Next, depending on whether each quantized DCT block is a monotone group or a complex group, either a scan order table for a monotone group or a scan order table for a complex group is determined. 20 is selected (S011 in FIG. 20), the address signal D of the frequency component storage means (2 in FIG. 19) is generated by the read control means (3 in FIG. 19) in accordance with the scan order, and the frequency component storage means (FIG. From 19-2), the quantized DCT
Each coefficient of the block is read up to the last non-zero coefficient (S012 in FIG. 20). Then, coefficients arranged one-dimensionally in scan order are encoded for each complexity group by coefficient encoding means (5 in FIG. 19) (S013 in FIG. 20).

The above processing (S011, S01 in FIG. 20)
2, S013, S014) are performed for all DCT blocks. At this time, a conventional Huffman code may be applied as the encoding method, or the encoding may be performed after an optimal code is derived after performing statistical processing.

As described above, the merits of performing coding by grouping each complexity group will be described. In the monotone group, small values are concentrated as the absolute value of the AC component coefficient, and in the complex group, the large value is concentrated as the absolute value of the AC component coefficient, so that the information amount of each group can be reduced. In the example of FIG. 21, a scan table is created for DCT block 1, DCT block 2, DCT block 3, and DCT block 4 in the same manner as in the second embodiment of the present invention, and the same as in FIG. When a scan table that performs processing in order is used, DCT block 1, DCT block 2, DCT block 3, DCT
When block 4 is mixed, the result is as shown in Table 5, and when the information amount (excluding EOB) is obtained according to Equation 3, about 1 is obtained.
62.8 [bits]. On the other hand, complexity grouping is performed, and as shown in FIG. 21, DCT block 1 and DCT block 3 are classified into complexity groups, and DCT block 2 and DCT block 4 are classified into monotone groups. Then, when the results for each group are totaled, the results are as shown in Tables 6 and 7, respectively. When the information amount (excluding EOB) is obtained according to Equation 3 for each group, about 57.2 [bits] are obtained. ] And about 48.
1 [bit], and the total information amount of the two groups is about 105.2 [bit]. As described above, the amount of information can be significantly reduced by classifying each block according to the complexity.

[0177]

[Table 9]

[0178]

[Table 10]

[0179]

[Table 11]

<Fifth Embodiment> Next, a fifth embodiment of the present invention for solving the second problem of the conventional signal processing system will be described.

A second problem of the conventional signal processing system is that each coefficient (one-dimensional data in FIG. 83) of the one-dimensionally arranged DCT block is converted into coefficient coding means (FIG. 81).
51), 1 [bit] is assigned to one non-zero coefficient in the Huffman encoding process (S005 in FIG. 82).
] Sign information is used, the code amount is increased. For example, when the image size of one frame is composed of 720 × 480 pixels, there are 5400 DCT blocks per frame. If you want to improve the image quality of the data after compression and encoding (that is, if you want to reduce image quality degradation),
The value of the quantization step at the time of quantization is set small. In that case, each coefficient of the AC component of the quantized DCT block is likely to be non-zero. For example, if all coefficients are non-zero, then for one entire frame,
5400 × 63 = 340200 [bit] = 42.
525 [KB] signature information is required. As described above, there is a problem (a second problem) that a large number of codes are required for signature information.

In order to solve the second problem, in the fifth embodiment of the present invention, the sign information is collectively encoded to reduce the code amount of the sign information. Coefficients other than sign information are coded by a process similar to that of the related art or the first, second, third, and fourth embodiments of the present invention, except for the sign information as a mere absolute value. Should be done. In the fifth embodiment of the present invention, as an example, the sign information is processed by a sign information processing means (described later), and the absolute value part of the coefficient excluding the sign information is set in the second embodiment of the present invention. The description will be made on the assumption that the same processing as in the embodiment is performed.

FIG. 22 shows the configuration of a signal processing system (compression encoding side) according to the fifth embodiment of the present invention. FIG.
, 9 is a frequency decomposition means, 10 is a quantization means, 1 is a basic coding means, and 2 is a frequency component storage means in the basic coding means (1).
Is a coefficient readout control means, 4 is a readout control information generating means, and 401 is a scan order determining means inside the readout control information generating means (4).
2 is a scan order table. 403 is a scan order information encoding unit. The scan order information E is used as read control information C. The coefficient read control unit (3) uses the scan order information E based on the read control information C. An address signal D is generated, and 412 inside the scan order determining means (401) is a coefficient strength calculating means, 411 is a priority order determining means, and the above is basically the same as that of the present invention described above. The function is the same as that of the second embodiment. The difference from the second embodiment of the present invention resides in that the basic encoding means (1) includes an absolute value conversion means (6) for converting each coefficient of the frequency component into an absolute value, And a sine information processing means (7) for processing the sine information of the inputted coefficients. For each coefficient of the frequency component converted into the absolute value, a coefficient encoding means (5) Do the processing,
Coding coefficients, sign information processing means (7)
A non-zero determination signature information extracting means (701) for extracting whether or not the input coefficient is non-zero and extracting the signature information if the input coefficient is non-zero; a signature information storage means for storing the signature information (702) and a sign information encoding means (703) for performing statistical processing on the sign information and encoding the sign information.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (on the compression encoding side) according to the fifth embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 23).
01).

<DCT and Quantization> Next, the frequency decomposition means (9 in FIG. 22) performs a DCT operation as shown in Equation 1 on the input original image block (S002 in FIG. 23), and In the converting means (10 in FIG. 22), D
Each coefficient (DCT block) after CT is quantized (FIG. 2).
3 S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 22) (FIG. 23).
S004). Here, as an example, all coefficients are
Uniform quantization is performed by applying 1 as a quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 22) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S005 in FIG. 23).

The above series of processing (S001 and S001 in FIG. 23)
002, S003, S004, and S005) are determined for all the original image blocks (S0 in FIG. 23).
06), until the processing of all the original image blocks is completed.

<Creation of Scan Order Table> Next, the priority order determining means (411 in FIG. 22) determines a scan order in which sampling is performed preferentially from a coefficient having a large intensity value (sum of absolute values of coefficients). It is determined and stored in the scan order table (402 in FIG. 22) (S00 in FIG. 23).
7). Here, the scan order table (402 in FIG. 22)
Scan order information (eg, coordinate values) stored in
Are encoded by the scan order information encoding means (403 in FIG. 22) and output as an output signal (S in FIG. 23).
008). In the encoding of the scan order information performed here, a value obtained by simply expressing a coordinate value in a binary number may be output, or a non-zero coefficient hardly appears in a low frequency component in image data or audio data. Alternatively, a method of outputting a variable-length code such that a code becomes shorter as the coordinate value is closer to the low-frequency component may be used.

<Execution and Encoding of Scan> Next, read control means (FIG. 2) is executed in accordance with the scan order stored in the created scan order table (402 in FIG. 22).
In 2-3), an address signal D of the frequency component storage means (2 in FIG. 22) is generated, and each coefficient of the quantized DCT block is converted from the frequency component storage means (2 in FIG. 22) to the last non-zero value. The coefficients are read out (S009 in FIG. 23). The read coefficients are converted into absolute values by the absolute value conversion means (6 in FIG. 22) (S010 in FIG. 23), and each coefficient converted into the absolute value is converted into coefficient encoding means (5 in FIG. 22). (S011 in FIG. 23). That is, the encoding process (S011 in FIG. 23) is performed only on the absolute value part of the coefficient excluding the sign information. The processing for obtaining the absolute value of the coefficient may be performed before storing the coefficient in the frequency component storage means (2 in FIG. 22). In this case, each coefficient is stored in the frequency component storage means (2 in FIG. 22). , And may be stored separately for the absolute value and the sign information.

<Processing of Signature Information> Next, by the reading process in the scan order (S009 in FIG. 23), the coefficient to be read is set to zero by the non-zero determination signature information extracting means (701 in FIG. 22). It is determined whether the signature information is non-zero or not, and the signature information is extracted (S0 in FIG. 23).
12) The signature information is stored in the signature information storage means (702 in FIG. 22) (S013 in FIG. 23).

The above processing (S009, S01 in FIG. 23)
0, S011, S012, and S013) are performed on all DCT blocks (S014 in FIG. 23). As a result, the sign information of all the non-zero coefficients is stored in the sign information storage means (702 in FIG. 22).

<Encoding of signature information> Next, signature information encoding means (703 in FIG. 22) is applied to all non-zero coefficient signature information stored in the signature information storage means (702 in FIG. 22). ) To perform statistical processing, and compress and encode all the sign information (S01 in FIG. 23).
5). In this encoding process, statistical processing is performed, a dedicated Huffman table is created, Huffman encoding is performed, arithmetic encoding is performed, or any encoding method may be used.

As described above, in the present embodiment (the fifth embodiment of the present invention), by adopting the above-mentioned configuration, in the prior art, one non-zero coefficient has a sign of 1 [bit]. A second problem that a code amount is increased by using information is solved.

As described above, the fact that the amount of information can be reduced by collectively encoding the sign information will be described with reference to the following example. FIG. 24 shows an example of an original image block. In this example, the range of the luminance value of each pixel of the original image is “0 to 25”.
5 ". FIG. 25 shows an example in which the original image block (FIG. 24) is subjected to a DCT operation as shown in Equation 1 and quantized to the nearest integer. In the DCT block of FIG. 25, the 63 coefficients of the AC component are all non-zero, so that the conventional signal processing system requires 63 [bit] of sign information. As in the fifth embodiment), these pieces of signature information are collected, statistically processed collectively, and compression-coded. In the DCT block (FIG. 25), there are 28 positive coefficients (AC components) and 35 negative coefficients (AC components). When the information amount of these pieces of signature information is obtained according to Equation 3, it is about 62.4 [bits]. One DCT
When looking at the block alone, the code amount is slightly reduced by about 0.6 [bit]. However, when there are as many as 5400 DCT blocks in one frame as in a general television, the whole frame is reduced. And 3240 [bit
] Can be reduced.

<Sixth Embodiment> Next, a sixth embodiment of the present invention for solving the second problem of the conventional signal processing system will be described.

In the fifth embodiment of the present invention described above, signature information is collectively encoded by signature information processing means (7 in FIG. 22), thereby reducing the code amount of signature information. 2 issues have been solved. In the sixth embodiment of the present invention, the signature information processing means (7 in FIG. 22) is provided with a function of estimating each signature information, and after sign information is estimated, encoding is performed. This is to reduce the code amount of information.

The image data, which is an input signal, is obtained by converting a signal from an image taken by a video camera or the like or from a photograph by a scanner. Things. Thus, the input signal (image data) is generated based on optical natural phenomena. Therefore, the contour (edge) of the subject
Information such as the shape of the input signal (image data)
Is reflected in the numerical value. These input signals (image data)
In the signal processing systems according to the related art and the present invention, which are encoded after frequency decomposition of the
Is characterized in that the energy is biased to any of the frequency components after the frequency decomposition. For example, if the contour (edge) of the subject has a complicated shape, energy concentrates on high frequency components, and if the contour (edge) has a monotonous shape, energy concentrates on low frequency components. I do. In other words, the frequency components where energy is concentrated reflect the characteristics of the input signal (image data) well and contain much information on the input signal (image data). In view of such characteristics of the input signal (image data), in the sixth embodiment of the present invention, the code amount of the sign information is reduced by the following configuration as shown in FIG. As in the fifth embodiment of the present invention, coefficients other than the sign information are used as a mere absolute value except for the sign information as in the prior art or the first, second, and third embodiments of the present invention described above. The encoding may be performed by the same processing as that of the fourth embodiment. However, in the sixth embodiment of the present invention, as an example, the absolute value part of the coefficient excluding the sign information is the second value of the present invention. The following description is made on the assumption that the same processing as that of the embodiment is performed.

The configuration of the sixth embodiment of the present invention will be described below with reference to FIG. In FIG. 26, 9
Is frequency decomposition means, 10 is quantization means, 1
Is a basic encoding unit, 2 is a frequency component storage unit, 3 is a coefficient readout control unit, and 4 is a readout control information generation unit inside the basic encoding unit (1).
Then, inside the read control information generating means (4),
401 is a scan order determining means, 402 is a scan order table, 403 is a scan order information encoding means, and the scan order information E is used as read control information C, and the coefficient read control means (3) An address signal D is generated based on the read control information C, and 412 inside the scan order determining means (401)
411 is a priority order determining means.
The above functions basically in the same manner as the above-described second embodiment of the present invention. The difference from the second embodiment of the present invention is that an absolute value conversion means (6) for converting each coefficient of a frequency component into an absolute value is provided inside the basic coding means (1).
And a sine information processing means (7) which receives each coefficient of the frequency component as input and processes sine information of the input coefficient. For each coefficient of the frequency component converted into an absolute value,
The coefficient encoding means (5) performs statistical processing to encode the coefficients, and the sign information processing means (7) has:
Among the coefficients of each frequency component in the frequency component storage means (2), the intensity of the energy of each coefficient in the unit of frequency decomposition (in the DCT block) (hereinafter, the intensity) is measured, and the intensity of each coefficient is measured. , A coefficient strength rank determining means (704) for determining the rank (hereinafter referred to as a coefficient strength rank), a coefficient strength rank table (705) for storing the coefficient strength ranks, and determining whether each coefficient is zero. Only in this case, a non-zero determination signature information extracting means (701) for extracting signature information;
Sign information estimating means (706) for preferentially estimating signature information from a coefficient having a higher coefficient strength rank, and generating and outputting new signature information from the estimation result; signature information estimating means (706) , Temporary signature information (hereinafter, temporary signature information) and, if necessary, true signature information (hereinafter, true signature information) for each frequency component coefficient in the frequency component storage means (2). ), And a signature information setting means (70).
7), a signature information-added coefficient storage means (708) for storing the temporary signature information and each coefficient for which true signature information is set as necessary (hereinafter, a coefficient with signature information);
Frequency component synthesizing means (709) for synthesizing each frequency component of the coefficient with signature information and outputting a frequency synthesis result with signature information; frequency synthesis result storing means (710) for storing the frequency synthesis result with signature information; First signature information estimation parameter calculating means (711) for calculating a first signature information estimation parameter (hereinafter, first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which the temporary signature information is set; From the first signature information estimation parameter storage means (712) for storing the first signature information estimation parameter and the value of the first signature information estimation parameter corresponding to each temporary signature information, it is determined which temporary signature information is true signature information. Estimated, and coincidence mismatch information indicating whether the estimated signature information matches the true signature information A first signature information generating means (713) for outputting the signature information as signature information, and the signature information setting means (707) converts the true signature information with respect to the coefficient for which the estimation of whether the temporary signature information is the true signature information is completed. It has a function of setting, and a frequency synthesis result (hereinafter, a true sine frequency synthesis result) by a coefficient in which true signature information is set (hereinafter, a coefficient with true signature information) and a coefficient (temporary signature information is set) Hereinafter, the sum of the absolute value of the difference from the frequency synthesis result by the temporary signature information-added coefficient (hereinafter, the temporary signature frequency synthesis result) is used as the first variation, and the mixture of the true signature information-added coefficient and the temporary signature information-added coefficient Regarding the frequency synthesis result in the state (hereinafter, “mixed frequency synthesis result”), the waveform of the waveform of the mixed frequency synthesis result, the waveform in the region surrounded by the waveform, and the waveform The sum of the absolute value of the difference from the central axis of the first is regarded as the first similarity, and the first sign information estimation parameter computing means (711) computes the first change amount or the first similarity as the first sign information estimation parameter. Then, the first signature information generating means (71
In 3), when the first signature information estimation parameter corresponding to each temporary signature information is the first variation, the temporary signature information when the value of the first signature information estimation parameter is small is estimated to be true signature information. If the first signature information estimation parameter corresponding to each temporary signature information is the first similarity,
The point is that provisional signature information when the value of the first signature information estimation parameter is large is estimated to be true signature information.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (compression encoding side) according to the sixth embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 27).
01).

<DCT and Quantization> Next, the frequency decomposition means (9 in FIG. 26) performs a DCT operation as shown in Equation 1 on the input original image block (S002 in FIG. 27), and DC (10 in FIG. 26)
Quantize each coefficient (DCT block) after T (FIG. 27)
S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 26) (S004 in FIG. 27). Here, as an example, any coefficient is quantized by applying 1 as a uniform quantization step. This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 26) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S005 in FIG. 27).

The above series of processing (S001 and S001 in FIG. 27)
002, S003, S004, and S005) are completed for all original image blocks (S0 in FIG. 27).
06), until the processing of all the original image blocks is completed.

<Preparation of Scan Order Table> Next, the scan order is determined by the priority order determination means (411 in FIG. 26) such that sampling is performed from the coefficient having the largest intensity value (sum of the absolute values of the coefficients). It is stored in the scan order table (402 in FIG. 26) (S007 in FIG. 27). Here, scan order information (eg, coordinate values) stored in the scan order table (402 in FIG. 26) is encoded by a scan order information encoding unit (403 in FIG. 26) and output as an output signal. (S008 in FIG. 27).
The encoding of the scan order information performed here simply involves the coordinate value of 2
A value expressed in a hexadecimal number may be output, or in image data or audio data, since a non-zero coefficient is unlikely to appear in a low-frequency component, a coordinate value closer to the low-frequency component has a shorter sign. A method such as outputting as a long code may be used.

<Execution and Encoding of Scan> Next, read control means (FIG. 2) is executed in accordance with the scan order stored in the created scan order table (402 in FIG. 26).
In 6-3), an address signal D of the frequency component storage means (2 in FIG. 26) is generated, and each coefficient of the quantized DCT block is converted from the frequency component storage means (2 in FIG. 26) to the last non-zero value. The coefficients are read out (S009 in FIG. 27). The read coefficients are converted into absolute values by absolute value conversion means (6 in FIG. 26) (S010 in FIG. 27), and each coefficient converted into an absolute value is converted into coefficient encoding means (5 in FIG. 26). (S011 in FIG. 27). That is, this encoding process (S011 in FIG. 27) is performed only on the absolute value portion of the coefficient excluding the sign information. The processing for obtaining the absolute value of the coefficient may be performed before storing the coefficient in the frequency component storage means (2 in FIG. 26). In this case, each coefficient is stored in the frequency component storage means (2 in FIG. 26). , And may be stored separately for the absolute value and the sign information.

<Processing of Signature Information> Next, by the reading process in the scan order (S009 in FIG. 27), the coefficient to be read is determined by the coefficient strength rank determining means (7 in FIG. 26).
04), the energy intensity (hereinafter, intensity) of each coefficient is measured within one DCT block, and the intensity order of each coefficient (hereinafter, coefficient intensity order) is determined (S01 in FIG. 27).
2). Then, the value of the coefficient strength ranking is stored in the coefficient strength ranking table (705 in FIG. 26) (S01 in FIG. 27).
3). As described above, if the contour (edge) of the subject has a complicated shape, if the energy is concentrated on the high-frequency component or if the contour (edge) has a monotonous shape,
Energy concentrates on low frequency components. The frequency component where energy is concentrated reflects the characteristics of the input signal (image data) well, and the input signal (image data)
The above processing (FIG. 27)
S012) is for determining to which frequency component in the DCT block energy is concentrated.
The intensity in this process (S012 in FIG. 27) is specifically the absolute value of each coefficient or the magnitude of the square of each coefficient. As an example, in the present embodiment, the intensity is an absolute value of the coefficient. The determination of the intensity order is performed only for the AC component in the DCT block of interest. For example, in the case of the same DCT block (FIG. 25) used in the description of the fifth embodiment of the present invention, a coefficient having the largest intensity value among AC components (hereinafter, referred to as a coefficient A1). Is “−784” at the coordinates (1, 0). 2
Coefficient with the second largest intensity value (hereinafter referred to as coefficient A2)
Is “212” at the coordinates (3,0). The coefficient having the third largest intensity value (hereinafter, referred to as coefficient A3) is “−103” at the coordinates (5, 0). The coefficient having the fourth largest intensity value (hereinafter, referred to as coefficient A4) is “71” at coordinates (2, 4) and “−” at coordinates (4, 1).
71 "and two coefficients having the same magnitude. In such a case, either of them may be prioritized, but in this example, “71” of the coordinates (2, 4) is the fourth (coefficient A4). In this way, the ranking is performed in the order of the magnitudes, the coefficient strength ranking is determined, and information of the coordinates (n, m) corresponding to the ranking is stored in the coefficient strength ranking table (705 in FIG. 26). Keep it.

Next, whether or not each coefficient is non-zero is determined by the non-zero determination sign information extracting means (701 in FIG. 26) (S014 in FIG. 27). According to (706 in FIG. 26), it is estimated whether the sign information of each coefficient is positive or negative (FIG. 27).
S015 to S023). In the sign information estimation process (S015 to S023 in FIG. 27), in accordance with the previously obtained coefficient strength rank, the strength rank is low (strength value) using information of a coefficient having a high strength rank (large strength value) as a clue. (Small value) is estimated. That is, in the example of the DCT block (FIG. 25), for example, six coefficients A1 to A62 having the highest strength order are used.
Based on the information of the two coefficients, the signature information of only the coefficient A63 (“−2” of the coordinates (4, 0) in FIG. 25) having the lowest strength order is estimated. (In this example, the sign information of the 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength order is not estimated.) Therefore, first, the sign information setting means (707 in FIG. 26) sets the coefficient A1 to the coefficient A1. True sign information is set for AC components up to A62 (S015 in FIG. 27), and this DCT block (hereinafter, referred to as a reference DCT block) is stored in a coefficient storage unit with sign information (708 in FIG. 26). (S01 in FIG. 27)
6). At the same time, signature information setting means (707 in FIG. 26)
Then, positive sign information is set as temporary sign information in the AC component of the coefficient A63 for estimating sign information (S015 in FIG. 27), and this DCT block (hereinafter, DCT) is set.
Block 0) is stored in the coefficient storage unit with signature information (708 in FIG. 26) (S016 in FIG. 27). The sign information setting means (707 in FIG. 26) sets negative sign information as temporary sign information in the AC component of the coefficient A63 for estimating the sign information (S01 in FIG. 27).
5) This DCT block (hereinafter referred to as DCT block 1) is stored as a coefficient storage unit with signature information (70 in FIG. 26).
8) (S016 in FIG. 27). That is, in the coefficient storage unit with sign information (708 in FIG. 26), three DCT blocks of the reference DCT block, DCT block 0, and DCT block 1 are stored. Each DCT block is shown in FIGS. 28, 29 and 30.

Next, a coefficient storage unit with signature information (FIG. 2)
6, 708), DCT block 0 (FIG. 29), and DCT block 0 (FIG. 29).
The frequency synthesis means (709 in FIG. 26) performs the inverse DCT operation of Equation 2 on the three DCT blocks in block 1 (FIG. 30) (S017 in FIG. 27), and obtains the respective frequency synthesis results (inverse DCT result) is stored in the frequency synthesis result storage means (710 in FIG. 26) (S01 in FIG. 27).
8). Note that the above inverse DCT operation (S017 in FIG. 27)
It is preferable that the inverse quantization is performed on each DCT block before performing the inverse quantization. However, in the above series, the quantization step is set to 1 as an example, and the inverse quantization process is omitted here. did. (Inverse DCT may be provided to perform inverse quantization.) Note that inverse DCT is a real number calculation, and there are values after the decimal point. Usually, the result of the real number calculation is treated as an integer. . At this time, the value after the decimal point is truncated, causing an error with respect to the actual operation result. In order to minimize this error, the above inverse DCT operation (S017 in FIG. 27)
Is an example in which the inverse DCT result is multiplied by 10,000 and output as an integer value with higher bit precision. The inverse DCT result of the reference DCT block (FIG. 28) is 10,000
FIG. 31 shows a multiplied result (hereinafter referred to as a reference inverse DCT block) and FIG. 32 shows a result obtained by multiplying the inverse DCT result of DCT block 0 (FIG. 29) by 10000 (hereinafter referred to as an inverse DCT block 0). Inverse DC of block 1 (FIG. 30)
FIG. 33 shows a result obtained by multiplying the T result by 10,000 (hereinafter, referred to as an inverse DCT block 1).

Since the reference inverse DCT block (FIG. 31) includes information of all AC components except for the coefficient A63 having the lowest intensity rank (“−2” of the coordinates (4, 0) in FIG. 25), the original It has a shape very similar to the contour (edge) shape of the image block (FIG. 24). FIG. 36 shows the contour (edge) shape of the original image block (FIG. 24). this is,
The average value (about 122) of the pixels of the original image block (FIG. 24) is subtracted from each pixel of the original image block (FIG. 24). FIG. 37 shows a value obtained by dividing the value of the reference inverse DCT block (FIG. 31) by 1 / 10,000. This is the value obtained by dividing the value of the reference inverse DCT block (FIG. 31) by 1 / 10,000 and comparing the scale with the contour (edge) shape (FIG. 36) of the original image block (FIG. 24). is there. In FIG. 38, based on the contour (edge) shape (FIG. 36) of the original image block (FIG. 24), the reference inverse DCT block (FIG.
The value obtained by subtracting the value of the corresponding coordinate of 1 / 10,000 (FIG. 37) of the value of 1) is shown. As shown in FIG. 38, the contour (edge) shape (FIG. 36) of the original image block (FIG. 24)
And 10000 of the value of the reference inverse DCT block (FIG. 31)
In 1 / (FIG. 37), the difference is small, and it can be seen that the reference inverse DCT block (FIG. 31) has substantially the same contour (edge) shape as the original image block (FIG. 24). Now, the signature information of the coefficient A63 is to be estimated, but the coefficient A63 is originally the original image block (FIG. 2).
4) is obtained by frequency decomposition of the coefficient A63.
The inverse DCT block 0 (FIG. 32) when positive sign information is temporarily set, and the inverse DCT block 1 (FIG. 33) when negative sign information is temporarily set for the coefficient A63 are the reference inverse DCT block. It is possible to estimate the sign information of the coefficient A63 by checking whether the shape is similar to that of FIG. Based on such a concept, the first sign information estimation parameter calculation means (711 in FIG. 26) calculates the first change amount as the first signature information estimation parameter (S019 in FIG. 27), and outputs the calculation result to the first sign information estimation parameter. 1 signature information estimation parameter storage means (7 in FIG. 26)
12) (S020 in FIG. 27).

The first change amount calculated as described above is obtained by a frequency synthesis result (hereinafter, true sign frequency synthesis result) based on a coefficient in which true signature information is set (hereinafter, a coefficient with true signature information) and a temporary signature information. Is set to the coefficient (hereinafter,
Frequency synthesis result (hereinafter, referred to as a coefficient with temporary signature information)
This is defined as the sum of the absolute values of the difference from the temporary sine frequency synthesis result). Specifically, in the above example, the absolute value of the difference of each pixel between the inverse DCT block 0 (FIG. 32) and the reference inverse DCT block (FIG. 31) when positive coefficient information is temporarily set to the coefficient A63. The sum of the "positive first change amount"
The sum of the absolute value of the difference between each pixel of the inverse DCT block 1 (FIG. 33) and the reference inverse DCT block (FIG. 31) in the case where negative sign information is temporarily set to the coefficient A63 is expressed as “negative negative 1 change amount ”and this“ positive first change amount ”
And the value of the “negative first change amount” are stored in the first sign information estimation parameter storage means (712 in FIG. 26) (S020 in FIG. 27). In the above example, the value of the “positive first change amount” is “66771562”, and the value of the “negative first change amount” is “66771558”.
A reference inverse DCT block, an inverse DCT block 0 (tentatively set to positive), and an inverse DCT block 1 (tentatively set to negative)
FIG. 34 and FIG. 35 respectively show the waveforms and the first variation in a conceptual image. That is, the size (FIG. 34) of the area surrounded by the waves of the reference inverse DCT block and the inverse DCT block 0 is “positive first variation”, and the size of the reference inverse DCT block and the inverse DCT block 1 is The size of the area surrounded by the waves (FIG. 35)
"Negative first change amount".

Next, the first sign information generating means (713 in FIG. 26) compares the magnitudes of the "positive first change amount" and the "negative first change amount" and finds that the first change amount is "1. It is estimated that the sign information of "smaller" is the true sign information of the coefficient A63. If the estimation result is correct, “zero (1 [bit])” is output as new signature information, and if the estimation result is incorrect, “1 (1 [bit])” is generated as new signature information. If the magnitudes of the “positive first change amount” and the “negative first change amount” are equal, it is regarded as “impossible to estimate” and the true sign information itself of the coefficient A63 is output (FIG. 27). S02
1). In the above example, the value of the “positive first change amount” is “66771562”, and the “negative first change amount”
Is "66771558", the coefficient A6
It is estimated that the sign information of No. 3 is “negative”. This is the correct answer. Therefore, "zero (1 [bit])" is output as new signature information. The signature information output from the first signature information generation means (713 in FIG. 26) is stored in the signature information storage means (702 in FIG. 26) (S022 in FIG. 27).

The above processing (from S014 to S02 in FIG. 27)
2) is repeated until the processing is completed for all the coefficients in one DCT block (S023 in FIG. 27). In the above example, the signature information of only the last coefficient (coefficient A63) is estimated.

The above processing (from S009 to S02 in FIG. 27)
Step 3) is repeated until completion for all DCT blocks (S024 in FIG. 27).

Finally, signature information storage means (7 in FIG. 26)
02) is compression-encoded by the signature information encoding means (703 in FIG. 26) (S025 in FIG. 27). This encoding process (S0 in FIG. 27)
25) is the same as the example described above, in each of the DCT blocks, 62 from the coefficient A1 to the coefficient A62 having the highest strength order.
When the sign information is estimated only for the coefficient A63 having the lowest strength rank based on the information of the coefficients, the reference inverse DCT block composed of 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength rank is obtained. , And an original image block, the contour (edge) shape is substantially equal to the original image block, so that the probability that the estimation result is correct is increased. Then, “0” increases as the value of the sign information indicating the correct answer, and only the coefficient A63 of each DCT block is collected, statistically processed and encoded, so that the code amount can be reduced. The other coefficients A1 to A63 are separately and collectively encoded. As a method for compressing and encoding the sign information, any encoding method such as conventional Huffman encoding or arithmetic encoding may be used.

<Seventh Embodiment> Next, a seventh embodiment of the present invention for solving the second problem of the conventional signal processing system will be described.

In the fifth embodiment of the present invention described above, the sign information is collectively encoded by the sign information processing means (7 in FIG. 22), thereby reducing the code amount of the sign information. Solved 2 problems. In the sixth embodiment of the present invention, in each DCT block, estimation of the sign information of the coefficient A63 having the lowest strength order is performed based on information of 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength order. And the coefficient A newly generated by estimation
The second problem was solved by reducing the code amount of the sign information by collecting and signing 63 sign information collectively. In the seventh embodiment of the present invention, similar to the sixth embodiment of the present invention, in each DCT block, estimation of the sine information of a coefficient having a low strength order is performed based on information of a coefficient having a high strength order. , And encodes
The difference from the sixth embodiment of the present invention is that, instead of estimating the sign information of only the coefficient A63 having the lowest intensity rank, the sign information of all the coefficients A1 to A63 is estimated. The encoding significantly reduces the code amount of the signature information. Also, in the sixth embodiment of the present invention, the first change amount is used as the first sign information estimation parameter, and the first sign information generating means (713 in FIG. 26) uses the first change amount when the first change amount is small. Is estimated to be the correct answer, but this has the same meaning. However, in the seventh embodiment of the present invention, the first similarity is used as the first signature information estimation parameter, and the first similarity is used. The sign information when the degree is large is estimated to be the correct answer.

[0220] The image data as the input signal is obtained by converting the signal of a photograph or the like by a video camera or the like, or by converting the signal of a photograph or the like by a scanner. Things. Thus, the input signal (image data) is generated based on optical natural phenomena. Therefore, the contour (edge) of the subject
Information such as the shape of the input signal (image data)
Is reflected in the numerical value. These input signals (image data)
In the signal processing systems according to the related art and the present invention, the contour (edge) of the subject is
Is characterized in that the energy is biased to any of the frequency components after the frequency decomposition. For example, if the contour (edge) of the subject has a complicated shape, energy concentrates on high frequency components, and if the contour (edge) has a monotonous shape, energy concentrates on low frequency components. I do. In other words, the frequency components where energy is concentrated reflect the characteristics of the input signal (image data) well and contain much information on the input signal (image data). In view of such characteristics of the input signal (image data), in the seventh embodiment of the present invention, as in the sixth embodiment of the present invention, a signature as shown in FIG. The code amount of information is reduced. As in the fifth and sixth embodiments of the present invention, except for the sign information as a mere absolute value, the coefficients other than the sign information, except for the sign information, are the same as those of the first and second embodiments of the present invention described above. 2,
Encoding may be performed by the same processing as in the third and fourth embodiments. However, in the seventh embodiment of the present invention, as an example, the absolute value part of the coefficient excluding the sign information is the same as that of the present invention. Description will be made assuming that the same processing as that of the second embodiment is performed.

Hereinafter, the configuration of the seventh embodiment of the present invention will be described with reference to FIG. In FIG. 39, 9
Is frequency decomposition means, 10 is quantization means, 1
Is a basic encoding unit, 2 is a frequency component storage unit, 3 is a coefficient readout control unit, and 4 is a readout control information generation unit inside the basic encoding unit (1).
Then, inside the read control information generating means (4),
401 is a scan order determining means, 402 is a scan order table, 403 is a scan order information encoding means, and the scan order information E is used as read control information C, and the coefficient read control means (3) An address signal D is generated based on the read control information C, and 412 inside the scan order determining means (401)
411 is a priority order determining means.
The above functions basically in the same manner as the above-described second embodiment of the present invention. The difference from the second embodiment of the present invention is that an absolute value conversion means (6) for converting each coefficient of a frequency component into an absolute value is provided inside the basic coding means (1).
And a sine information processing means (7) which receives each coefficient of the frequency component as input and processes sine information of the input coefficient. For each coefficient of the frequency component converted into an absolute value,
The coefficient encoding means (5) performs statistical processing to encode the coefficients, and the sign information processing means (7) has:
Among the coefficients of each frequency component in the frequency component storage means (2), the intensity (hereinafter, the intensity) of each coefficient in the unit of frequency decomposition (DCT block) is measured, and the intensity of each coefficient is measured. Coefficient strength rank determining means (704) for determining a rank (hereinafter, coefficient strength rank); a coefficient strength rank table (705) for storing the coefficient strength rank; and determining whether each coefficient is zero. Only the non-zero determination sign information extracting means (701) for extracting the sign information, preferentially estimates the sign information from the coefficient having the highest coefficient strength rank, and generates and outputs new sign information from the estimation result. The signature information estimating means (706) performs a temporary sign information (hereinafter, referred to as a temporary sign information) for the coefficient of each frequency component in the frequency component storage means (2) inside the signature information estimating means (706). Sign And distribution), the true signature information (hereinafter optionally the sign information setting means (707) for setting a true signature information), signature information setting means (707)
The sign information-added coefficient storage means (708) for storing the temporary sign information and each coefficient (hereinafter, the coefficient with the sign information) for which the true sign information is set as necessary, and each frequency component of the coefficient with the sign information Frequency component synthesizing means (70) for synthesizing
9), a frequency synthesis result storage means (710) for storing a frequency synthesis result with signature information, and a first signature information estimation parameter corresponding to the signature synthesis result with signature information in which each temporary signature information is set (hereinafter referred to as a first parameter). , A first signature information estimation parameter), a first signature information estimation parameter calculation means (711), a first signature information estimation parameter storage means (712) for storing the first signature information estimation parameter, From the value of the first signature information estimation parameter corresponding to the above, it is estimated which temporary signature information is true signature information, and match / mismatch information indicating whether the estimated signature information matches true signature information, If it is not possible to estimate which temporary signature information is the true signature information, the new signature information The first signature information generating means (713) for outputting as
7) has a function of setting true signature information with respect to a coefficient for which estimation of whether or not the temporary signature information is true signature information is performed, and a coefficient in which the true signature information is set (hereinafter, a signature with true signature information Frequency synthesis result (hereinafter referred to as a true sine frequency synthesis result) and a frequency synthesis result based on a coefficient for which provisional signature information is set (hereinafter a coefficient with provisional sine information) (hereinafter a provisional sine frequency combination result).
The sum of the absolute values of the differences between the coefficients is referred to as a first variation, and the frequency synthesis result in the mixed state of the coefficient with the true signature information and the coefficient with the temporary signature information (hereinafter, the mixed frequency synthesis result) A first sine information estimation parameter calculating means (711), wherein the sum of the absolute values of the differences between the waveform and the central axis of the waveform in the region surrounded by the waveform is set as the first similarity. , A first change amount or a first similarity is calculated as a first signature information estimation parameter, and the first signature information generation means (713) calculates the first signature information estimation parameter corresponding to each temporary signature information into the first signature information estimation parameter. In the case of the amount of change, the temporary signature information when the value of the first signature information estimation parameter is small is estimated as true signature information, and the first signature information estimation parameter corresponding to each temporary signature information is the first signature information estimation parameter. If similarity is that the provisional signature information when the value of the first signature information estimation parameter is greater is configured to estimate that the true signature information.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (compression encoding side) according to the seventh embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 40).
01).

<DCT and Quantization> Next, in the frequency decomposition means (9 in FIG. 39), the input original image block is subjected to a DCT operation as shown in Equation 1 (S002 in FIG. 40), and the quantization is performed. In the converting means (10 in FIG. 39), D
Each coefficient (DCT block) after CT is quantized (FIG. 4).
0 S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 39) (FIG. 40).
S004). Here, as an example, all coefficients are
Uniform quantization is performed by applying 1 as a quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 39) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S005 in FIG. 40).

The above series of processing (S001 and S001 in FIG. 40)
002, S003, S004, and S005) are determined for all the original image blocks (S0 in FIG. 40).
06), until the processing of all the original image blocks is completed.

<Creation of Scan Order Table> Next, the scan order is determined by the priority order determining means (411 in FIG. 39) such that sampling is performed from the coefficient having the largest intensity value (sum of the absolute values of the coefficients). It is stored in the scan order table (402 in FIG. 39) (S007 in FIG. 40). Here, scan order information (eg, coordinate values) stored in the scan order table (402 in FIG. 39) is encoded by a scan order information encoding unit (403 in FIG. 39) and output as an output signal. (S008 in FIG. 40).
The encoding of the scan order information performed here simply involves the coordinate value of 2
A value expressed in a hexadecimal number may be output, or in image data or audio data, since a non-zero coefficient is unlikely to appear in a low-frequency component, a coordinate value closer to the low-frequency component has a shorter sign. A method such as outputting as a long code may be used.

<Execution and Encoding of Scan> Next, read control means (FIG. 3) is executed in accordance with the scan order stored in the created scan order table (402 in FIG. 39).
In 9-3), an address signal D of the frequency component storage means (2 in FIG. 39) is generated, and each coefficient of the quantized DCT block is converted from the frequency component storage means (2 in FIG. 39) to the last non-zero value. The coefficients are read out (S009 in FIG. 40). The read coefficient is converted into an absolute value by the absolute value conversion means (6 in FIG. 39) (S010 in FIG. 40), and each coefficient converted into the absolute value is converted into a coefficient encoding means (5 in FIG. 39). Then, encoding is performed (S011 in FIG. 40). That is, this encoding process (S011 in FIG. 40) is performed only on the absolute value part of the coefficient excluding the sign information. The processing for obtaining the absolute value of the coefficient may be performed before storing the coefficient in the frequency component storage means (2 in FIG. 39). In this case, each coefficient is stored in the frequency component storage means (2 in FIG. 39). , And may be stored separately for the absolute value and the sign information.

<Processing of Signature Information> Next, by the reading process in the scan order (S009 in FIG. 40), the coefficient to be read out is determined by the coefficient strength ranking determining means (7 in FIG. 39).
04), the energy intensity (hereinafter, intensity) of each coefficient is measured within one DCT block, and the intensity order of each coefficient (hereinafter, coefficient intensity order) is determined (S01 in FIG. 40).
2). Then, the value of the coefficient strength rank is stored in the coefficient strength rank table (705 in FIG. 39) (S01 in FIG. 40).
3). As described above, if the contour (edge) of the subject has a complicated shape, if the energy is concentrated on the high-frequency component or if the contour (edge) has a monotonous shape,
Energy concentrates on low frequency components. The frequency component where energy is concentrated reflects the characteristics of the input signal (image data) well, and the input signal (image data)
Of the above-described processing (FIG. 40).
S012) is for determining to which frequency component in the DCT block energy is concentrated.
The strength in this processing (S012 in FIG. 40) is specifically the magnitude of the absolute value of each coefficient or the square value of each coefficient. As an example, in the present embodiment (seventh embodiment of the present invention), the intensity is defined as the absolute value of the coefficient. It should be noted that the determination of the strength order is based on A in the DCT block of interest.
Perform only for the C component. For example, in the case of the same DCT block (FIG. 25) used in the description of the fifth embodiment of the present invention, a coefficient having the largest intensity value among AC components (hereinafter, referred to as a coefficient A1). Is the coordinates (1,0)
"-784". The coefficient having the second largest intensity value (hereinafter, referred to as coefficient A2) is “212” at the coordinates (3, 0). The coefficient having the third largest intensity value (hereinafter, referred to as coefficient A3) is represented by “−” of coordinates (5, 0).
103 ". The coefficient having the fourth largest intensity value (hereinafter, referred to as coefficient A4) is “71” of the coordinates (2, 4).
"And two coefficients having the same magnitude in intensity at" -71 "of the coordinates (4, 1). In such a case, either of them may be prioritized, but in this example, “71” of the coordinates (2, 4) is the fourth (coefficient A4). As described above, the ranking is performed in the order of the magnitudes, the coefficient strength ranking is determined, and the information of the coordinates (n, m) corresponding to the ranking is stored in the coefficient strength ranking table (7 in FIG. 39).
05). The above processing (S001 in FIG. 40)
To S013) are the same as in the above-described sixth embodiment of the present invention, except for the following processing.

Next, whether or not each coefficient is non-zero is determined by the non-zero determination sign information extracting means (701 in FIG. 39) (S014 in FIG. 40). Based on (706 in FIG. 39), it is estimated whether the sign information of each coefficient is positive or negative (FIG. 40).
S015 to S023). In the sign information estimation process (S015 to S023 in FIG. 40), information of a coefficient having a high strength order (a large strength value) is used as a clue in accordance with the coefficient strength order obtained earlier, and the strength order is low (strength). Is estimated). The sign information of the coefficient is estimated. In contrast to the example of the DCT block (FIG. 25), in the above-described sixth embodiment of the present invention, the information of the 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength rank is used as a clue to determine the highest strength. Although the sign information is estimated only for the coefficient A63 having the lower rank (“−2” of the coordinates (4, 0) in FIG. 25), in the seventh embodiment of the present invention, the sign information is estimated from the coefficient A1 having the higher strength rank. The sign information of all the coefficients up to the coefficient A63 is estimated in order.

First, the sign information setting means (70 in FIG. 39)
In 7), true sign information is set to the DC component, and the AC component of the coefficient A1 (“−784” of the coordinates (1, 0) in FIG. 25) for estimating the sign information is added as the temporary sign information. Is set (S015 in FIG. 40), and this DCT block (hereinafter referred to as DCT block 0) is stored in the coefficient storage unit with sign information (708 in FIG. 39) (S016 in FIG. 40). The sign information setting means (707 in FIG. 39) sets the true sign information to the DC component, and estimates the AC component of the coefficient A1 for estimating the sign information (“−784 of the coordinates (1, 0) in FIG. 25). )), Negative sign information is set as temporary sign information (S015 in FIG. 40), and this DCT block (hereinafter, referred to as DCT block 1) is stored in the coefficient storage unit with sign information (708 in FIG. 39). (S016 in FIG. 40). That is,
In the coefficient storage unit with signature information (708 in FIG. 39),
DCT block 0 and DCT block 1
T blocks are stored. Each DCT block is shown in FIGS.

Next, a coefficient storage unit with signature information (FIG. 3)
9 of the DCT block 0 (FIG. 4).
1) and two DCTs of DCT block 1 (FIG. 42).
Frequency synthesis means for the block (709 in FIG. 39)
The inverse DCT operation of Equation 2 is performed (S01 in FIG. 40).
7) The respective frequency synthesis results (inverse DCT results) are stored in the frequency synthesis result storage means (710 in FIG. 39) (S018 in FIG. 40). In addition, before performing the above-described inverse DCT operation (S017 in FIG. 40), it is preferable to perform the inverse quantization on each DCT block. 1, and the process of inverse quantization is omitted here. (An inverse quantization means may be provided to perform inverse quantization.)
Inverse DCT is a real number calculation, and there are values after the decimal point. However, usually, the result of the real number calculation is handled as an integer.
At this time, the value after the decimal point is truncated,
An error occurs with respect to the actual calculation result. In order to reduce this error as much as possible, the above inverse DCT operation (FIG. 40)
S017) is an example in which the inverse DCT result is 10000.
The output is multiplied and output as an integer value with higher bit precision. FIG. 43 shows a result obtained by multiplying the inverse DCT result of DCT block 0 (FIG. 41) by 10000 (hereinafter referred to as “inverse DCT block 0”).
FIG. 44 shows a result obtained by multiplying the T result by 10,000 (hereinafter, referred to as an inverse DCT block 1).

Next, the first signature information estimation parameter calculation means (711 in FIG. 39) calculates the first similarity as the first signature information estimation parameter (S01 in FIG. 40).
9), storing the calculation result in the first signature information estimation parameter storage means (712 in FIG. 39) (S02 in FIG. 40)
0). The first similarity calculated as described above is a frequency synthesis result (hereinafter, temporary signature frequency synthesis result) based on a coefficient for which temporary signature information is set (hereinafter, a coefficient with temporary signature information).
Is defined as the sum of the absolute values of the differences between the waveform and the central axis of the waveform (in this example, the average value). Specifically, in the above example, the difference between the waveform of the inverse DCT block 0 (FIG. 43) when the coefficient A1 is temporarily set to positive sine information and the center axis (the average value in this example) of this waveform. Is the positive first similarity, and the inverse DCT block 1 when the coefficient A1 is temporarily set to negative sign information (FIG. 4).
4) Waveform and center axis of this waveform (average value in this example)
And the sum of the absolute values of the differences with “negative first similarity”
The values of the “positive first similarity” and the “negative first similarity” are stored in the first sign information estimation parameter storage unit (712 in FIG. 39) (S020 in FIG. 40).
In the above example, the value of “positive first similarity” is “56
823328 ", and the value of" negative first similarity "is" 5683328 ", both of which are equal. FIG. 45 shows the relationship between the waveform of the inverse DCT block and the first similarity in a conceptual image.

Next, the first sign information generating means (713 in FIG. 39) compares the magnitudes of “positive first similarity” and “negative first similarity” and finds that the first similarity is “ It is estimated that the sign information of the "larger" one is the true sign information of the coefficient A1. If the estimation result is correct, “zero (1 [bit])” is output as new signature information. If the estimation result is incorrect, “signature” is output as new signature information.
1 (1 [bit]) ", and when the magnitudes of the" positive first similarity "and the" negative first similarity "are equal, it is determined that" the estimation is impossible "and the true sign of the coefficient A1 is obtained. The information itself is output (S021 in FIG. 40). In the above example, the value of “positive first similarity” is “56832”.
328 ”and the value of“ negative first similarity ”is“ 56
832328 ", and both are equal values, and the coefficient A
The signature information of No. 1 is "cannot be estimated" and the true signature information itself is output as the signature information. That is, since the true sine information of the coefficient A1 is negative, “1 (1 [bit])” meaning negative is output. The signature information output from the first signature information generation means (713 in FIG. 39) is stored in the signature information storage means (702 in FIG. 39) (S022 in FIG. 40).

In the present embodiment (seventh embodiment of the present invention), the first signature information estimation parameter
Using the similarity, the signature information with the first similarity “larger” is estimated to be true signature information. However, the estimation processing of the seventh embodiment of the present invention is based on the sixth embodiment of the present invention. In the embodiment, the first signature information estimation parameter
The process of using the amount of change and estimating the sign information with the first smaller amount of “small” as true sign information is substantially the same as the process, except for the way of expression. The reason will be described with reference to FIG. In FIG.
Reference inverse DCT used in the description of the sixth embodiment of the present invention
An example of the sum and difference between the block and the DCT block 0 and the sum and difference between the reference inverse DCT block and the DCT block 1 are shown. Set the value of each pixel of the reference inverse DCT block to "
a, b, c, d "and DCT block 0 (positive wave)
Is defined as “e, f, g, h”, the reference inverse D
The sum of the pixels of the CT block and the DCT block 0 (positive wave) (the first similarity having a sum of the absolute values of the sum is positive) and the difference of the pixels (the sum of the absolute values of the differences is a positive value) The first change amounts) are as shown in FIG. In the lower table of FIG. 46, each pixel of DCT block 1 (negative wave)
"-E, -f, -g, -h" and the reference inverse DCT
Block and the sum of each pixel of DCT block 1 (negative wave) (the sum of the absolute value of this sum is negative first similarity) and the difference of each pixel (the sum of the absolute value of this difference is negative 1 change amount), the positive first similarity and the negative first change amount have the same value, and the positive first change amount and the negative first similarity have the same value. is there. As can be seen from this example, the seventh embodiment of the present invention in which the first similarity is used as the first signature information estimation parameter and the sign information having the first similarity “larger” is estimated to be true signature information. The estimation process of the embodiment of
Estimation processing according to the sixth embodiment of the present invention in which the first change amount is used as the first sign information estimation parameter, and the sign information whose first change amount is “smaller” is estimated to be true sign information; The only difference is in the way they are expressed, but they have virtually the same meaning.

As described above, the estimation process of the sign information of the coefficient A1 having the highest strength rank has been completed.
It is determined whether or not (“212” of the coordinates (3, 0) in FIG. 25) is non-zero. Since it is non-zero, the following signature information is estimated.

First, the signature information setting means (70 in FIG. 39)
In 7), the true sign information is set to the DC component (it has already been set and may be used as it is), and the coefficient A1 (coordinates (1, 0) in FIG. 25) for which the estimation of the sign information has already been completed.
The coefficient A2 (“212” of the coordinates (3, 0) in FIG. 25) for setting negative sign information as true sign information in “−784” of FIG.
In the DCT block (hereinafter, DCT), positive sign information is set as temporary sign information (S015 in FIG. 40).
(Referred to as block 0) is stored in the coefficient storage unit with signature information (708 in FIG. 39) (S016 in FIG. 40). The signature information setting means (707 in FIG. 39)
The true sign information is set for the component (it has already been set, so it can be used as it is), and the coefficient A1 for which estimation of the sign information has already been completed (“−784” of the coordinates (1, 0) in FIG. 25).
)), Negative sign information is set as true sign information, and a coefficient A2 (FIG. 25) for estimating the sign information from this is set.
The negative sign information is set as the temporary sign information at “212” of the coordinates (3, 0) (S01 in FIG. 40).
5) This DCT block (hereinafter referred to as DCT block 1) is stored in a coefficient storage unit with signature information (70 in FIG. 39).
8) (S016 in FIG. 40). That is, the coefficient storage unit with signature information (708 in FIG. 39) stores the DC
Two DCT blocks, a T block 0 and a DCT block 1, are stored. Each DCT block is shown in FIGS.

The processing from S016 to S022 in FIG. 40 is performed on DCT block 0 (FIG. 47) and DCT block 1 (FIG. 48) newly updated as described above.

Coefficient storage means with signature information (7 in FIG. 39)
08) DCT block 0 (FIG. 47)
Then, the two DCT blocks of DCT block 1 (FIG. 48) are subjected to the inverse DCT operation of Expression 2 by frequency synthesis means (709 in FIG. 39) (S017 in FIG. 40),
Each frequency synthesis result (inverse DCT result) is stored in the frequency synthesis result storage means (710 in FIG. 39) (FIG. 40).
S018). The above inverse DCT operation (S in FIG. 40)
017) is desirably configured to perform inverse quantization on each DCT block, but in the above series, the quantization step is set to 1 as an example,
Here, the processing of the inverse quantization is omitted. (An inverse quantization means may be provided to perform the inverse quantization.) Note that the inverse DCT is a real number calculation and has a value below the decimal point. . At this time, the value after the decimal point is truncated, causing an error with respect to the actual operation result. In order to minimize this error, the above-described inverse DCT operation (S017 in FIG. 40), for example, converts the inverse DCT result to 1
0000 times and output as an integer value with higher bit precision. Inverse D of DCT block 0 (FIG. 47)
FIG. 49 shows a result obtained by multiplying the CT result by 10000 (hereinafter referred to as an inverse DCT block 0).
FIG. 50 shows a result obtained by multiplying the inverse DCT result of 8) by 10000 (hereinafter, referred to as an inverse DCT block 1).

Next, the first signature information estimation parameter calculation means (711 in FIG. 39) calculates the first similarity as the first signature information estimation parameter (S01 in FIG. 40).
9), storing the calculation result in the first signature information estimation parameter storage means (712 in FIG. 39) (S02 in FIG. 40)
0). The first similarity calculated as described above is a frequency synthesis result (hereinafter, temporary signature frequency synthesis result) based on a coefficient for which temporary signature information is set (hereinafter, a coefficient with temporary signature information).
Is defined as the sum of the absolute values of the differences between the waveform and the central axis of the waveform (in this example, the average value). Specifically, in the above example, the difference between the waveform of the inverse DCT block 0 (FIG. 49) when the coefficient A2 is set to positive sine information and the central axis (in this example, the average value) of this waveform. Is the positive first similarity, and the inverse DCT block 1 (FIG. 5) when the coefficient A2 is temporarily set to negative sign information.
0) and the center axis of this waveform (in this example, the average value)
And the sum of the absolute values of the differences with “negative first similarity”
The values of the “positive first similarity” and the “negative first similarity” are stored in the first sign information estimation parameter storage unit (712 in FIG. 39) (S020 in FIG. 40).
In the above example, the value of “positive first similarity” is “62”.
228808 ", and the value of" negative first similarity "is" 51435816 ".

Next, the first sign information generating means (713 in FIG. 39) compares the magnitudes of “positive first similarity” and “negative first similarity” and finds that the first similarity is “ It is estimated that the sign information of the "larger" one is the true sign information of the coefficient A2. If the estimation result is correct, “zero (1 [bit])” is output as new signature information. If the estimation result is incorrect, “signature” is output as new signature information.
1 (1 [bit]) ", and if the magnitudes of the" positive first similarity "and the" negative first similarity "are equal, it is determined to be" impossible to estimate "and the true sign of the coefficient A2 The information itself is output (S021 in FIG. 40). In the above example, the value of the “positive first similarity” is “62228”.
808 ”and the value of“ negative first similarity ”is“ 51
435816 ”, the“ positive first similarity ”is larger, so the true sign information is estimated to be“ positive ”.
This estimation result is a correct answer, and "zero (1 [bit])" which means a correct answer is output as new sign information. The signature information output from the first signature information generation means (713 in FIG. 39) is stored in the signature information storage means (702 in FIG. 39) (S022 in FIG. 40).

The above processing (from S014 to S02 in FIG. 40)
Until 2) is completed for all the coefficients in one DCT block, or the processing is performed in the order of the strength order, and the processing is repeated until the last non-zero coefficient (S023 in FIG. 40).
(In the example of the DCT block in FIG. 25, since no zero coefficient exists, the processing is repeated up to the coefficient A63.)

As described above, the DCT block (FIG. 25)
, The coordinates (n, m) of each coefficient when the sign information is estimated up to all the coefficients (coefficients A1 to A63)
The value of each coefficient, the positive first similarity, the negative first similarity, the estimation result, whether the estimation result is a correct answer (O) or an incorrect answer (X), or newly generated signature information (0, In case of incorrect answer 1,
Tables 8, 9, and 10 show 0 as a sign information and 1) as a negative sign information as it is when estimation is impossible.

[0245]

[Table 12]

[0246]

[Table 13]

[0247]

[Table 14]

The above processing (from S009 to S02 in FIG. 40)
Step 3) is repeated until completion for all DCT blocks (S024 in FIG. 40).

Finally, signature information storage means (7 in FIG. 39)
02) is compression-encoded by the signature information encoding means (703 in FIG. 39) (S025 in FIG. 40). This encoding process (S0 in FIG. 40)
25), as in the above example, in each DCT block, the sign information is estimated by estimating the sign information of the coefficient having the lower strength rank in the order of the coefficient having the higher strength rank based on the information of the coefficient having the higher strength rank. Since “0” increases as the value of the information, the sign amount can be reduced by collecting the sign information, performing statistical processing, and encoding. As a method for compressing and encoding the sign information, any encoding method such as conventional Huffman encoding or arithmetic encoding may be used.

Table 11 shows the DCT block (FIG. 2).
5) As a totaling result of the estimation of the signature information, the total number (total number of coefficients), the correct answer (number and percentage of the total number), the incorrect answer (number and percentage of the total number), the impossible estimation (number and Sign information 0 (the number using 0 as sign information and the ratio in the total number), sign information 1 (number using 1 as the sign information and the ratio in the total number), signature The information amount of information (the number of bits and the ratio to the total number of bits (63 bits)) is shown. As a result, in the prior art, 1 [bit] was originally used for one coefficient as the signature information.
t], but as described above, by estimating the signature information and treating it as new signature information, the information amount becomes about 49.9 [bit]. About 13.1 [bit] (about 20.8
%), The amount of information can be reduced.

[0251]

[Table 15]

That is, based on the same concept as in the sixth embodiment of the present invention, the sign information is estimated in order from the coefficient with the highest strength order (the coefficient with the largest amplitude absolute value), and the following is performed. By using it as a clue for estimating the sign information of the coefficient, it is possible to estimate the sign information of many coefficients, and by compressing and encoding the estimation result, the sign information remains lossless, The code amount required for information can be greatly reduced.

<Eighth Embodiment> Next, an eighth embodiment of the present invention for solving the second problem of the conventional signal processing system will be described.

In the fifth embodiment of the present invention described above, the sign information is collectively encoded by the sign information processing means (7 in FIG. 22), thereby reducing the code amount of the sign information. Solved 2 problems. In the sixth embodiment of the present invention, in each DCT block, estimation of the sign information of the coefficient A63 having the lowest strength rank is performed based on information of 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength rank. Then, the sign information of the estimated coefficient A63 is collected and collectively encoded, thereby reducing the code amount of the sign information and solving the second problem. Seventh of the present invention
In the embodiment of the present invention, in each DCT block, signature information of not only the coefficient A63 having the lowest strength rank but also all the coefficients A1 to A63 is estimated based on the information of the coefficient having the highest strength rank. By encoding,
The second problem has been solved by greatly reducing the code amount of the sign information. In the seventh embodiment of the present invention, the first similarity is used as the first signature information estimation parameter, and the signature information when the first similarity is large is estimated to be the correct answer. In the eighth embodiment of the present invention, similar to the seventh embodiment of the present invention, the first similarity is used as the first signature information estimation parameter, and further, the second similarity information estimation parameter is used. , By using the value of the degree of difference (details will be described later), compared with the seventh embodiment of the present invention,
This is to solve the second problem by greatly reducing the code amount of the sign information.

The image data as the input signal is obtained by converting a signal of a photograph or the like by a video camera or the like, or by converting a signal of a photograph or the like by a scanner. Things. Thus, the input signal (image data) is generated based on optical natural phenomena. Therefore, the contour (edge) of the subject
Information such as the shape of the input signal (image data)
Is reflected in the numerical value. These input signals (image data)
In the signal processing systems according to the related art and the present invention in which the frequency is resolved after the frequency decomposition, the energy is biased to one of the frequency components after the frequency decomposition depending on the shape of the contour (edge) of the subject. For example, if the contour (edge) of the subject has a complicated shape, energy concentrates on high frequency components, and if the contour (edge) has a monotonous shape, energy concentrates on low frequency components. I do. In other words, the frequency components where energy is concentrated reflect the characteristics of the input signal (image data) well and contain much information on the input signal (image data). In view of such characteristics of the input signal (image data), in the eighth embodiment of the present invention, the following FIG.
With the configuration as in 1, the code amount of the signature information is reduced. As in the fifth, sixth, and seventh embodiments of the present invention, the coefficients other than the sign information, except for the sign information as a mere absolute value, are the same as those of the prior art or the already described present invention. Encoding may be performed by the same processing as in the first, second, third, and fourth embodiments. However, in the eighth embodiment of the present invention, as an example, the absolute value part of the coefficient excluding the sign information is used. Will be described as performing the same processing as in the second embodiment of the present invention.

The configuration of the eighth embodiment of the present invention will be described below with reference to FIG. In FIG. 51, 9
Is frequency decomposition means, 10 is quantization means, 1
Is a basic encoding unit, 2 is a frequency component storage unit, 3 is a coefficient readout control unit, and 4 is a readout control information generation unit inside the basic encoding unit (1).
Then, inside the read control information generating means (4),
401 is a scan order determining means, 402 is a scan order table, 403 is a scan order information encoding means, and the scan order information E is used as read control information C, and the coefficient read control means (3) An address signal D is generated based on the read control information C, and 412 inside the scan order determining means (401)
411 is a priority order determining means.
The above functions basically in the same manner as the above-described second embodiment of the present invention. The difference from the second embodiment of the present invention is that an absolute value conversion means (6) for converting each coefficient of a frequency component into an absolute value is provided inside the basic coding means (1).
And a sine information processing means (7) which receives each coefficient of the frequency component as input and processes sine information of the input coefficient. For each coefficient of the frequency component converted into an absolute value,
The coefficient encoding means (5) performs statistical processing to encode the coefficients, and the sign information processing means (7) has:
Among the coefficients of each frequency component in the frequency component storage means (2), the intensity (hereinafter, the intensity) of each coefficient in the unit of frequency decomposition (DCT block) is measured, and the intensity of each coefficient is measured. Coefficient strength rank determining means (704) for determining a rank (hereinafter, coefficient strength rank); a coefficient strength rank table (705) for storing the coefficient strength rank; and determining whether each coefficient is zero. Only the non-zero determination sign information extracting means (701) for extracting the sign information, preferentially estimates the sign information from the coefficient having the highest coefficient strength rank, and generates and outputs new sign information from the estimation result. The signature information estimating means (706) performs a temporary sign information (hereinafter, referred to as a temporary sign information) for the coefficient of each frequency component in the frequency component storage means (2) inside the signature information estimating means (706). Sign And distribution), the true signature information (hereinafter optionally the sign information setting means (707) for setting a true signature information), signature information setting means (707)
The sign information-added coefficient storage means (708) for storing the temporary sign information and each coefficient (hereinafter, the coefficient with the sign information) for which the true sign information is set as necessary, and each frequency component of the coefficient with the sign information Frequency component synthesizing means (70) for synthesizing
9), a frequency synthesis result storage means (710) for storing a frequency synthesis result with signature information, and a first signature information estimation parameter corresponding to the signature synthesis result with signature information in which each temporary signature information is set (hereinafter referred to as a first parameter). , A first signature information estimation parameter), a first signature information estimation parameter calculation means (711), a first signature information estimation parameter storage means (712) for storing the first signature information estimation parameter, A second signature information estimation parameter calculation means (714) for calculating a second signature information estimation parameter (hereinafter, a second signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which is set; Second signature information estimation parameter storage means (7) for storing information estimation parameters
15) and using the first signature information estimation parameter and the second signature information estimation parameter, a third signature information estimation parameter (hereinafter, referred to as a third signature information parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. A third signature information estimation parameter calculating means (716) for calculating the third signature information estimation parameter; a third signature information estimation parameter storage means (717) for storing the third signature information estimation parameter; From the value of the third signature information estimation parameter thus obtained, which temporary signature information is the true signature information is estimated, and the match / mismatch information indicating whether the estimated signature information matches the true signature information is newly added. It has a second signature information generating means (718) for outputting as signature information, and has a signature information setting means (70).
7) has a function of setting true signature information with respect to a coefficient for which estimation of whether or not the temporary signature information is true signature information is performed, and a coefficient in which the true signature information is set (hereinafter, a signature with true signature information Frequency synthesis result (hereinafter referred to as a true sine frequency synthesis result) and a frequency synthesis result based on a coefficient for which provisional signature information is set (hereinafter a coefficient with provisional sine information) (hereinafter a provisional sine frequency combination result).
The sum of the absolute values of the differences between the coefficients is referred to as a first variation, and the frequency synthesis result in the mixed state of the coefficient with the true signature information and the coefficient with the temporary signature information (hereinafter, the mixed frequency synthesis result) The sum of the absolute value of the difference between the waveform and the central axis of the waveform in the region surrounded by the waveform and the central axis of the waveform is defined as the first similarity, and the waveform of the mixed frequency synthesis result is in a predetermined value range. And the waveform in the area
The sum of the absolute value of the difference from the value range is defined as the degree of difference, and the first signature information estimation parameter calculating means (711) performs
The change amount or the first similarity is calculated as the first signature information estimation parameter, and the second signature information estimation parameter calculating means (714) calculates the difference as the second signature information estimation parameter to calculate the first change amount or the first signature amount. The difference is normalized by a normalization coefficient, and a value obtained by adding the normalized difference to the normalized first change is defined as a second change.
A first similarity or dissimilarity is normalized by a normalization coefficient, and a value obtained by subtracting the normalized dissimilarity from the normalized first similarity is set as a second similarity. In the means (716), the second change amount or the second similarity is calculated as a third signature information estimation parameter, and in the second signature information generating means (718),
When the third signature information estimation parameter corresponding to each temporary signature information is the second change amount, the temporary signature information when the value of the third signature information estimation parameter is small is estimated as true signature information, and each temporary signature information is estimated. When the third signature information estimation parameter corresponding to the information is the second similarity, the temporary signature information when the value of the third signature information estimation parameter is large is estimated to be the true signature information. .

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the signal processing system (compression encoding side) according to the eighth embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 52).
01).

<DCT and Quantization> Next, in the frequency decomposition means (9 in FIG. 51), the input original image block is subjected to a DCT operation as shown in Expression 1 (S002 in FIG. 52), and the quantization is performed. In the conversion means (10 in FIG. 51), D
Each coefficient (DCT block) after CT is quantized (FIG. 5).
2 S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 51) (FIG. 52).
S004). Here, as an example, all coefficients are
Uniform quantization is performed by applying 1 as a quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 51) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S005 in FIG. 52).

The above series of processing (S001 and S001 in FIG. 52)
002, S003, S004, and S005) are determined for all the original image blocks (S0 in FIG. 52).
06), until the processing of all the original image blocks is completed.

<Creation of Scan Order Table> Next, the priority order determining means (411 in FIG. 51) determines a scan order in which sampling is performed preferentially from a coefficient having a larger intensity value (sum of absolute values of coefficients). Determined and stored in the scan order table (402 in FIG. 51) (S00 in FIG. 52).
7). Here, the scan order table (402 in FIG. 51)
Scan order information (eg, coordinate values) stored in
Is encoded by the scan order information encoding means (403 in FIG. 51) and output as an output signal (S in FIG. 52).
008). In the encoding of the scan order information performed here, a value obtained by simply expressing a coordinate value in a binary number may be output, or a non-zero coefficient hardly appears in a low frequency component in image data or audio data. Alternatively, a method of outputting a variable-length code such that a code becomes shorter as the coordinate value is closer to the low-frequency component may be used.

<Execution and Encoding of Scan> Next, read control means (FIG. 5) is executed in accordance with the scan order stored in the created scan order table (402 in FIG. 51).
In 1) 3), an address signal D of the frequency component storage means (2 in FIG. 51) is generated, and each coefficient of the quantized DCT block is converted from the frequency component storage means (2 in FIG. 51) to the last non-zero value. The coefficients are read out (S009 in FIG. 52). The read coefficients are converted into absolute values by the absolute value conversion means (6 in FIG. 51) (S010 in FIG. 52), and the coefficients converted into the absolute values are converted into coefficient coding means (5 in FIG. 51). Then, encoding is performed (S011 in FIG. 52). That is, this encoding process (S011 in FIG. 52) is performed only on the absolute value part of the coefficient excluding the sign information. The process of obtaining the absolute value of the coefficient may be performed before storing the coefficient in the frequency component storage means (2 in FIG. 51). In this case, each coefficient is stored in the frequency component storage means (2 in FIG. 51). , And may be stored separately for the absolute value and the sign information.

<Processing of Signature Information> Next, by the reading process in the scan order (S009 in FIG. 52), the coefficient to be read is determined by the coefficient strength ranking determining means (7 in FIG. 51).
04), the energy intensity (hereinafter, intensity) of each coefficient is measured within one DCT block, and the order of the intensity of each coefficient (hereinafter, coefficient intensity order) is determined (S01 in FIG. 52).
2). Then, the value of the coefficient strength rank is stored in the coefficient strength rank table (705 in FIG. 51) (S01 in FIG. 52).
3). As described above, if the contour (edge) of the subject has a complicated shape, if the energy is concentrated on the high-frequency component or if the contour (edge) has a monotonous shape,
Energy concentrates on low frequency components. The frequency component where energy is concentrated reflects the characteristics of the input signal (image data) well, and the input signal (image data)
52, the above processing (FIG. 52).
S012) is for determining to which frequency component in the DCT block energy is concentrated.
The intensity in this processing (S012 in FIG. 52) is specifically the absolute value of each coefficient or the magnitude of the square of each coefficient. As an example, in the present embodiment (the eighth embodiment of the present invention), the intensity is defined as the absolute value of the coefficient. It should be noted that the determination of the strength order is based on A in the DCT block of interest.
Perform only for the C component. For example, in the case of the same DCT block (FIG. 25) used in the description of the fifth embodiment of the present invention, a coefficient having the largest intensity value among AC components (hereinafter, referred to as a coefficient A1). Is the coordinates (1,0)
"-784". The coefficient having the second largest intensity value (hereinafter, referred to as coefficient A2) is “212” at the coordinates (3, 0). The coefficient having the third largest intensity value (hereinafter, referred to as coefficient A3) is represented by “−” of coordinates (5, 0).
103 ". The coefficient having the fourth largest intensity value (hereinafter, referred to as coefficient A4) is “71” of the coordinates (2, 4).
"And two coefficients having the same magnitude in intensity at" -71 "of the coordinates (4, 1). In such a case, either of them may be prioritized, but in this example, “71” of the coordinates (2, 4) is the fourth (coefficient A4). In this way, the ranking is performed in the order of the magnitude, and the coefficient strength ranking is determined, and the information of the coordinates (n, m) corresponding to the ranking is stored in the coefficient strength ranking table (7 in FIG. 51).
05).

Next, whether or not each coefficient is non-zero is determined by the non-zero determination sign information extracting means (701 in FIG. 51) (S014 in FIG. 52). According to (706 in FIG. 51), it is estimated whether the sign information of each coefficient is positive or negative (FIG. 52).
S015 to S026). In the sign information estimation process (S015 to S026 in FIG. 52), in accordance with the coefficient strength rank obtained earlier, the strength rank is low (strength value is large), using the information of the coefficient having the high strength rank (large strength value) as a clue. (Small value) is estimated. In contrast to the example of the DCT block (FIG. 25), in the above-described sixth embodiment of the present invention, the information of the 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength rank is used as a clue to determine the highest strength. The estimation of the sign information of only the coefficient A63 having a low rank (“−2” of the coordinates (4, 0) in FIG. 25) was performed, but the eighth embodiment of the present invention is different from the seventh embodiment of the present invention. In the same manner as in the embodiment, the sign information of all the coefficients up to the coefficient A63 is estimated in order from the coefficient A1 having the highest strength order.

First, signature information setting means (70 in FIG. 51)
In 7), the true sign information is set to the DC component, and the AC component of the coefficient A1 (“−784” of the coordinates (1, 0) in FIG. 25) for estimating the sign information is added as the temporary sign information. (S015 in FIG. 52), and this DCT block (hereinafter referred to as DCT block 0) is stored in the coefficient storage unit with signature information (708 in FIG. 51) (S016 in FIG. 52). The sign information setting means (707 in FIG. 51) sets the true sign information to the DC component, and estimates the AC component of the coefficient A1 for estimating the sign information (“−784 of the coordinates (1, 0) in FIG. 25). )), Negative sign information is set as temporary sign information (S015 in FIG. 52), and this DCT block (hereinafter, referred to as DCT block 1) is stored in the coefficient storage unit with sign information (708 in FIG. 51). (S016 in FIG. 52). That is,
In the coefficient storage unit with signature information (708 in FIG. 51),
DCT block 0 and DCT block 1
T blocks are stored. Each DCT block is shown in FIGS.

Next, coefficient storage means with signature information (FIG. 5)
DCT block 0 (FIG. 5)
3) and two DCTs of DCT block 1 (FIG. 54).
For the block, frequency synthesis means (709 in FIG. 51)
The inverse DCT operation of Equation 2 is performed (S01 in FIG. 52).
7), storing each inverse DCT result in the frequency synthesis result storage means (710 in FIG. 51) (S01 in FIG. 52).
8). Note that the above inverse DCT operation (S017 in FIG. 52)
It is preferable that the inverse quantization is performed on each DCT block before performing the inverse quantization. However, in the above series, the quantization step is set to 1 as an example, and the inverse quantization process is omitted here. did. (An inverse quantization means may be provided to perform the inverse quantization.) Note that the inverse DCT is a real number calculation and has a value below the decimal point. . At this time, the value after the decimal point is truncated, causing an error with respect to the actual operation result. In order to minimize this error, the above-described inverse DCT operation (S017 in FIG. 52), for example, converts the inverse DCT result to 1
0000 times and output as an integer value with higher bit precision. Inverse DC of DCT block 0 (FIG. 53)
FIG. 55 shows a result obtained by multiplying the T result by 10,000 (hereinafter referred to as an inverse DCT block 0), and a DCT block 1 (FIG. 54)
10000 times the inverse DCT result of
FIG. 56 shows T block 1).

Next, the first signature information estimation parameter calculation means (711 in FIG. 51) calculates the first similarity as the first signature information estimation parameter (S01 in FIG. 52).
9), and store the calculation result in the first signature information estimation parameter storage means (712 in FIG. 51) (S02 in FIG. 52).
0). The first similarity calculated as described above is a frequency synthesis result (hereinafter, temporary signature frequency synthesis result) based on a coefficient for which temporary signature information is set (hereinafter, a coefficient with temporary signature information).
Is defined as the sum of the absolute values of the differences between the waveform and the central axis of the waveform (in this example, the average value). Specifically, in the above example, the difference between the waveform of the inverse DCT block 0 (FIG. 55) when the coefficient A1 is temporarily set to the positive sign information and the center axis (in this example, the average value) of this waveform. Is the positive first similarity, and the inverse DCT block 1 when the coefficient A1 is temporarily set to negative sign information (FIG. 5).
6) and the central axis of this waveform (average value in this example)
And the sum of the absolute values of the differences with “negative first similarity”
The values of the “positive first similarity” and the “negative first similarity” are stored in the first signature information estimation parameter storage unit (712 in FIG. 51) (S020 in FIG. 52).
In the above example, the value of “positive first similarity” is “56
823328 ", and the value of" negative first similarity "is" 5683328 ", both of which are equal. FIG. 57 shows the relationship between the waveform of the inverse DCT block and the first similarity in a conceptual image.

The above processing (from S014 to S02 in FIG. 52)
0) is exactly the same processing as in the seventh embodiment of the present invention, except for the following processing.

Next, the second sign information estimation parameter calculation means (714 in FIG. 51) calculates the degree of difference as the second signature information estimation parameter (S02 in FIG. 52).
1), storing the calculation result in the second signature information estimation parameter storage means (715 in FIG. 51) (S02 in FIG. 52)
2). The degree of difference calculated as described above is determined based on the area where the waveform of the frequency synthesis result (hereinafter, temporary signature frequency synthesis result) based on the coefficient for which the temporary signature information is set (hereinafter, the coefficient with temporary signature information) is out of the predetermined value range. Is defined as the sum of the absolute values of the differences between the waveform and the value range. In the above example, the range of the input original image data is “0 to 25”.
5 ". However, the inverse DCT block 0 (FIG. 5)
5) and the inverse DCT block 1 (FIG. 56) both have a value of 10,000 times at the time of the inverse DCT operation.
Here, the value range is also 10,000 to match the range.
Treated as a double value (0 to 2550000). Coefficient A1
The waveform of the inverse DCT block 0 (FIG. 55) when positive sign information is temporarily set to a predetermined value range (in this example, “
0 to 2550000 ") and the waveform and value range (in this example," 0 to 2550000 ").
The sum of the absolute value of the difference between the coefficient and the inverse DCT block 1 (FIG. 56) when the coefficient A1 is temporarily set to the negative sign information has a predetermined value range (this In the example, the waveform and the value range (“0 to 25
The sum of the absolute value of the difference from the positive difference and the negative difference is stored in the second sign information estimation parameter storage means (FIG. 51 (715 of FIG. 52).
2). In the above example, the value of "positive difference" is "1
348784 "and the value of" negative difference "is" 1
348784 ", which are both equal values. Inverse DC
FIG. 58 shows a conceptual image of the relationship between the waveform of the T block and the degree of difference.

In the seventh embodiment of the present invention described above, signature information is estimated using only the first signature information estimation parameter (first similarity or first variation). In the embodiment (the eighth embodiment of the present invention),
By using information of the second signature information estimation parameter (degree of difference) together, the accuracy of signature information estimation is improved and finally the code amount of signature information is significantly increased as compared with the seventh embodiment of the present invention. It is to reduce. The reason is as follows. In the seventh embodiment of the present invention, the sign information having the larger value of the first similarity is estimated to be the correct answer. This means that as shown in FIG. 57, it is estimated that the farther the waveform is from the central axis, the more correct the waveform. However, the range of the original image data is originally “0 to 255”.
In the situation where "", the sign information is considered to be unnatural if it goes beyond this range. Therefore, the sign information is estimated in consideration of both the first similarity (or the first amount of change) and the difference by using the amount of the difference outside the range as the difference, thereby obtaining a more accurate estimation result. Obtainable. Specifically, when the first signature information estimation parameter is the first change amount,
The first change amount or the degree of difference is normalized by a normalization coefficient, and a value obtained by adding the normalized difference amount to the normalized first change amount is set as a second change amount. , The first similarity is normalized by multiplying the first similarity or the dissimilarity by a normalization coefficient, and the value obtained by subtracting the normalized dissimilarity from the normalized first similarity is used as the second similarity. The third sign information estimation parameter calculating means (716 in FIG. 51) calculates the second variation or the second similarity as the third sign information estimation parameter (S023 in FIG. 52), and generates the second sign information. Means (71 in FIG. 51)
In 8), when the third signature information estimation parameter corresponding to each temporary signature information is the second variation, the temporary signature information when the value of the third signature information estimation parameter is small is estimated as true signature information. If the third signature information estimation parameter corresponding to each temporary signature information is the second similarity,
The temporary signature information when the value of the third signature information estimation parameter is large is estimated to be true signature information, and if the estimation result is correct, “zero (1)” is set as new signature information.
[Bit])], and if the estimation result is incorrect, “1 (1 [bit]
]) Is output, and when estimation is impossible, the true sign information itself of the coefficient A1 is output (S02 in FIG. 52).
5).

In the present embodiment (eighth embodiment of the present invention), the second signature information
The similarity is calculated, and specifically, in order to normalize the first similarity and the difference, the sum of the positive first similarity and the negative first similarity (hereinafter, the sum of the first similarity) ) And the sum of the positive and negative dissimilarities (hereinafter referred to as the sum of dissimilarities), and as a normalization coefficient α, the ratio of the sum of the first similarities and the sum of the dissimilarities (= (Sum of first similarity / sum of difference). Then, a value obtained by subtracting the value of the dissimilarity multiplied by α from the first similarity (= first similarity−α × dissimilarity) is obtained as the second similarity. In the above example (coefficient A1), the value of the normalization coefficient α is
“42”, and “positive second similarity (= positive first similarity)
(Similarity-α × positive difference)) is “183400
", The negative second similarity (= negative first similarity−α ×
The value of “negative dissimilarity” is “183400”, both of which are equal. The sign information of the coefficient A1 is “cannot be estimated”, and true sign information is output as sign information. That is, since the true sine information of the coefficient A1 is negative, “1 (1 [bit])” meaning negative is output.

If the second change amount is calculated as the third sign information estimation parameter, the sum of the positive first change amount and the negative first change amount (hereinafter, referred to as the sum of the first change amounts)
And the sum of the dissimilarities (= the sum of the positive dissimilarity and the negative dissimilarity), and as the normalization coefficient α, the ratio of the sum of the first changes and the sum of the dissimilarities (= the sum of the first changes) / Sum of differences). Then, a value obtained by adding the value of the difference degree multiplied by α to the first change amount (= first change amount + α × difference degree) may be obtained as the second change amount.

In the above description, the sum of the differences may be used as it is when the normalization coefficient α is obtained. However, the inverse DCT block may be used because of the quantization error and not using all the coefficients. Even if the sign information of is correct, it may be slightly out of range. Therefore, the normalization coefficient α
When calculating the difference, the value of the degree of difference is adjusted (subtracted or multiplied by a coefficient less than 1) or the normalization coefficient α itself is adjusted (subtracted) so as to have a slightly smaller value. Or a coefficient less than 1).

The second signature information generating means (71 in FIG. 51)
The sign information output from 8) is stored in the sign information storage means (702 in FIG. 51) (S02 in FIG. 52).
6).

In the present embodiment (eighth embodiment of the present invention), the second signature information
The similarity was used to estimate the sign information having the second similarity “larger” as the true sign information. However, when the second change amount is used as the third sign information estimation parameter, the second change amount is The process of estimating the sign information of the "small" one as true sign information is different from the process of expressing it,
It has substantially the same meaning.

As described above, the estimation processing of the sign information of the coefficient A1 having the highest strength rank has been completed.
It is determined whether or not (“212” of the coordinates (3, 0) in FIG. 25) is non-zero. Since it is non-zero, the following signature information is estimated.

First, signature information setting means (70 in FIG. 51)
In 7), the true sign information is set to the DC component (it has already been set and may be used as it is), and the coefficient A1 (coordinates (1, 0) in FIG. 25) for which the estimation of the sign information has already been completed.
The coefficient A2 (“212” of the coordinates (3, 0) in FIG. 25) for setting negative sign information as true sign information in “−784” of FIG.
In the DCT block (hereinafter, DCT), positive sign information is set as temporary sign information (S015 in FIG. 52).
Block 0) is stored in the coefficient storage unit with signature information (708 in FIG. 51) (S016 in FIG. 52). Also, the signature information setting means (707 in FIG. 51)
The true sign information is set for the component (it has already been set, so it can be used as it is), and the coefficient A1 for which estimation of the sign information has already been completed (“−784” of the coordinates (1, 0) in FIG. 25).
)), Negative sign information is set as true sign information, and a coefficient A2 (FIG. 25) for estimating the sign information from this is set.
Negative sign information is set as temporary sign information at “212” of the coordinates (3, 0) (S01 in FIG. 52).
5) This DCT block (hereinafter referred to as DCT block 1) is stored in a coefficient storage unit with signature information (70 in FIG. 51).
8) (S016 in FIG. 52). That is, the coefficient storage unit with signature information (708 in FIG. 51) stores the DC
Two DCT blocks, a T block 0 and a DCT block 1, are stored. Each DCT block is shown in FIGS.

Hereinafter, the newly updated DCT block 0
The processing of S016 to S026 in FIG. 52 is performed on (FIG. 59) and DCT block 1 (FIG. 60), respectively.

Coefficient storage means with signature information (7 in FIG. 51)
08) DCT block 0 (FIG. 59)
Then, the inverse DCT operation of Expression 2 is performed on the two DCT blocks of DCT block 1 (FIG. 60) by the frequency synthesizing means (709 in FIG. 51) (S017 in FIG. 52).
Each inverse DCT result is stored in the frequency synthesis result storage means (710 in FIG. 51) (S018 in FIG. 52). Before performing the above-described inverse DCT operation (S017 in FIG. 52), it is desirable to perform an inverse quantization on each DCT block. 1, and the process of inverse quantization is omitted here. (An inverse quantization means may be provided to perform the inverse quantization.) Note that the inverse DCT is a real number calculation and has a value below the decimal point. . At this time, the value after the decimal point is truncated, causing an error with respect to the actual operation result. In order to minimize this error, the above-described inverse DCT operation (S017 in FIG. 52), for example, converts the inverse DCT result to 1
0000 times and output as an integer value with higher bit precision. Inverse DC of DCT block 0 (FIG. 59)
FIG. 61 shows a result obtained by multiplying the T result by 10,000 (hereinafter referred to as an inverse DCT block 0), and FIG. 61 shows a DCT block 1 (FIG. 60).
10000 times the inverse DCT result of
62 is shown in FIG.

Next, the first signature information estimation parameter calculation means (711 in FIG. 51) calculates the first similarity as the first signature information estimation parameter (S01 in FIG. 52).
9), and store the calculation result in the first signature information estimation parameter storage means (712 in FIG. 51) (S02 in FIG. 52).
0). The first similarity calculated as described above is a frequency synthesis result (hereinafter, temporary signature frequency synthesis result) based on a coefficient for which temporary signature information is set (hereinafter, a coefficient with temporary signature information).
Is defined as the sum of the absolute values of the differences between the waveform and the central axis of the waveform (in this example, the average value). Specifically, in the above example, the difference between the waveform of the inverse DCT block 0 (FIG. 61) when the coefficient A2 is set to positive sine information and the central axis (in this example, the average value) of this waveform. Is the positive first similarity, and the inverse DCT block 1 (FIG. 6
2) Waveform and center axis of this waveform (average value in this example)
The sum of the absolute value of the difference between the first sign information estimation parameter and the negative first similarity is stored in the first sign information estimation parameter storage. This is stored in the means (712 in FIG. 51) (S020 in FIG. 52).
In the above example, the value of “positive first similarity” is “62”.
228808 ", and the value of" negative first similarity "is" 51435816 ".

Next, the second sign information estimation parameter calculating means (714 in FIG. 51) calculates the degree of difference as the second sign information estimation parameter (S02 in FIG. 52).
1), storing the calculation result in the second signature information estimation parameter storage means (715 in FIG. 51) (S02 in FIG. 52)
2). In the above example (FIGS. 61 and 62 for the coefficient A2),
The value of "positive difference" is "53768" and "
The value of the “negative difference” is “63334496”.

Next, the third similarity estimation parameter calculating means (716 in FIG. 51) calculates the second similarity to the third similarity.
It is calculated as a signature information estimation parameter (S0 in FIG. 52).
23), second signature information generating means (718 in FIG. 51)
In the above, when the third signature information estimation parameter corresponding to each temporary signature information is the second change amount, the temporary signature information when the value of the third signature information estimation parameter is small is estimated as true signature information. When the third signature information estimation parameter corresponding to the temporary signature information is the second similarity, the temporary signature information when the value of the third signature information estimation parameter is large is estimated to be true signature information, and the estimation result is If the answer is correct, “0 (1 [bi
t])) ”, and if the estimation result is incorrect,“ 1 (1 [bit]) ”is used as new signature information.
Is output. If estimation is impossible, the true sign information of the coefficient A1 itself is output (S025 in FIG. 52). In the above example (coefficient A2 in FIGS. 61 and 62), the value of the normalization coefficient α is “17” and “positive second similarity (= positive first similarity−α × positive dissimilarity) ) ”Is“ 6131
4752, and the value of “negative second similarity (= negative first similarity−α × negative dissimilarity)” is “−5625061”.
6 ”, the“ positive first similarity ”is larger, so the true sign information is estimated to be“ positive ”. Since this estimation result is a correct answer, "zero (1 [bit])" which means a correct answer is output as new signature information. The signature information output from the second signature information generation means (718 in FIG. 51) is stored in the signature information storage means (70 in FIG. 51).
2) (S026 in FIG. 52).

The above processing (from S014 to S02 in FIG. 52)
6) is repeated for all the coefficients in one DCT block, or until the last non-zero coefficient is processed and processed in the order of higher intensity rank (S027 in FIG. 52).
(In the example of the DCT block in FIG. 25, since no zero coefficient exists, the processing is repeated up to the coefficient A63.)

As described above, the DCT block (FIG. 25)
, The coordinates (n, m) of each coefficient when the sign information is estimated up to all the coefficients (coefficients A1 to A63)
The value of each coefficient, the positive second similarity, the negative second similarity, the estimation result, whether the estimation result is a correct answer (O) or an incorrect answer (X), a newly generated signature information (0 in the case of a correct answer, In case of incorrect answer 1,
Table 12, Table 13, and Table 14 show 0 in the case of positive sign and 1 in the case of negative sign information as it is when estimation is impossible.

[0287]

[Table 16]

[0288]

[Table 17]

[0289]

[Table 18]

The above processing (from S009 to S02 in FIG. 52)
Step 7) is repeated until completion for all DCT blocks (S028 in FIG. 52).

Finally, the signature information storage means (7 in FIG. 51)
02) is compression-encoded by the signature information encoding means (703 in FIG. 51) (S029 in FIG. 52). This encoding process (S0 in FIG. 52)
29), as in the above example, in each DCT block, the sign information is estimated by estimating the sign information of the coefficient having the lower strength rank in the order of the coefficient having the higher strength rank based on the information of the coefficient having the higher strength rank. Since “0” increases as the value of the information, the sign amount can be reduced by collecting the sign information, performing statistical processing, and encoding. As a method for compressing and encoding the sign information, any encoding method such as conventional Huffman encoding or arithmetic encoding may be used. Also, Table 15 shows the total number (total number of coefficients) as a totaling result of the estimation of the signature information of the DCT block (FIG. 25).
Correct answer (number and percentage of total number), Incorrect answer (number and percentage of total number), impossible to estimate (number and percentage of total number), signature information 0 (number using 0 as signature information) And the percentage of the total), signature information 1
(Number using 1 as signature information and percentage of total number), and information amount of signature information (number of bits and percentage of total number of bits (63 bits)). As a result, in the prior art, 1 [bit] was originally used for the signature information for one coefficient, so that 63 [bit] was required as the signature information for the coefficients A1 to A63. By estimating the sign information and treating it as new sign information as described above, the information amount becomes about 3
7.3 [bit], which means that the information amount can be reduced by about 25.7 [bit] (about 40.8%) from the original information amount. In the above-described seventh embodiment of the present invention, the original information amount is
Although the information amount can be reduced by about 13.1 [bit] (about 20.8%), in the eighth embodiment of the present invention,
This means that the amount of information can be reduced significantly more than in the seventh embodiment. As described above, with the above-described configuration, the sign information is estimated in consideration of both the first similarity (or the first change amount) and the dissimilarity information, so that the code amount necessary for the sign information is reduced. It can be significantly reduced.

[0292]

[Table 19]

Here, the signature information storage means (7 in FIG. 51)
02) in the compression encoding process (703 in FIG. 51) of the signature information stored in the signature information (703 in FIG. 51) (S029 in FIG. 52). A method will be described in which information (the rightmost values in Tables 12, 13, and 14) is classified into a plurality of groups, and statistical processing is performed for each group and encoded, thereby further reducing the amount of information. Looking at the values of the sign information for each coefficient in Tables 12, 13, and 14, the first half (the one closer to the coefficient A1) has some unpredictable and incorrect answers, but the second half (the coefficient A63). Closer)
The more you go, the higher the accuracy rate. In particular, the coefficient A
35 consecutive coefficients from 29 to coefficient A63 are:
In all cases, the estimation result of the signature information is correct. This is because in the first half, the number of coefficients for which the signature information is already known is small, but the further to the second half, the more signature information of the coefficient is known, and the subsequent signature information is estimated based on the information. Therefore, in the latter half, the signature information can be estimated using information close to the original image as a clue. Therefore, the latter half is more accurate. By utilizing this feature, the sign information of the 63 coefficients (the rightmost values in Tables 12, 13, and 14) are not combined and compression-coded.
For example, “first half group (31 coefficients from coefficient A1 to coefficient A31)” and “second half group (coefficient A32
, And a coefficient A63), the compression ratio is further improved by dividing into a plurality of groups and compressing and encoding each group. In the following, we calculate the “information amount” of the first half group and the second half group. Table 16 summarizes the situation of the first half group, and Table 17 summarizes the situation of the second half group. According to this result, the information amount of the first half group is about 26.9 [bit], the information amount of the second half group is 0.0 [bit], and the total information amount of the first half group and the second half group is about 26. [bit]. 9 [bit
] (About 42.8%). This is because the sum of the first half group and the second half group is originally 63% in the conventional technology.
[Bit], the information amount is about 42.8% (less than half), which means that the information amount of the sign information can be reduced by about 57.2%. As shown in Table 15 described above, the amount of information when performing compression encoding without discrimination between the former half group and the latter half group is about 37.3 [bit].
Therefore, when the compression coding is performed by dividing the data into the first half group and the second half group, the information amount becomes about 26.9.
[Bit], and the compression encoding is performed for each of the plurality of groups as described above, so that the information amount of the signature information can be significantly reduced.

[0294]

[Table 20]

[0295]

[Table 21]

<Ninth Embodiment> Next, a ninth embodiment of the present invention for solving the second problem of the conventional signal processing system will be described.

In the fifth embodiment of the present invention described above, the sign information is collectively encoded by the signature information processing means (7 in FIG. 22), thereby reducing the code amount of the signature information. 2 issues have been solved. In the sixth embodiment of the present invention, in each DCT block, estimation of the sign information of the coefficient A63 having the lowest strength rank is performed based on information of 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength rank. Then, the sign information of the estimated coefficient A63 is collected and collectively encoded, thereby reducing the code amount of the sign information and solving the second problem. Seventh of the present invention
In the embodiment of the present invention, in each DCT block, signature information of not only the coefficient A63 having the lowest strength rank but also all the coefficients A1 to A63 is estimated based on the information of the coefficient having the highest strength rank. By encoding,
The second problem has been solved by greatly reducing the code amount of the sign information. In the seventh embodiment of the present invention, the first similarity is used as the first signature information estimation parameter, and the signature information when the first similarity is large is estimated to be the correct answer. The eighth embodiment of the present invention uses the first similarity (or the first amount of change) as the first signature information estimation parameter, similarly to the seventh embodiment of the present invention.
By using the dissimilarity as the second signature information estimation parameter, the code amount of the signature information is significantly reduced as compared with the seventh embodiment of the present invention, and the second problem is solved. In the ninth embodiment of the present invention, as in the eighth embodiment of the present invention, the first signature information
The similarity (or the first variation) is used, and the difference is used as the second signature information estimation parameter, but the signature information is not estimated for each coefficient in descending order of the strength order. By collectively estimating the sign information of a plurality of coefficients, the code amount of the sign information is further reduced as compared with the eighth embodiment of the present invention, and the second problem is solved.

[0298] The image data as the input signal is obtained by converting a signal of a photograph or the like by a video camera or the like, or by converting a signal of a photograph or the like by a scanner. Things. Thus, the input signal (image data) is generated based on optical natural phenomena. Therefore, the contour (edge) of the subject
Information such as the shape of the input signal (image data)
Is reflected in the numerical value. These input signals (image data)
In the signal processing systems according to the related art and the present invention in which the frequency is resolved after the frequency decomposition, the energy is biased to one of the frequency components after the frequency decomposition depending on the shape of the contour (edge) of the subject. For example, if the contour (edge) of the subject has a complicated shape, energy concentrates on high frequency components, and if the contour (edge) has a monotonous shape, energy concentrates on low frequency components. I do. In other words, the frequency components where energy is concentrated reflect the characteristics of the input signal (image data) well and contain much information on the input signal (image data). In view of such characteristics of the input signal (image data), the ninth embodiment of the present invention provides the following FIG.
With the configuration as in No. 3, the code amount of the signature information is reduced. Note that, as in the fifth, sixth, seventh, and eighth embodiments of the present invention, coefficients other than the sign information are excluded from the conventional technology and the already described book except for the sign information as a mere absolute value. Encoding may be performed by the same processing as in the first, second, third, and fourth embodiments of the present invention. However, in the ninth embodiment of the present invention, as an example, the coefficient The description will be made assuming that the absolute value portion performs the same processing as that of the second embodiment of the present invention.

The configuration of the ninth embodiment of the present invention will be described below with reference to FIG. In FIG. 63, 9
Is frequency decomposition means, 10 is quantization means, 1
Is a basic encoding unit, 2 is a frequency component storage unit, 3 is a coefficient readout control unit, and 4 is a readout control information generation unit inside the basic encoding unit (1).
Then, inside the read control information generating means (4),
401 is a scan order determining means, 402 is a scan order table, 403 is a scan order information encoding means, and the scan order information E is used as read control information C, and the coefficient read control means (3) An address signal D is generated based on the read control information C, and 412 inside the scan order determining means (401)
411 is a priority order determining means.
The above functions basically in the same manner as the above-described second embodiment of the present invention. The difference from the second embodiment of the present invention is that an absolute value conversion means (6) for converting each coefficient of a frequency component into an absolute value is provided inside the basic coding means (1).
And a sine information processing means (7) which receives each coefficient of the frequency component as input and processes sine information of the input coefficient. For each coefficient of the frequency component converted into an absolute value,
The coefficient encoding means (5) performs statistical processing to encode the coefficients, and the sign information processing means (7) has:
Among the coefficients of each frequency component in the frequency component storage means (2), the intensity (hereinafter, the intensity) of each coefficient in the unit of frequency decomposition (DCT block) is measured, and the intensity of each coefficient is measured. Coefficient strength rank determining means (704) for determining a rank (hereinafter, coefficient strength rank); a coefficient strength rank table (705) for storing the coefficient strength rank; and determining whether each coefficient is zero. Only the non-zero determination sign information extracting means (701) for extracting the sign information, preferentially estimates the sign information from the coefficient having the highest coefficient strength rank, and generates and outputs new sign information from the estimation result. The signature information estimating means (706) performs a temporary sign information (hereinafter, referred to as a temporary sign information) for the coefficient of each frequency component in the frequency component storage means (2) inside the signature information estimating means (706). Sign And distribution), the true signature information (hereinafter optionally the sign information setting means (707) for setting a true signature information), signature information setting means (707)
The sign information-added coefficient storage means (708) for storing the temporary sign information and each coefficient (hereinafter, the coefficient with the sign information) for which the true sign information is set as necessary, and each frequency component of the coefficient with the sign information Frequency component synthesizing means (70) for synthesizing
9), a frequency synthesis result storage means (710) for storing a frequency synthesis result with signature information, and a first signature information estimation parameter corresponding to the signature synthesis result with signature information in which each temporary signature information is set (hereinafter referred to as a first parameter). , A first signature information estimation parameter), a first signature information estimation parameter calculation means (711), a first signature information estimation parameter storage means (712) for storing the first signature information estimation parameter, A second signature information estimation parameter calculation means (714) for calculating a second signature information estimation parameter (hereinafter, a second signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which is set; Second signature information estimation parameter storage means (7) for storing information estimation parameters
15) and using the first signature information estimation parameter and the second signature information estimation parameter, a third signature information estimation parameter (hereinafter, referred to as a third signature information parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. A third signature information estimation parameter calculating means (716) for calculating the third signature information estimation parameter; a third signature information estimation parameter storage means (717) for storing the third signature information estimation parameter; From the value of the obtained third signature information estimation parameter, the combination of the temporary signature information that matches the combination of the true signature information is
A third signature information generating unit (719) for adding new signature information and outputting new signature information; and in the signature information setting unit (707), estimating whether the temporary signature information is true signature information. Has a function of setting true signature information for the coefficient for which the signature has been completed, and a frequency synthesis result (hereinafter, a true sine frequency synthesis result) based on a coefficient for which true signature information is set (hereinafter, a coefficient with true signature information).
And the sum of the absolute value of the difference between the result and the frequency synthesis result (hereinafter referred to as the provisional sign frequency combination result) by the coefficient in which the provisional sign information is set (hereinafter referred to as the provisional sign frequency combination coefficient) Regarding the frequency synthesis result in the mixed state of the information-added coefficient and the temporary signature information-added coefficient (hereinafter, the mixed frequency synthesis result), the waveform of the mixed frequency synthesis result in the center axis of the waveform and the region surrounded by the waveform The sum of the absolute value of the difference between the waveform and the center axis of the waveform is defined as the first similarity, and the sum of the absolute values of the difference between the waveform and the value range in the region where the waveform of the mixed frequency synthesis exceeds the predetermined value range. As the degree of difference, and the first sign information estimation parameter calculating means (71
In 1), the first change amount or the first similarity is set to the first
The second signature information estimation parameter calculation means (714) calculates the degree of difference as a second signature information estimation parameter.
The change amount or dissimilarity is normalized by a normalization coefficient, and a value obtained by adding the normalized dissimilarity to the normalized first change amount is set as a second change amount, and is normalized to the first similarity or dissimilarity. A value obtained by subtracting the normalized dissimilarity from the normalized first similarity is used as a second similarity, and the second sign information estimation parameter calculating means (716) calculates the second change The amount or the second similarity is calculated as a third signature information estimation parameter, and the third signature information generation means (719) calculates the third signature information corresponding to each temporary signature information.
If the signature information estimation parameter is the second variation, the third
Either a combination of temporary signature information with a small value of the signature information estimation parameter is preferentially used as a code containing a large number of identical symbols (hereinafter, a first code), or temporary signature information with a small value of the third signature information estimation parameter. Is given a short code (hereinafter referred to as a second code),
The first code or the second code corresponding to the combination of the temporary signature information that matches the combination of the true signature information is output as new signature information, and the third signature information estimation parameter corresponding to each temporary signature information is set to the second signature information. In the case of two similarities, the combination of temporary signature information having a large value of the third signature information estimation parameter is preferentially used as the first code, or the combination of temporary signature information having a large value of the third signature information estimation parameter is used. The combination is preferentially used as the second code, and the first code or the second code corresponding to the combination of the temporary signature information that matches the combination of the true signature information is output as new signature information. is there.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the coefficient coding means (5) performs Huffman coding.

The operation of the ninth embodiment of the signal processing system (compression encoding side) will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) is input (S0 in FIG. 64).
01).

<DCT and Quantization> Next, the frequency decomposition means (9 in FIG. 63) performs a DCT operation as shown in Equation 1 on the input original image block (S002 in FIG. 64), and In the conversion means (10 in FIG. 63), D
Each coefficient (DCT block) after CT is quantized (FIG. 6).
4 S003). Then, the quantized DCT block is stored in the frequency component storage means (2 in FIG. 63) (FIG. 64).
S004). Here, as an example, all coefficients are
Uniform quantization is performed by applying 1 as a quantization step.
This DCT and quantization processing is the same as the above-described conventional signal processing system.

<Calculation of Strength> Next, the coefficient strength calculation means (412 in FIG. 63) calculates the strength of the coefficient corresponding to each component of the quantized DCT block (S005 in FIG. 64).

The above series of processes (S001 and S001 in FIG. 64)
002, S003, S004, and S005) are completed for all original image blocks (S0 in FIG. 64).
06), until the processing of all the original image blocks is completed.

<Creation of Scan Order Table> Next, the scan order is determined by the priority order determining means (411 in FIG. 63) such that sampling is performed from the coefficient having the largest intensity value (sum of absolute values of coefficients). It is stored in the scan order table (402 in FIG. 63) (S007 in FIG. 64). Here, scan order information (eg, coordinate values) stored in the scan order table (402 in FIG. 63) is encoded by a scan order information encoding unit (403 in FIG. 63) and output as an output signal. (S008 in FIG. 64).
The encoding of the scan order information performed here simply involves the coordinate value of 2
A value expressed in a hexadecimal number may be output, or in image data or audio data, since a non-zero coefficient is unlikely to appear in a low-frequency component, a coordinate value closer to the low-frequency component has a shorter sign. A method such as outputting as a long code may be used.

<Execution and Encoding of Scan> Next, in accordance with the scan order stored in the created scan order table (402 in FIG. 63), read control means (FIG. 6)
In 3-3), an address signal D of the frequency component storage means (2 in FIG. 63) is generated, and each coefficient of the quantized DCT block is converted from the frequency component storage means (2 in FIG. 63) to the last non-zero value. The coefficients are read out (S009 in FIG. 64). The read coefficients are converted into absolute values by the absolute value conversion means (6 in FIG. 63) (S010 in FIG. 64), and each coefficient converted into the absolute value is converted into coefficient encoding means (5 in FIG. 63). (S011 in FIG. 64). That is, this encoding process (S011 in FIG. 64) is performed only on the absolute value part of the coefficient excluding the sign information. The processing for obtaining the absolute value of the coefficient may be performed before storing the coefficient in the frequency component storage means (2 in FIG. 63). In this case, each coefficient is stored in the frequency component storage means (2 in FIG. 63). , And may be stored separately for the absolute value and the sign information.

<Processing of Signature Information> Next, by the reading process in the scan order (S009 in FIG. 64), the coefficient to be read is determined by the coefficient strength ranking determining means (7 in FIG. 63).
04), the energy intensity (hereinafter, intensity) of each coefficient is measured in one DCT block, and the order of the intensity of each coefficient (hereinafter, coefficient intensity order) is determined (S01 in FIG. 64).
2). Then, the value of the coefficient strength rank is stored in the coefficient strength rank table (705 in FIG. 63) (S01 in FIG. 64).
3). As described above, if the contour (edge) of the subject has a complicated shape, if the energy is concentrated on the high-frequency component or if the contour (edge) has a monotonous shape,
Energy concentrates on low frequency components. The frequency component where energy is concentrated reflects the characteristics of the input signal (image data) well, and the input signal (image data)
The above processing (FIG. 64)
S012) is for determining to which frequency component in the DCT block energy is concentrated.
The intensity in this processing (S012 in FIG. 64) is specifically the absolute value of each coefficient or the magnitude of the square value of each coefficient. As an example, in the present embodiment, the intensity is an absolute value of the coefficient. The determination of the intensity order is performed only for the AC component in the DCT block of interest. For example, in the case of the same DCT block (FIG. 25) used in the description of the fifth embodiment of the present invention, a coefficient having the largest intensity value among AC components (hereinafter, referred to as a coefficient A1). Is “−784” at the coordinates (1, 0). 2
Coefficient with the second largest intensity value (hereinafter referred to as coefficient A2)
Is “212” at the coordinates (3,0). The coefficient having the third largest intensity value (hereinafter, referred to as coefficient A3) is “−103” at the coordinates (5, 0). The coefficient having the fourth largest intensity value (hereinafter, referred to as coefficient A4) is “71” at coordinates (2, 4) and “−” at coordinates (4, 1).
71 "and two coefficients having the same magnitude. In such a case, either of them may be prioritized, but in this example, “71” of the coordinates (2, 4) is the fourth (coefficient A4). In this way, the ranking is performed in the order of the strengths, the coefficient strength ranking is determined, and information of the coordinates (n, m) corresponding to the ranking is stored in the coefficient strength ranking table (705 in FIG. 63). Keep it.

The above processing (S001 to S01 in FIG. 64)
3) is the same as the above-described eighth embodiment of the present invention.

Next, whether or not each coefficient is non-zero is determined by the non-zero determination sign information extracting means (701 in FIG. 63) (S014 in FIG. 64). Based on (706 in FIG. 63), it is estimated whether the sign information of each coefficient is positive or negative (FIG. 64).
S015 to S026). In the sign information estimation process (S015 to S026 in FIG. 64), information of a coefficient having a high strength order (a large value of the strength) is used as a clue in accordance with the coefficient strength order obtained earlier, and the strength order is low (strength). Is estimated). The sign information of the coefficient is estimated. However, for the example of the DCT block (FIG. 25),
In the above-described sixth embodiment of the present invention, the coefficient A63 having the lowest strength order (FIG. 25) is obtained based on information of 62 coefficients from the coefficient A1 to the coefficient A62 having the highest strength order.
The sign information of only the coordinate (4, 0) of “−2”) was estimated, but the ninth embodiment of the present invention is similar to the seventh and eighth embodiments of the present invention. The sign information of all the coefficients up to the coefficient A63 is estimated in order from the coefficient A1 having the highest intensity rank. In addition, the seventh and eighth aspects of the present invention
In the embodiment of the present invention, in order from the coefficient A1 having the highest intensity ranking,
The sign information of each coefficient is estimated up to the coefficient A63, but the present embodiment (the ninth embodiment of the present invention).
In this method, the sign information of a plurality of coefficients is collectively estimated preferentially from the coefficient having the highest strength order, thereby further reducing the information amount of the sign information. In the following, as an example, the sign information of four coefficients is collectively estimated in descending order of the strength rank. However, there are 63 coefficients in total,
Since it is not divisible by 4, the last coefficient A61 and coefficient A61
62, only the coefficient A63 is assumed to collectively estimate the sign information.

First, the sign information is estimated for four coefficients A1 to A4 in the descending order of the strength order. The sign information setting means (707 in FIG. 63) sets true sign information in the DC component, and sets temporary sign information in four coefficients A1 to A4 (S015 in FIG. 64). There are 16 (= 2 to the fourth power) combinations of provisional signature information set for the four coefficients from the coefficient A1 to the coefficient A4. These 16 combinations are managed by a signature number N. For example, when positive sign information is set as temporary sign information for all four coefficients from the coefficient A1 to the coefficient A4, this combination is managed as “sign number N = 0000”. This indicates whether the temporary sign information set for each coefficient is positive (= 0) or negative (= 1), as shown in FIG. A DCT block in which temporary signature information is set when the signature number N = 0000 is referred to as a DCT block 0000. The DCT block 0000 is shown in FIG.

The sign information setting means (707 in FIG. 63) sets true sign information in the DC component, and assigns four coefficients A1 to A4 to provisional signs corresponding to all 16 sign numbers N. (S01 in FIG. 64)
5) Coefficient storage means with signature information (708 in FIG. 63)
(S016 in FIG. 64). That is, the 16 DCT blocks (DCT block 0000, DCT block 0001,..., DCT block) corresponding to the sign number N are stored in the coefficient storage unit with sign information (708 in FIG. 63).
Block 1111) is stored.

Next, a coefficient storage unit with signature information (FIG. 6)
3 DCT blocks (DCT blocks 0000, DT) corresponding to the signature number N stored in
CT block 0001,..., DCT block 111
For 1), the inverse DCT operation of Expression 2 is performed by frequency synthesis means (709 in FIG. 63) (S017 in FIG. 64),
Each inverse DCT result is stored in the frequency synthesis result storage means (710 in FIG. 63) (S018 in FIG. 64). In addition, before performing the above-described inverse DCT operation (S017 in FIG. 64), it is preferable to perform the inverse quantization on each DCT block. 1, and the process of inverse quantization is omitted here. (Inverse DCT may be provided to perform inverse quantization.) Note that inverse DCT is a real number calculation, and there are values after the decimal point. Usually, the result of the real number calculation is treated as an integer. . At this time, the value after the decimal point is truncated, causing an error with respect to the actual operation result. In order to minimize this error, the above-described inverse DCT operation (S017 in FIG. 64) is performed, for example, by multiplying the inverse DCT result by 10,000 times to obtain a higher bit.
It shall be output as an integer value of precision. A result obtained by multiplying the inverse DCT result of the DCT block 0000 by 10000 is called an inverse DCT block 0000. Similarly, a result obtained by multiplying the inverse DCT result of each DCT block (DCT block 0001,..., DCT block 1111) by 10000 is used. Using the signed number (inverse DCT block 000
1,..., Inverse DCT block 1111). A result obtained by multiplying the inverse DCT result of DCT block 0000 (FIG. 66) by 10,000 (hereinafter, inverse DCT block 0)
000) is shown in FIG.

Next, 16 inverse DCT blocks (inverse DCT block 0000, inverse DCT block 000) are calculated by the first sign information estimation parameter calculating means (711 in FIG. 63).
,..., Inverse DCT block 1111), a first similarity calculation is performed as a first signature information estimation parameter (S019 in FIG. 64), and the calculation result is stored in the first DCT block 1111).
Sign information estimation parameter storage means (712 in FIG. 63)
(S020 in FIG. 64). The calculation of the first similarity is exactly the same processing as that described in the eighth embodiment of the present invention. Although the first variation may be used as the first signature information estimation parameter, the first similarity is used in this embodiment.

Then, the second sign information estimation parameter calculating means (714 in FIG. 63) calculates 16 inverse DCT blocks (inverse DCT block 0000, inverse DCT block 000).
,..., Inverse DCT block 1111), the difference is calculated as the second signature information estimation parameter (S021 in FIG. 64), and the calculation result is stored in the second signature information estimation parameter storage unit (S021 in FIG. 64). (715 in FIG. 63) (S022 in FIG. 64). This calculation of the degree of difference is exactly the same processing as that described in the eighth embodiment of the present invention.

Next, in the third sign information estimation parameter calculation means (716 in FIG. 63), each of the 16 inverse DCT blocks (inverse DCT block 0000, inverse DCT block 0001,..., Inverse DCT block 1111) On the other hand, the second signature information
The amount of change or the second similarity is calculated (S02 in FIG. 64).
3). In the present embodiment (a ninth embodiment of the present invention), a second similarity is calculated as a third signature information estimation parameter. Specifically, in order to normalize the first similarity and the dissimilarity, the first similarity of each of the 16 inverse DCT blocks (inverse DCT block 0000, inverse DCT block 0001,..., Inverse DCT block 1111) is determined. Sum (hereinafter referred to as sum of first similarity) and 16 inverse DCs
The sum of the dissimilarities of the T blocks (the inverse DCT block 0000, the inverse DCT block 0001,..., The inverse DCT block 1111) (hereinafter referred to as the sum of the dissimilarities) is determined, and the first similarity is calculated as the normalization coefficient α. The ratio of the sum of degrees and the sum of differences (= sum of first similarities / sum of differences) is determined. Then, 16 inverse DCT blocks (inverse DCT block 00
00, inverse DCT block 0001,..., Inverse DCT block 1111), as the second similarity corresponding to each, a value obtained by subtracting the value of the degree of difference multiplied by α from the first similarity (= first similarity− α × difference). That is, 16 values of the second similarity corresponding to the signature number N are obtained. Table 1 shows the second similarity for each of the sign numbers N for the four coefficients (coefficient A1, coefficient A2, coefficient A3, and coefficient A4).
Summarize in 8. The value on the right end of Table 18 is “size order number (0 when the second similarity is the maximum, 15 when the second similarity is the minimum)” which means the order of the magnitude of the second similarity. However, when the second similarities are of the same size, priority is given to the “size order number” with the smaller signature number N.

[0317]

[Table 22]

In the above example, the second similarity is calculated as the third signature information estimation parameter, but the second change amount may be calculated instead. If the second change amount is calculated as the third sign information estimation parameter, 16 inverse DCT blocks (inverse DCT block 0000, inverse DCT block)
T block 0001,..., Inverse DCT block 111
1) The sum of the first change amounts (hereinafter, referred to as the first change amount sum) and 16 inverse DCT blocks (inverse DCT block 0000, inverse DCT block 0001,.
The inverse DCT block 1111) obtains the sum of the differences, and calculates, as the normalization coefficient α, the ratio of the sum of the first variation and the sum of the differences (= sum of the first variation / sum of the differences). Then, 16 inverse DCT blocks (inverse DCT block 00
00, the inverse DCT block 0001,..., The inverse DCT block 1111), as the second variation corresponding to each, a value obtained by adding the value of the difference degree multiplied by α to the first variation (= first variation + α X difference). In addition,
In the above, when calculating the normalization coefficient α, the sum of the dissimilarities may be used as it is, but the sign information of the inverse DCT block may not be correct due to the quantization error or not using all the coefficients. However, it may be slightly out of range. Therefore, when obtaining the normalization coefficient α, the value of the degree of difference is adjusted so as to be slightly smaller (subtraction or
Alternatively, the normalization coefficient α itself may be adjusted (subtracted or multiplied by a coefficient less than 1) to a slightly smaller value.

Next, in the third signature information generating means (719 in FIG. 63), when the third signature information estimation parameter corresponding to each temporary signature information is the second variation, the value of the third signature information estimation parameter is obtained. Is preferentially used as a code including a large number of identical symbols (hereinafter, referred to as a first code), or a combination of temporary signature information having a small value of the third signature information estimation parameter is preferentially used. , A short code (hereinafter, a second code)
The first code or the second code corresponding to the combination of the true signature information and the matching temporary signature information is output as new signature information, and the third signature information estimation parameter corresponding to each temporary signature information is In the case of the second similarity, a combination of temporary signature information having a large value of the third signature information estimation parameter is preferentially used as the first code, or temporary signature information having a large value of the third signature information estimation parameter. Is given priority as the second code, and the first code or the second code corresponding to the combination of the true signature information and the matching combination of the temporary signature information is output as new signature information (FIG. 64). S
025).

In the present embodiment (the ninth embodiment of the present invention), the third signature information estimation parameter is set to the second similarity, and the processing in the third signature information generating means (719 in FIG. 63) is performed. (S025 in FIG. 64) specifically performs processing as shown in the following example. Based on the same concept as in the eighth embodiment of the present invention described above, in Table 18, the sign number N (combination of temporary sign information) may be more correct as the second similarity is larger. It is considered high. However, actually, the case where the second similarity is the largest is not always the correct answer. In the example of Table 18, “size number” is “
In the case of "3", the sign number N (combination of temporary sign information) is correct. If this “3 (size order number)” is represented by a binary number, it becomes “0011”. By devising a method of expressing the “size order number”, it is expressed by a code (first code) containing many identical symbols,
The code is divided into four coefficients (coefficient A1, coefficient A2, coefficient A
3, the coefficient A4) is used as new sign information. That is, based on the same concept as in the eighth embodiment of the present invention described above, the sign number N (combination of temporary sign information) is more likely to be correct as the second similarity increases. That is, the smaller the numerical value of the numerical order of the size, the higher the possibility of the correct answer. Therefore, the first code is configured such that the smaller the numerical value of the size order number, the larger the number of zeros (or any one, but either one) appears more frequently, and assigns the first code corresponding to the size order number to a new signature. Treat it as information. Hereinafter, an example of the first code is shown. The “size order number” is “0”, and the first code is “0000”.
(All zero). After that, "size order number"
1 "in the first code," 1000 "(1 is 1)
And The “size order number” is “2”, and the first code is “0100” (1 is 1). The “size order number” is “3”, and the first code is “001”.
0 "(1 is 1). These size order numbers and the first
Table 19 summarizes the correspondence between the codes.

[0321]

[Table 23]

As shown in Table 18, four coefficients (coefficient A
1, the coefficient A2, the coefficient A3, and the coefficient A4), the correct combination of the sign information is that the sign number N is “101”.
1 ". At this time, the size order number is "
3 ", which can be expressed in binary notation as" 00
11 ". Simply treating this “0011” as sign information of four coefficients (coefficient A1, coefficient A2, coefficient A3, and coefficient A4), 1 becomes 2 [bit] (2
) Are present, and the compression ratio is deteriorated during the subsequent compression encoding. Therefore, the first code (“0010” in Table 19) in which the smaller the size order number, the more zeros appear.
"). In this case, 1 is 1 [bit]
(One), and the number of ones can be reduced as compared with the case where the size order number is simply expressed in a binary number. If the sign information can be expressed in a state where the number of 1s is small and the number of zeros is as large as possible, it is possible to improve the compression ratio at the time of compression encoding later. (Note that, instead of using the first code, a code having a smaller size order may be replaced with a shorter code (second code).) Four coefficients (coefficient A1, coefficient A2, coefficient A3, Table 20 summarizes the result of newly creating the signature information by replacing the size order number of the coefficient A4) with the first code. Table 20
, Four coefficients (coefficient A1, coefficient A2, coefficient A
3, coefficient A4) (newly generated) signature information "0
010 (4 bits) "are described in order from the upper bit. As described above, the third signature information generating means (7 in FIG. 63)
The signature information newly generated in 19) is stored in the signature information storage means (702 in FIG. 63) (S02 in FIG. 64).
6).

[0323]

[Table 24]

As described above, the four coefficients (coefficient A1, coefficient A2, coefficient A
3, the process of estimating the signature information corresponding to the coefficient A4) is completed.

Next, the non-zero determination sign information extracting means (701 in FIG. 63) determines whether or not the four coefficients A5 to A8 are non-zero. Is estimated.

First, the signature information setting means (70 in FIG. 63)
In 7), the true sign information is set to the DC component (it has already been set and can be used as it is), and the true sign information is added to the four coefficients A1 to A4 for which the sign information has already been estimated. Is set, and temporary sign information corresponding to all 16 sign numbers N is set to four coefficients A5 to A8 from which sign information is estimated (S015 in FIG. 64). It is stored in the coefficient storage means (708 in FIG. 63) (S01 in FIG. 64).
6). That is, the coefficient storage unit with signature information (FIG. 63)
708) includes 16 DCs corresponding to the signature number N.
T blocks (DCT block 0000, DCT block 0001,..., DCT block 1111) are stored.

Thereafter, the processing of S017 to S026 in FIG. 64 is performed in the same manner as in the case of the coefficients A1 to A4.

The above processing (from S014 to S02 in FIG. 64)
Until step 6) is completed for all coefficients in one DCT block, or until the last non-zero coefficient is processed by processing in descending order of intensity (S027 in FIG. 64).
(In the example of the DCT block in FIG. 25, since no zero coefficient exists, the processing is repeated up to the coefficient A63.)

As described above, the DCT block (FIG. 25)
Table 21, Table 22, Table 23, and Table 24 show the results of creating signature information for all the coefficients (coefficients A1 to A63). In Table 21, Table 22, Table 23, and Table 24, the coordinates of each coefficient (coefficient A1 to coefficient A63), the coefficient value, and a set of four coefficients (eg, coefficient A1, coefficient A2, Coefficient A3, coefficient A4, except for the lowest coefficient A6
1, the coefficient A62 and the coefficient A63 are three sets), the sign number of the correct answer, the second similarity corresponding to the sign number of the correct answer, the size order number (in decimal notation), the sign information (the first code ).

[0330]

[Table 25]

[0331]

[Table 26]

[0332]

[Table 27]

[0333]

[Table 28]

The above processing (from S009 to S02 in FIG. 64)
Step 7) is repeated until completion for all DCT blocks (S028 in FIG. 64).

Finally, signature information storage means (7 in FIG. 63)
02) is compressed and encoded by the signature information encoding means (703 in FIG. 63) (S029 in FIG. 64). This encoding process (S0 in FIG. 64)
29), as in the above example, in each DCT block, the sign information is estimated by estimating the sign information of the coefficient having the lower strength rank in the order of the coefficient having the higher strength rank based on the information of the coefficient having the higher strength rank. Since “0” increases as the value of the information, the sign amount can be reduced by collecting the sign information, performing statistical processing, and encoding. As a method for compressing and encoding the sign information, any encoding method such as conventional Huffman encoding or arithmetic encoding may be used. Table 25 also shows the total number (total number of coefficients),
Sign information 0 (the number using 0 as the sign information and the ratio in the total number), sign information 1 (the number using 1 as the sign information and the ratio in the total number), the information amount of the sign information (the number of bits And the number of bits (the ratio of the total number of bits to 63 bits). As a result, in the prior art, the signature information originally includes 1 [bit] for each coefficient.
], 63 [bit] was required as the sign information of the coefficients A1 to A63. However, as described above, the sign information is estimated and treated as new sign information. The amount of information is about 34.6 [bi
t], which is about 28.4 with respect to the original information amount.
[Bit] (about 45.1%) can be reduced. In the above-described eighth embodiment of the present invention, about 25.7 [
bit] (approximately 40.8%), but in the present embodiment (the ninth embodiment of the present invention), the information amount is much greater than in the eighth embodiment. That is,

[0336]

[Table 29]

As described above, with the above-described configuration, the sign information of a plurality of coefficients is collectively estimated by considering both the first similarity (or the first variation) and the difference. Thus, the code amount required for the sign information can be significantly reduced. In the above description, an example is shown in which the coefficients of the AC components are grouped "by four" and the signature information is estimated.
It goes without saying that the number of coefficients to be collectively estimated need not be limited to four.

Here, the signature information storage means (7 in FIG. 63)
02) in the process (S029 in FIG. 64) of compressing and encoding the sign information stored in the sign information (703 in FIG. 63). Is classified into a plurality of groups, statistical processing is performed for each group, and the group is coded to further reduce the amount of information. Looking at the values of the sign information of each coefficient in Table 21, Table 22, Table 23 and Table 24, the first half (Table 21, Table 22, Table 2
3 and the value closer to the coefficient A1 in Table 24), there are some values of “1” as sign information (first code), but in the latter half (coefficients in Tables 21, 22, 23 and 24). The closer you go to A63, the more sign information (No. 1)
Code), the value of "zero" appears more frequently. In particular, from the coefficient A28 to the coefficient A63,
The sign information (first code) of all 36 consecutive coefficients has a value of “zero”. This is because in the first half, there are few coefficients whose signature information is already known,
The more you go, the more sign information of the coefficient is known, and since the information is used as a clue to estimate the sign information after that, the latter half uses the information closer to the original image as a clue. Can be estimated. Therefore, in the latter half, the accuracy increases, and the probability that the second similarity having a small size order number (close to zero) is correct is higher. By utilizing this feature, the sign information of 63 coefficients (Table 21, Table 22, Table 2)
3 and the values at the right end of Table 24) are not combined and compression-coded. For example, among the 63 coefficients, for example, “first half group (31 coefficients from coefficient A1 to coefficient A31)” and “second half group” (From coefficient A32 to coefficient A63,
32), the compression rate is further improved by dividing into a plurality of groups and performing compression encoding on each group. In the following, we calculate the “information amount” of the first half group and the second half group. Table 26 summarizes the situation of the first half group, and Table 27 summarizes the situation of the second half group. From this result, the information amount of the first half group is about 25.5 [
bit], the information amount of the latter half group is 0.0 [b
It], the total information amount of the first half group and the second half group is about 25.5 [bit] (about 40.5%). This means that the sum of the first half group and the second half group is about 40.5% (less than half), compared to 63 bits in the prior art, and about 5
This means that the amount of signature information can be reduced by 9.5%. As described in Table 25 above, the amount of information when performing compression encoding without distinction between the former group and the latter group is:
Since it was about 34.6 [bits], when compression encoding is performed by dividing the data into the first half group and the second half group, the information amount becomes about 25.5 [bits]. By performing compression encoding for each group, the amount of signature information can be significantly reduced. Similarly, in the above-described eighth embodiment of the present invention, the case where compression encoding is performed by dividing the data into the first half group and the second half group is similarly performed at about 26.9 [b].
it], but the present embodiment (the ninth embodiment of the present invention) is about 1.4 [bit] in comparison with this.
] Can be reduced, compared to the fifth, sixth, seventh and eighth embodiments of the present invention described above.
With good results.

[0339]

[Table 30]

[0340]

[Table 31]

<Tenth Embodiment> Next, in order to solve the third problem of the conventional signal processing system,
A tenth embodiment of the present invention will be described.

A third problem of the above conventional signal processing system is that each coefficient (one-dimensional data in FIG. 83) of the one-dimensionally arranged DCT block is converted into coefficient coding means (FIG. 81).
51), the Huffman table used in the process of Huffman coding (S005 in FIG. 82) generally has a shorter code (amp) as the amplitude absolute value (amp) of each coefficient of the AC component of the DCT block is smaller. Code), the amplitude absolute value (am
The coefficient with the larger p) is the coefficient encoding means (51 in FIG. 81).
When the frequency of inputting to the is high, the code length (code length) becomes longer as a whole. In the tenth embodiment of the present invention, before performing frequency decomposition on each input signal (in this example, image data), as preprocessing, a difference value between adjacent samples (in this example, pixel values) is calculated. Then, a frequency decomposition is performed on this difference value,
The third problem of the conventional signal processing system is solved by performing compression encoding.

FIG. 68 shows the configuration of the signal processing system (compression encoding side) according to the tenth embodiment of the present invention. Fig. 68
, 8 is an adjacent sample difference calculating means for calculating a difference between adjacent sample values (pixel values) of image data as an input signal (hereinafter, an adjacent sample difference value).
Is frequency decomposition means for decomposing an adjacent sample difference value into a plurality of frequency components and outputting a coefficient A of each frequency component.
10 is data after frequency decomposition (hereinafter referred to as coefficient A)
Is a basic encoding means that receives data after quantization of the coefficient A (hereinafter, referred to as coefficient B) and outputs compressed encoded data. In 2), 2 is a frequency component storage unit for storing data after quantization of the coefficient A (hereinafter, referred to as a coefficient B), and 3 is an address signal D supplied to the frequency component storage unit (2). , A coefficient readout control means for reading out the stored coefficient B. 450 is a frequency component storage means (2).
Is a fixed scan order table in which the order in which each coefficient B is read from the memory (hereinafter, referred to as a scan order) is stored as a fixed value in advance. Reference numeral 51 denotes a compression code for each coefficient B read from the frequency component storage means (2). This is a coefficient encoding means to be transformed.

In the following description, as an example, it is assumed that the frequency resolving means (9) performs DCT, the fixed scan order table (450) stores a scan order called zigzag scan, and performs coefficient coding. Means (5
In 1), Huffman coding is performed. The range of the input signal (image data) is, for example, “0 to 255”.
".

The operation of the signal processing system (compression encoding side) according to the tenth embodiment of the present invention will be described below with reference to FIG.

<Preprocessing> First, in the adjacent sample difference calculating means (8 in FIG. 68), the difference (hereinafter, referred to as the pixel value) between all adjacent sample values (pixel values) for one frame of image data as an input signal. The adjacent sample difference value is calculated, and the adjacent sample difference value is stored in the adjacent sample difference value storage means (37 in FIG. 68) (S001 in FIG. 69).
A characteristic feature of a natural image is that pixel values are very similar between neighboring pixels (particularly between neighboring pixels). Therefore, when a difference value (adjacent sample difference value) is obtained between neighboring pixels, the absolute value of the difference value (adjacent sample difference value) becomes smaller. Difference value (adjacent sample difference value)
Is small means that the amplitude absolute value of each coefficient after frequency decomposition is small. This is because in orthogonal transform such as DCT, the sum of the squares of the pixel values of the block before the DCT is equal to the sum of the squares of the coefficients of the block after the DCT. Therefore, when the difference values (adjacent sample difference values) are obtained, and these adjacent sample difference values are subjected to frequency decomposition (DCT in the following example) and encoded, the AT of the DCT block is obtained.
Since the amplitude absolute value (amp) of each coefficient of the C component becomes small, the code amount can be finally reduced. Processing for calculating the adjacent sample difference value (S001 in FIG. 69)
Is specifically performed as in the example of FIG. For example, assume that the pixel value (original image block) of one block of the input signal (original image data) is as shown in FIG. FIG. 72 shows the result of calculating adjacent sample difference values for the original image block (FIG. 71) in the order shown in FIG. In the example of FIG. 70, in the horizontal direction, a value obtained by subtracting the value of the pixel at the coordinate x1 (FIG. 71) from the pixel at the coordinate x0 (FIG. 71) is stored at the position of the coordinate x1 in FIG. Next, from the pixel at the coordinate x1 (FIG. 71) to the pixel at the coordinate x2 (FIG.
The value obtained by subtracting the value of 1) is stored at the position of the coordinate x2 in FIG. Such processing is repeated up to the right end. (In this example, only one block is shown.
What is necessary is just to go to the right end of the screen of one frame. Do this for each row. As shown in FIG. 70, only the leftmost value is
The pixel at the coordinate y0 (FIG. 71) to the pixel at the coordinate y1 (FIG. 7)
The value obtained by subtracting the value of 1) is stored at the position of the coordinate y1 in FIG. Next, a value obtained by subtracting the value of the pixel at the coordinate y2 from the pixel at the coordinate y1 is stored at the position of the coordinate y2 in FIG. Such a process is repeated to the bottom. (In this example,
Although only one block is shown, this processing may be performed up to the bottom of the screen of one frame. In the above processing, focusing on only one block, the adjacent sample difference value shown in FIG. 72 is stored in the adjacent sample difference value storage means (37 in FIG. 68).

<Input of Block> One block (8 pixels × 8 pixels) is stored in the adjacent sample difference value storage means (37 in FIG. 68).
An adjacent sample difference value (for example, the value in FIG. 72) of the eight pixels is read out and input to the frequency decomposition means (9 in FIG. 68) (S002 in FIG. 69).

<DCT and Quantization> Next, in the frequency decomposition means (9 in FIG. 68), the input adjacent sample difference value of one block (8 pixels × 8 pixels) (FIG. 72) 1 is performed (S0 in FIG. 69).
03), each coefficient (DCT block) after DCT is quantized by the quantization means (10 in FIG. 68), and the quantized DCT block is stored in the frequency component storage means (2 in FIG. 68) (FIG. 69). S004). Here, as an example, any coefficient is quantized by applying 1 as a uniform quantization step. This DCT and quantization processing is the same as the above-described conventional signal processing system. FIG. 73 shows a result of performing DCT on the adjacent sample difference value (FIG. 72) of one block (8 pixels × 8 pixels). DC for this adjacent sample difference value (FIG. 72)
In the T result (FIG. 73), the sum of the squares of the AC components is
In “247452”, the sum of the absolute values of the AC components is
"2612". For reference, FIG. 74 shows the result of directly performing DCT on the original image block (FIG. 71) without taking the adjacent sample difference value. (A method of performing DCT directly on an original image block in this way is as follows.
It is used in MPEG intra pictures. )
In the DCT result (FIG. 74) for the original image block (FIG. 71), the sum of the squares of the AC components is “7353”.
04 ", the sum of the absolute values of the AC components is" 2720
". For reference, FIG. 75 shows the result of uniformly subtracting 128 from all pixels of the original image block (FIG. 71) without taking the adjacent sample difference value. FIG. 76 shows the result of performing.
(In this way, from all pixels of the original image block, "12
A method of performing DCT on the value obtained by subtracting 8 is as follows.
Used in JPEG. 7) A value of “128 subtracted” from each pixel of the original image block (FIG. 71) (FIG. 7)
In the DCT result with respect to 5) (FIG. 76), the sum of the squares of the AC component is “735381”, and the sum of the absolute values of the AC components is “2721”. As can be seen from the above results, as in the present embodiment (the tenth embodiment of the present invention), the result of performing DCT on the adjacent sample difference value (FIG. 73) is similar to that of MPEG. The sum of the square of the AC component and the sum of the absolute values of the AC component are smaller than either the result of the processing (FIG. 74) or the result of the processing such as JPEG (FIG. 76). And
It can be seen that the code amount can be reduced in the later Huffman coding.

<Zigzag Scan> Next, in accordance with the scan order of the zigzag scan stored in the fixed scan order table (450 in FIG. 68), the quantization stored in the frequency component storage means (2 in FIG. 68). The subsequent DCT block is read and arranged one-dimensionally (S00 in FIG. 68).
5).

<Huffman Coding> Next, Huffman coding is performed on each coefficient of the one-dimensionally arranged DCT blocks by the coefficient coding means (51 in FIG. 68) (S006 in FIG. 69).

The above processing (from S002 to S00 in FIG. 69)
6) is performed on all blocks (S007 in FIG. 69).
Thus, data of one frame can be compression-encoded. In the above processing, S003 to S0 in FIG.
06 is the same processing as the conventional signal processing system.

With the above-described configuration, after obtaining adjacent sample difference values, frequency decomposition is performed and encoding is performed, so that the magnitude of the absolute value of each coefficient after frequency decomposition can be reduced. The code amount can be reduced as compared with the related art. Note that the above process of obtaining adjacent sample difference values is an example, and the order of obtaining difference values between adjacent pixels is as follows.
It is not necessary to limit to the example of FIG. Further, the internal configuration of the basic encoding means (1) in FIG. 68 need not be limited to this example, and the configurations of the first to ninth embodiments of the present invention may be used.

<Eleventh Embodiment> Next, in order to solve the fourth problem of the conventional signal processing system,
An eleventh embodiment of the present invention will be described.

The fourth problem of the above-mentioned conventional signal processing system is that frequency decomposition such as DCT operation is performed as an integer and compression-encoded even though it is a real number operation.
Inverse DCT for the block once DCT operated
The original image block f (i, j) shown in FIG.
88, an error occurs between the original image block and the restored image block as in the example of the restored image block f (i, j) in FIG. (That is, it is not possible to obtain a completely restored image identical to the original image.) In the eleventh embodiment of the present invention, after performing frequency decomposition on each of the input signals (the original image blocks in this example), Once the frequency synthesis is performed, a restored image block is obtained, and the difference value between the original image block and the restored image block is encoded.The result of encoding the coefficient after frequency decomposition and the difference value are encoded. The fourth problem of the conventional signal processing system is solved by deriving a quantization step value that minimizes the total code amount as a result of the above, and outputting the code and the quantization step value at that time. .

FIG. 77 shows the configuration of a signal processing system (compression encoding side) according to the eleventh embodiment of the present invention. Figure 77
, The image data (or audio data) itself,
Alternatively, a difference between adjacent sample values of image data (or audio data) (hereinafter, adjacent sample difference value) is set as an original signal H, and 9 is an input of the original signal H, which is decomposed into a plurality of frequency components and Frequency decomposition means for outputting the coefficient A of the component, 10 is a quantization means for quantizing the coefficient A obtained from the frequency decomposition means (9) and outputting a coefficient B after quantization; 1 is at least With coefficient B as input,
Basic encoding means for outputting compressed encoded data.
Is a compression coded data storage means for storing the compression coded data, 13 is a dequantization means for dequantizing the coefficient B, and 709 is a frequency synthesis means for frequency synthesizing the dequantized coefficient B. 710 is a frequency synthesizing unit (7
09) is a frequency-synthesis result storage means for storing the frequency-synthesis result according to the following equation. Difference value storage means for storing a difference value M
Reference numeral 18 denotes a difference value encoding unit that encodes the difference value M and outputs difference value encoded data. 19 denotes a difference value encoded data storage unit that stores the difference value encoded data.
Reference numeral 20 denotes the total code amount of the compressed coded data stored in the compressed coded data storage means (11) and the difference value coded data stored in the difference value coded data storage means (19) ( Hereinafter, 21 is a total code amount calculation means for calculating the total code amount, and 2 is a total code amount storage means for storing the total code amount. 3 is a coefficient reading control means for generating an address signal D of the coefficient B stored in the frequency component storing means (2) and reading out the coefficient B from the frequency component storing means (2). Numeral 5 denotes coefficient encoding means for encoding the coefficient B read from the frequency component storage means (2). In the total code amount storage means (21), quantization means (10)
And a function of storing the total code amount corresponding to the quantization information used in the inverse quantization means (13), and 26 is a plurality of total code amounts stored in the total code amount storage means (21). Is a minimum code amount determining means for determining a minimum value from among the values of (i), and determining quantization information (hereinafter, minimum quantization information) that minimizes the total code amount. Quantized information encoding means for outputting quantized quantized information data. Reference numeral 28 denotes compressed encoded data (hereinafter, referred to as selection data R) corresponding to the minimum quantized information, and differential value encoding corresponding to the minimum quantized information. This is a selection output means for selecting data (hereinafter, selected data T) and outputting the selected data R, the selected data T, and the coded quantization information data as final compressed coded data.

In the following description, as an example, it is assumed that the frequency decomposition means (9) performs DCT, and the frequency synthesis means (7)
09) performs inverse DCT, and the difference value encoding means (18) and the coefficient encoding means (5) perform Huffman encoding.

The operation of the signal processing system (compression encoding side) according to the eleventh embodiment of the present invention will be described below with reference to FIG.

<Input> First, one block (8 pixels × 8 pixels) of the original image data for one frame.
(Original image block) (S0 in FIG. 78)
01).

<DCT> Next, frequency decomposition means (FIG. 77)
In 9) above, for the input original image block,
A DCT operation as shown in Equation 1 is performed (S002 in FIG. 78).

<Quantization> Next, quantization means (1 in FIG. 77)
In step 0), the quantization step value is set to “1” (S003 in FIG. 78). This quantization step value is used as quantization information. Then, each coefficient (DCT block) after DCT is quantized, and the quantized DCT block is stored in the frequency component storage means (2 in FIG. 77) (S0 in FIG. 78).
04). In the present embodiment (the eleventh embodiment of the present invention), the same DCT block is quantized and coded with a plurality of quantization step values. A quantization step value (minimum quantization information) that minimizes the amount is determined, and a code corresponding to the minimum quantization information is output. Here, as an example, for the same DCT block, the quantization step value is “
1 and the quantization with a quantization step value of “2”. For this purpose, the above processing (S in FIG. 78) is performed.
In 003), the quantization is first performed by setting the quantization step value to “1”. In addition, the quantization information is
It may be treated as a shift amount. For example, when the quantization step value is “2”, the binary value is changed to “right 1b”.
It is the same as "it shift".
It can also be handled as a bit shift amount. Further, a numerical value smaller than “1” may be used as the quantization step value. For example, when “0.5” is used, b
Considering the amount of it shift, "shift 1 bit to the left"
Same as what you did.

<Encoding> Next, read control means (FIG. 7)
In 7-3), an address signal D of the frequency component storage means (2 in FIG. 77) is generated, and each coefficient of the quantized DCT block is read out from the frequency component storage means (2 in FIG. 77). The outputted coefficients are encoded by coefficient encoding means (5 in FIG. 77) (S005 in FIG. 78). Then, the encoded data (compressed encoded data) is stored in the compressed encoded data storage means (11 in FIG. 77) (S006 in FIG. 78). At this point, the quantization step value is "1" in the compression-encoded data storage means (11 in FIG. 77).
, The corresponding compressed and encoded data is stored.

<Generation of Reconstructed Image Block> Next, each coefficient of the quantized DCT block is inversely quantized by the inverse quantization means (13 in FIG. 77) (S007 in FIG. 78).
Since the quantization step value is "1", "1 time"
I do. Next, a frequency synthesizing unit (709 in FIG. 77) performs an inverse DCT operation shown in Expression 2 on the inversely quantized DCT block to obtain a restored image block (FIG. 78).
S008). Then, the restored image block is stored in the frequency synthesis result storage means (710 in FIG. 77) (S009 in FIG. 78). At this point, the quantization step value is "1" in the frequency synthesis result storage means (710 in FIG. 77).
", The corresponding restored image block is stored.

<Acquisition of difference value> Next, the difference value calculating means (16 in FIG. 77) calculates the difference between each pixel of the original image block and each pixel of the restored image block corresponding to the case where the quantization step value is "1". The value is calculated (S010 in FIG. 78) and stored in the difference value storage means (17 in FIG. 77) (S010 in FIG. 78).
011). Then, the difference value encoding means (18 in FIG. 77)
Thus, the difference value is encoded to obtain difference value encoded data (S12 in FIG. 78), and stored in the difference value encoded data storage means (19 in FIG. 77) (S013 in FIG. 78). At this point, the difference value encoded data storage means (19 in FIG. 77) stores the difference value encoded data corresponding to the case where the quantization step value is “1”. This difference value may be encoded using an encoding method such as Huffman encoding or arithmetic encoding.

<Calculation of Total Code Amount> Next, the compressed code data corresponding to the case where the quantization step value is "1" and the quantization step value are calculated by the total code amount calculation means (20 in FIG. 77). The total code amount of the difference value encoded data corresponding to the case of “1” is calculated (S014 in FIG. 78),
It is stored in the total code amount storage means (21 in FIG. 77) (FIG. 7).
8 S015). At this time, the total code amount storage means (21 in FIG. 77) stores the total code amount corresponding to the case where the quantization step value is “1”.

The above processing (from S003 to S01 in FIG. 78)
5) is processing when the quantization step value is “1”. In the present embodiment, as an example, the same DCT block is subjected to quantization with a quantization step value of “1” and quantization with a quantization step value of “2”. Since only the processing when the step value is “1” has been completed, the quantization step value is set to “2” next.
”And perform the processing from S003 to S015 in FIG. 78. At the point of time when the processing up to this point is completed, the quantization step value is set to “1” in the total code amount storage means (21 in FIG. 77).
"And the total code amount corresponding to the case where the quantization step value is" 2 ".

Next, the minimum code amount determining means (2 in FIG. 77)
6), the plurality of total code amounts stored in the total code amount storage means (21 in FIG. 77) are compared, and the minimum total code amount and the quantization step value (minimum) corresponding to the minimum total code amount are compared. (Quantization information) is determined (S01 in FIG. 78).
7). Then, quantization information encoding means (27 in FIG. 77)
Thus, the minimum quantization information is encoded and output as encoded quantization information data (S018 in FIG. 78). Then, by the selection output means (28 in FIG. 77), the compressed coded data stored in the compressed coded data storage means (11 in FIG. 77) and the difference coded data storage means (FIG. 7)
7) From among the difference value encoded data stored in 19), compressed encoded data corresponding to the minimum quantization information and difference value encoded data corresponding to the minimum quantization information are selected, respectively. It is output as final compression encoded data. It also outputs encoded quantization information data.

The above processing (from S001 to S01 in FIG. 78)
9) is performed on all blocks.

When a large value is used as the quantization step value, the value of the coefficient after quantization becomes small. Although the number decreases, the error due to the quantization increases, so that the code amount of the difference value encoded data stored in the difference value encoded data storage means (19 in FIG. 77) may increase. Conversely, if a small value is used as the quantization step value, the error due to the quantization will be small. May be reduced, but since the value of the quantized coefficient increases, the code amount of the compressed coded data stored in the compressed coded data (11 in FIG. 77) may increase. However, in the present embodiment (the eleventh embodiment of the present invention), the input signal can be compression-encoded losslessly by the configuration shown in FIG. ) And the difference value coded data stored in the difference value coded data storage means (19 in FIG. 77), the output which always minimizes the total code amount. A signal can be obtained. The internal configuration of the basic encoding means (1 in FIG. 77) need not be limited to this, and the basic code of the other embodiments (the first to tenth embodiments of the present invention) already described. The configuration may be the same as that of the conversion means. Further, an adjacent sample difference calculating means is provided,
An adjacent sample difference value of the input signal is obtained, and
You may use as.

<Twelfth Embodiment> Next, in order to solve the fifth problem of the conventional signal processing system,
A twelfth embodiment of the present invention will be described.

[0370] The fifth problem of the conventional signal processing system is that it cannot cope with an operation error caused by a physical problem of the signal processing system itself. For example, a semiconductor component or the like used inside a signal processing system (hardware or a computer that performs signal processing) may not operate normally depending on a temperature condition in which the component is used. Generally, it is considered that a semiconductor component or the like is commercialized and put on the market after obtaining a sufficient temperature resistance test or the like. Therefore, it is considered that the device normally operates normally even under severe conditions. However, in recent years, opportunities to use various electric appliances outdoors, such as mobile phones and notebook personal computers, have been increasing. As described above, the electric appliances themselves have been used under more severe conditions than before. Therefore, it is conceivable that the device does not operate normally instantaneously. Thus, there is a possibility that a failure based on a natural phenomenon occurs. If the input signal does not operate normally instantaneously during the compression encoding of the input signal by the above-mentioned conventional signal processing system, the compression-encoded data has an unintended configuration. There was a problem (fifth problem) that the restoration could not be performed correctly even if it was performed. In the twelfth embodiment of the present invention, after an input signal is encoded (or converted in form), it is temporarily decoded (or inverted in form), and then decoded (or inverted in form). If an error occurs during (conversion) processing (or the processing result), the input signal itself is output, and if there is no error, an encoded (or format-converted) signal (coded conversion signal) is output. By doing so, the fifth problem of the conventional signal processing system described above is solved.

FIG. 79 shows the configuration of the signal processing system (compression encoding side) according to the twelfth embodiment of the present invention. FIG.
9, at least image data (or audio data) itself or data encoding the image data (or audio data) is used as an input signal W;
Is input signal storage means for storing the input signal W;
Is a coding conversion means for coding (or converting the format of) the input signal W and outputting a coding conversion signal.
Encoded conversion signal storage means for storing an encoded conversion signal
, 32 is a decoding inverse conversion means for decoding the encoded conversion signal (or inverse format conversion) and outputting a decoded inverse conversion signal, and 33 is a decoding inverse conversion means for storing the decoded inverse conversion signal. A conversion signal storage means for detecting an error in the decoding (or format inversion) processing of the encoded conversion signal in the decoding inverse conversion means (32) and outputting a first error signal indicating the presence or absence of the error; Error detection means
Calculates a change amount by comparing the input signal W with the decoded inverse transform signal, sets an error state when the change amount exceeds a predetermined value, and outputs a second error signal indicating whether the state is an error state. Means 36 outputs a signal indicating whether the first error signal or the second error signal is in an error state, and outputs an input signal when the first error signal or the second error signal is in an error state. Select and output W,
When the first error signal and the second error signal are not in an error state, this is a selection output means for selecting and outputting an encoded conversion signal.

The operation of the twelfth embodiment of the present invention will be described below with reference to FIG.

First, the input signal W is stored in the input signal storage means (29 in FIG. 79) (S00 in FIG. 80).
1).

Next, the encoding conversion means (30 in FIG. 79) encodes (or converts the format of) the input signal W and outputs an encoded conversion signal (S002 in FIG. 80).
Then, the coded conversion signal is stored in the coded conversion signal storage means (31 in FIG. 79) (S003 in FIG. 80).

Next, decoding inverse conversion means (32 in FIG. 79)
Decoding (or inverse format conversion) of the encoded conversion signal, and outputs a decoded inverse conversion signal (S in FIG. 80).
004). The error detection means (34 in FIG. 79) checks whether an error occurs during the decoding (or format inversion) processing of the encoded conversion signal, and outputs a first error signal for notifying the presence or absence of the error. Output (Fig. 8
0 S005). The error during the decoding (or inverse format conversion) of the coded transform signal is, for example, a case where the Huffman code is converted into a Huffman code when the coded transform signal is a DCT coefficient obtained by Huffman coding. When decrypted, D
Assume a case where a problem such as detection of more than a predetermined number of CT coefficients occurs. Then, when an error is detected, the selection output means (36 in FIG. 79) outputs a signal indicating that there is an error, and selects and outputs the input signal W (S006 in FIG. 80). If S in FIG.
In 005, if there is no error, the decoded inverse transform signal is stored in the decoded inverse transform signal storage means (33 in FIG. 79) (S007 in FIG. 80).

Next, the comparison operation means (35 in FIG. 79)
The amount of change is obtained by comparing the input signal W and the decoded inverse transform signal, and when the amount of change exceeds a predetermined value, the state is regarded as an error state, and a second error signal indicating whether the state is an error state is output (FIG. 80). S008). This embodiment (first embodiment of the present invention)
Assuming that the input signal W is image data, the change amount determined in the second embodiment) is, for example, the sum of the absolute values of the difference values of the respective pixels in the input signal W and the decoded inverse transform signal. If the value of the amount of change is extremely large, it can be determined that the encoding conversion process (S002 in FIG. 80) has not been performed normally. Therefore, the presence or absence of an error is determined by comparing the amount of change with a predetermined value. It is determined whether the second error signal is in an error state (S009 in FIG. 80). If the second error signal is in an error state, a signal indicating that there is an error is output by the selection output means (36 in FIG. 79). Signal W
Is selected and output (S006 in FIG. 80). It is determined whether the second error signal is in an error state (S00 in FIG. 80).
9) If not in an error state, select output means (FIG. 79)
36), a signal indicating that there is no error is output, and an encoded conversion signal is selected and output (S0 in FIG. 80).
10).

With the above configuration and operation, in the twelfth embodiment of the present invention, after the input signal W is encoded (or converted in format), it is temporarily decoded (or inversely converted in format). ), And if an error occurs during the decoding (or inverse format conversion) processing (or the processing result), the input signal itself is output. If there is no error, the encoding (or format conversion) is performed. By outputting the signal (encoded conversion signal), the fifth problem of the conventional signal processing system can be solved.

[0378]

As is apparent from the above description, the signal processing system of the present invention has the following first to fifth effects.

<First Effect> The first effect of the signal processing system of the present invention is that a scan order table suitable for coefficients after frequency decomposition of an input signal (image data or audio data) is created, and the created scan order table is obtained. Since the coefficients are sampled according to the order table, the number of zero coefficients to be encoded can be reduced as compared with the related art using the fixed scan order table, and the amount of information can be greatly reduced as a whole. . In addition, as preprocessing, “complexity” of input signals (image data or audio data)
Also, by performing the grouping of the "coefficient distribution", there is an effect of further reducing the amount of information.

<Second Effect> The second effect of the signal processing system of the present invention is as follows. In the prior art, one non-zero coefficient in each coefficient after frequency decomposition requires sign information of 1 [bit]. Although the input signal (image data or audio data) as a whole is extremely wasteful, the sign information is collectively encoded and compressed, or whether each coefficient is positive or negative is determined. By performing compression encoding after using the estimation method, the amount of these codes (the amount of information) can be significantly reduced.

<Third Effect> A third effect of the signal processing system of the present invention is to obtain a difference value between adjacent samples by utilizing the feature that adjacent image data (or audio data) is similar. By performing the frequency decomposition after that, the amplitude absolute value of each coefficient after the frequency decomposition can be reduced as compared with the related art, and the amount of information can be greatly reduced.

<Fourth Effect> The fourth effect of the signal processing system of the present invention is that, since the frequency decomposition is a real number operation, an image restored by frequency synthesis is simply converted to an integer and compression-coded. Compared to the prior art in which a large error occurs in the original image, the quantization step after frequency decomposition is changed, and the sum of the coding result of the coefficient after frequency decomposition and the coding result of the difference between the restored image and the original image is calculated. By detecting the quantization step in which the code amount of the minimum becomes the minimum and outputting the coding result corresponding to the quantization step, the input signal can always be compression-coded with a small code amount and without loss. That's what it means

<Fifth Effect> The fifth effect of the signal processing system of the present invention is that, when an instantaneous operation error occurs due to a physical problem of the signal processing system itself, the signal cannot be restored by the conventional technology. When an error occurs while the compression-encoded data is output, the compression-encoding is temporarily expanded and decoded (decoded) during compression-encoding to check whether the compressed-encoded data can be restored. By outputting the original signal (input signal), the original signal (input signal W) can be restored from the output signal.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a signal processing system according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing an operation of the signal processing system according to the first embodiment of the present invention.

FIG. 3 is a diagram illustrating an outline of an operation of the signal processing system according to the first embodiment of the present invention.

FIG. 4 is a diagram illustrating an example of a DCT block (DCT block 2).

FIG. 5 is a diagram illustrating an example of a DCT block (DCT block 3).

FIG. 6 is a diagram illustrating the number of zero coefficients of each frequency component of a DCT block.

FIG. 7 is a diagram illustrating a scan order of coefficients of each frequency component of a DCT block.

FIG. 8 is a diagram showing a manner of scanning coefficients of each frequency component of the DCT block 2.

FIG. 9 is a diagram showing a manner of scanning coefficients of each frequency component of the DCT block 3;

FIG. 10 is a configuration diagram of a signal processing system according to a second embodiment of the present invention.

FIG. 11 is a flowchart showing an operation of the signal processing system according to the second embodiment of the present invention.

FIG. 12 is a diagram showing values of the intensity of each frequency component of the DCT block.

FIG. 13 is a diagram illustrating a value of a coefficient strength of each frequency component of a DCT block and a scan order.

FIG. 14 is a diagram illustrating a manner of scanning coefficients of each frequency component of the DCT block 2;

FIG. 15 is a diagram illustrating a manner of scanning coefficients of each frequency component of the DCT block 3;

FIG. 16 is a configuration diagram of a signal processing system according to a third embodiment of the present invention.

FIG. 17 is a flowchart showing an operation of the signal processing system according to the third embodiment of the present invention.

FIG. 18 is a diagram illustrating an example in which a DCT block is classified into two regions (region A and region B).

FIG. 19 is a configuration diagram of a signal processing system according to a fourth embodiment of the present invention.

FIG. 20 is a flowchart illustrating an operation of the signal processing system according to the fourth embodiment of the present invention.

FIG. 21 is a diagram illustrating an example in which DCT blocks are classified into two groups, a complex group and a monotone group.

FIG. 22 is a configuration diagram of a signal processing system according to a fifth embodiment of the present invention.

FIG. 23 is a flowchart illustrating an operation of the signal processing system according to the fifth embodiment of the present invention.

FIG. 24 is a diagram showing an original image block.

FIG. 25 is a diagram showing a DCT block.

FIG. 26 is a configuration diagram of a signal processing system according to a sixth embodiment of the present invention.

FIG. 27 is a flowchart illustrating an operation of the signal processing system according to the sixth embodiment of the present invention.

FIG. 28 is a diagram showing a reference DCT block.

FIG. 29 is a diagram showing DCT block 0 in which coefficient A63 is positive.

FIG. 30 is a diagram showing DCT block 1 in which coefficient A63 is negative.

FIG. 31 is a diagram showing a reference inverse DCT block.

FIG. 32 is a diagram showing an inverse DCT block 0;

FIG. 33 is a diagram showing an inverse DCT block 1;

FIG. 34 is a diagram showing a conceptual image of a state where the first change amount is large.

FIG. 35 is a diagram illustrating a conceptual image of a state in which the first change amount is small.

FIG. 36 is a diagram illustrating a contour (edge) shape of an original image block.

FIG. 37 is a diagram illustrating a state where each value of a reference inverse DCT block is reduced to 1 / 10,000.

FIG. 38 is a diagram illustrating a difference between an outline (edge) shape of an original image block and a reference inverse DCT block (1 / 10,000).

FIG. 39 is a configuration diagram of a signal processing system according to a seventh embodiment of the present invention.

FIG. 40 is a flowchart showing an operation of the signal processing system according to the seventh embodiment of the present invention.

FIG. 41 is a diagram showing DCT block 0 in which coefficient A1 is positive.

FIG. 42 is a diagram showing DCT block 1 in which coefficient A1 is negative.

FIG. 43 is a diagram showing an inverse DCT block 0;

FIG. 44 is a diagram showing an inverse DCT block 1;

FIG. 45 is a diagram showing a conceptual image of the first similarity.

FIG. 46 is a diagram showing that the positive / negative first similarity and the negative / positive first change amount have the same value.

FIG. 47 is a diagram showing DCT block 0 in which coefficient A2 is positive.

FIG. 48 is a diagram showing DCT block 1 in which coefficient A2 is negative.

FIG. 49 is a diagram showing an inverse DCT block 0;

FIG. 50 is a diagram showing an inverse DCT block 1;

FIG. 51 is a configuration diagram of a signal processing system according to an eighth embodiment of the present invention.

FIG. 52 is a flowchart showing an operation of the signal processing system according to the eighth embodiment of the present invention.

FIG. 53 is a diagram showing DCT block 0 in which coefficient A1 is positive.

FIG. 54 is a diagram showing DCT block 1 in which coefficient A1 is negative.

FIG. 55 is a diagram showing an inverse DCT block 0;

FIG. 56 is a diagram showing an inverse DCT block 1;

FIG. 57 is a diagram showing a conceptual image of the first similarity.

FIG. 58 is a diagram showing a conceptual image of a degree of difference.

FIG. 59 is a diagram showing DCT block 0 in which coefficient A2 is positive.

FIG. 60 is a diagram showing DCT block 1 in which coefficient A2 is negative.

FIG. 61 is a diagram showing an inverse DCT block 0;

FIG. 62 is a diagram showing an inverse DCT block 1;

FIG. 63 is a configuration diagram of a signal processing system according to a ninth embodiment of the present invention.

FIG. 64 is a flowchart illustrating an operation of the signal processing system according to the ninth embodiment of the present invention.

FIG. 65: Sign number N and sign of each coefficient (sign information)
FIG.

FIG. 66 shows coefficients A1, A2, A3, A4,
FIG. 10 is a diagram illustrating a DCT block 0000 in which is positive.

FIG. 67 is a diagram showing an inverse DCT block 0000;

FIG. 68 is a configuration diagram of a signal processing system according to a tenth embodiment of the present invention.

FIG. 69 is a flowchart showing an operation of the signal processing system according to the tenth embodiment of the present invention.

FIG. 70 is a diagram illustrating the direction of calculation of adjacent sample difference values.

FIG. 71 is a diagram showing an original image block.

FIG. 72 is a diagram showing adjacent sample difference values.

FIG. 73 is a diagram showing a DCT result for adjacent sample difference values.

FIG. 74 is a diagram illustrating a DCT result for an original image block.

FIG. 75 is a diagram illustrating a value obtained by subtracting 128 from each value of an original image block.

FIG. 76 is a diagram illustrating a DCT result for a value obtained by subtracting 128 from each value of an original image block.

FIG. 77 is a configuration diagram of a signal processing system according to an eleventh embodiment of the present invention.

FIG. 78 is a flowchart showing an operation of the signal processing system according to the eleventh embodiment of the present invention.

FIG. 79 is a configuration diagram of a signal processing system according to a twelfth embodiment of the present invention.

FIG. 80 is a flowchart showing an operation of the signal processing system according to the twelfth embodiment of the present invention.

FIG. 81 is a configuration diagram of a conventional signal processing system (compression encoding side).

FIG. 82 is a flowchart showing an operation of a conventional signal processing system (compression encoding side).

FIG. 83 is a diagram showing an outline of an operation of a conventional signal processing system (compression encoding side).

FIG. 84 is a diagram showing an original image block and a DCT block.

FIG. 85 is a configuration diagram of a conventional signal processing system (extension decoding side).

FIG. 86 is a flowchart showing an operation of a conventional signal processing system (extension decoding side).

FIG. 87 is a diagram showing an outline of operation of a conventional signal processing system (decompression decoding side).

FIG. 88 is a diagram showing a DCT block and a restored image block.

FIG. 89 is a diagram illustrating an example of a DCT block (DCT block 1).

FIG. 90 is a diagram illustrating an example of a DCT block (DCT block 2).

FIG. 91 is a diagram illustrating an example of a DCT block (DCT block 3).

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Basic encoding means 2 Frequency component storage means 21 Frequency component storage means 25 Coefficient writing control means 3 Coefficient reading control means 4 Read control information generation means 401 Scan order determination means 402 Scan order table 403 Scan order information encoding means 404 Complex S group grouping means 405 Complexity group information storage means 406 Complexity group information coding means 407 Coefficient distribution determination means 408 Coefficient distribution information storage means 409 Coefficient distribution information coding means 410 Zero coefficient counting means 411 Priority determination means 412 Coefficient strength Arithmetic means 450 Fixed scan order table 5 Coefficient encoding means 51 Coefficient encoding means 52 Coefficient decoding means 6 Absolute value conversion means 7 Signature information processing means 701 Non-zero determination signature information extraction means 702 Signature information storage means 703 Information encoding means 704 coefficient strength rank determination means 705 coefficient strength rank table 706 sign information estimation means 707 signature information setting means 708 coefficient information storage means with signature information 709 frequency synthesis means 710 frequency synthesis result storage means 711 first signature information estimation parameter Calculation means 712 First signature information estimation parameter storage means 713 First signature information generation means 714 Second signature information estimation parameter calculation means 715 Second signature information estimation parameter storage means 716 Third signature information estimation parameter calculation means 717 Third signature information Estimation parameter storage means 718 Second sign information generation means 719 Third sign information generation means 720 Inverse quantization means 8 Adjacent sample difference calculation means 9 Frequency decomposition means 10 Quantization means 11 Compressed encoded data storage means 1 Inverse quantization means 16 Difference value calculation means 17 Difference value storage means 18 Difference value encoding means 19 Difference value encoded data storage means 20 Total code amount calculation means 21 Total code amount storage means 26 Minimum code amount determination means 27 Quantization information Encoding means 28 Selection output means 29 Input signal storage means 30 Encoding conversion means 31 Encoding conversion signal storage means 32 Decoding inverse conversion means 33 Decoding inverse conversion signal storage means 34 Error detection means 35 Comparison operation means 36 Selection output means 37 adjacent sample difference value storage means 40 complexity determination means C coefficient read control information D address signal E scan order information F complexity information G complexity group information H original signal

Claims (28)

[Claims] 1. A compression-encoded data receiving at least a coefficient A obtained by decomposing at least image data (or audio data) into a plurality of frequency components (or a coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least a coefficient A (or coefficient B) inside the basic coding means (1); and a frequency component storage means ( 2) a coefficient reading control means (3) for generating an address signal D of the coefficient A (or coefficient B) stored in 2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2); A (or coefficient B) as input and read control information generating means for generating read control information C for reading coefficient A (or coefficient B) stored in frequency component storage means (2) And 4) the coefficients A (or coefficient B) has a coefficient encoding means for encoding (5),
Scanning order determining means (401) for determining the scanning order of each frequency component in the frequency component storage means (2) from the input coefficient A (or coefficient B) inside the read control information generating means (4); , Information on the scan order determined by the scan order determining means (401)
Scan order table (4) storing scan order information E)
02), and scan order information encoding means (403) for encoding the scan order information E. The coefficient read control means (3) uses at least the scan order information E as the read control information C, A signal processing system, wherein an address signal D is generated based on control information C. 2. A compression coded data receiving at least a coefficient A obtained by decomposing at least image data (or audio data) into a plurality of frequency components (or a coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least a coefficient A (or coefficient B) inside the basic coding means (1); and a frequency component storage means ( 2) a coefficient reading control means (3) for generating an address signal D of the coefficient A (or coefficient B) stored in 2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2); The complexity information F of data (or audio data) is input, and read control information C for reading the coefficient A (or coefficient B) stored in the frequency component storage means (2) is generated. A readout control information generating means (4); and a coefficient coding means (5) for coding the coefficient A (or coefficient B). A complexity group classification means (404) for classifying each frequency component in the frequency component storage means (2) into a plurality of complexity groups
A complexity group information storage means (405) for storing the complexity group information G classified by the complexity group classification means (404); and a complexity group information encoding means (104) for encoding the complexity group information G. 406), and at least the complexity group information G is used as the read control information C, and the coefficient read control means (3) generates the address signal D based on at least the read control information C. A signal processing system characterized by the above-mentioned. 3. A compression-encoded data input with at least a coefficient A (or a coefficient B obtained by quantizing the coefficient A) obtained by decomposing at least image data (or audio data) into a plurality of frequency components. And a frequency component storage means (2) for storing at least a coefficient A (or coefficient B) inside the basic coding means (1); and a frequency component storage means ( A coefficient reading control means (3) for generating an address signal D of the coefficient A (or coefficient B) stored in 2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2); An absolute value converting means (6) for obtaining the absolute value of the coefficient (or coefficient B); a coefficient coding means (5) for coding the absolute value of each coefficient A (or coefficient B) of the frequency component;
(Or coefficient B) as input, and a sign information processing means (7) for processing a code (hereinafter, sign information) indicating whether the input coefficient A (or coefficient B) is positive or negative, In the signature information processing means (7), the coefficient A
It is determined whether or not (or coefficient B) is non-zero, and when it is non-zero, non-zero determination signature information extraction means (701) for extracting signature information, and signature information storage means (702) for storing signature information And a sign information encoding unit (703) for performing statistical processing on the sign information and encoding the sign information. 4. At least a coefficient A (or a coefficient B obtained by quantizing the coefficient A) obtained by decomposing at least image data (or audio data) into a plurality of frequency components is input to the compression-encoded data. And a frequency component storage means (2) for storing at least a coefficient A (or coefficient B) inside the basic coding means (1); and a frequency component storage means ( 2) a coefficient reading control means (3) for generating an address signal D of the coefficient A (or coefficient B) stored in 2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2); A (or coefficient B) as input and read control information generating means for generating read control information C for reading coefficient A (or coefficient B) stored in frequency component storage means (2) And 4) the absolute value converting means for obtaining the absolute value of the coefficient A (or coefficient B) (6),
A coefficient encoding means (5) for encoding the absolute value of each coefficient A (or coefficient B) of the frequency component; and a sine of the inputted coefficient A (or coefficient B) with the coefficient A (or coefficient B) as an input. It has a signature information processing means (7) for processing information, and inside the read control information generating means (4),
Scan order determining means (401) for determining the scan order of each frequency component in the frequency component storage means (2) from the input coefficient A (or coefficient B), and the scan order determining means (401) A scan order table (402) for storing scan order information (hereinafter referred to as scan order information E); and a scan order information encoding unit (403) for encoding the scan order information E. , The coefficient read control means (3) generates an address signal D based on at least the read control information C, and the coefficient A ( Alternatively, it is determined whether or not the coefficient B) is non-zero. Sign information storage means (702) for storing sign information;
A signal processing system comprising a signature information encoding unit (703) for performing statistical processing and encoding signature information. 5. A compression-encoded data which receives at least a coefficient A obtained by decomposing at least image data (or audio data) into a plurality of frequency components (or a coefficient B obtained by quantizing the coefficient A). And a frequency component storage means (2) for storing at least a coefficient A (or coefficient B) inside the basic coding means (1); and a frequency component storage means ( 2) a coefficient reading control means (3) for generating an address signal D of the coefficient A (or coefficient B) stored in 2) and reading out the coefficient A (or coefficient B) from the frequency component storage means (2); The complexity information F of data (or audio data) is input, and read control information C for reading the coefficient A (or coefficient B) stored in the frequency component storage means (2) is generated. Extraction control information generating means (4); absolute value converting means (6) for obtaining the absolute value of coefficient A (or coefficient B); and encoding the absolute value of each coefficient A (or coefficient B) of the frequency component. A coefficient encoding means (5); a sign information processing means (7) which receives a coefficient A (or coefficient B) as input and processes sign information of the inputted coefficient A (or coefficient B); A complexity group classification unit (404) for classifying each frequency component in the frequency component storage unit (2) into a plurality of complexity groups based on the input complexity information F inside the generation unit (4); Group classification means (4
04) that includes a complexity group information storage unit (405) that stores the complexity group information G classified according to (4) and a complexity group information encoding unit (406) that encodes the complexity group information G 2. The group information G is used as read control information C, and the coefficient read control means (3) generates an address signal D based on at least the read control information C. In step (1), it is determined whether the coefficient A (or coefficient B) is non-zero. If the coefficient A (or coefficient B) is non-zero, the sign information is extracted. Means (702) and a signature information encoding means (703) for performing statistical processing on the signature information and encoding the signature information. A signal processing system characterized by. 6. A difference between adjacent sample values of image data (or audio data) (hereinafter, an adjacent sample difference value)
, And frequency decomposition means (9) for decomposing the adjacent sample difference value into a plurality of frequency components and outputting a coefficient A of each frequency component; and at least a coefficient A (or It has a basic encoding means (1) for receiving a coefficient B) which is a result of quantizing the coefficient A as input and outputting compressed encoded data, and the basic encoding means (1)
And a frequency component storage means (2) for storing at least a coefficient A (or coefficient B), and an address signal D for the coefficient A (or coefficient B) stored in the frequency component storage means (2). , A coefficient reading control means (3) for reading the coefficient A (or coefficient B) from the frequency component storage means (2), and a coefficient coding means (5) for coding the coefficient A (or coefficient B). A signal processing system characterized by the above-mentioned. 7. An image signal (or audio data) itself or a difference between adjacent sample values of image data (or audio data) (hereinafter, an adjacent sample difference value) is set as an original signal H, and the original signal H is input. Frequency decomposition means (9) for decomposing the original signal H into a plurality of frequency components and outputting coefficients A of each frequency component; and quantizing the coefficient A obtained from the frequency decomposition means (9), A quantizing means (10) for outputting a coefficient B; a basic encoding means (1) for receiving at least the coefficient B as input and outputting compressed and encoded data; and a compressed and encoded data storage means for storing the compressed and encoded data ( 11), an inverse quantization means (13) for inversely quantizing the coefficient B, a frequency synthesizing means (709) for frequency synthesizing the inversely quantized coefficient B, and a frequency synthesizing means by the frequency synthesizing means (709). 710, a difference value calculating means 16 for calculating a difference value (hereinafter, a difference value M) between the original signal H and the frequency synthesis result, and a difference storing the difference value M. Value storage means (17), difference value encoding means (18) for encoding the difference value M and outputting the difference value encoded data, and difference value encoded data storage means (19) for storing the difference value encoded data ), The compressed encoded data stored in the compressed encoded data storage means (11), and the differential encoded data stored in the differential encoded data storage means (19). Hereinafter, a total code amount calculation means (20) for calculating the total code amount) and a total code amount storage means (21) for storing the total code amount are provided. Frequency component storage means (2) for storing A coefficient reading control unit (3) for generating an address signal D of the coefficient B stored in the frequency component storage unit (2) and reading out the coefficient B from the frequency component storage unit (2); and a frequency component storage unit (2) And a coefficient encoding means (5) for encoding the coefficient B read from the storage means.
0) and a function of storing the total code amount corresponding to the quantization information used in the inverse quantization means (13), and a function of storing a plurality of total code amounts stored in the total code amount storage means (21). Minimum code amount determining means (26) for determining the minimum value from the values and determining quantization information (hereinafter referred to as minimum quantization information) for minimizing the total code amount; It has a quantization information encoding unit (27) for outputting quantization information data, and has compression encoding data (hereinafter, selection data R) corresponding to the minimum quantization information and a difference value code corresponding to the minimum quantization information. Data (hereinafter, selection data T)
And a selection output means (28) for outputting the selected data R, the selected data T, and the coded quantization information data as final compressed coded data. 8. An input signal storage means (29) for storing at least image data (or audio data) itself or data encoding image data (or audio data) as an input signal W, and storing the input signal W. , An encoding conversion means (30) for encoding the input signal W (or format conversion) and outputting an encoded conversion signal, and an encoded conversion signal storage means (31) for storing the encoded conversion signal.
Perform decoding (or inverse format conversion) of the coded conversion signal,
Decoding inverse conversion means (32) for outputting a decoding inverse conversion signal
A decoding inverse conversion signal storage means (33) for storing the decoding inverse conversion signal; and an error in decoding (or format inversion) processing of the encoded conversion signal in the decoding inverse conversion means (32). An error detecting means (34) for outputting a first error signal indicating the presence / absence of an error; and comparing the input signal W with the decoded inverse transform signal to determine a change amount, and when the change amount exceeds a predetermined value. A comparison operation means (35) for outputting a second error signal indicating whether the state is an error state, and outputting a signal indicating whether the first error signal or the second error signal is an error state; First
When the error signal or the second error signal is in the error state, the input signal W is selected and output. When the first error signal and the second error signal are not in the error state, the coded conversion signal is selected. A signal output system having a selection output means (36) for outputting the signal. 9. A configuration in which the coefficient A (or coefficient B) is input to the read control information generating means (4), and the input coefficient A (or coefficient B) is input inside the read control information generating means (4). A scan order determining means (401) for determining a scan order of each frequency component in the frequency component storage means (2); and information on the scan order determined by the scan order determining means (401) (hereinafter, scan order information) A scan order table (402) for storing E);
A scan order information encoding unit (403) for encoding the scan order information E;
6 is used as read control information C, and the coefficient read control means (3) generates the address signal D based on at least the read control information C. 6. The signal processing according to claim 5, wherein system. 10. A zero coefficient counting means (410) for counting the number of zero coefficients (or non-zero coefficients) for each frequency component for a coefficient A (or coefficient B) inside a scan order determining means (401). And the number of zero coefficients for each frequency component counted by the zero coefficient counting means (410) are compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is given priority. 2. The signal processing system according to claim 1, further comprising a priority order determining means (411) for ranking the data. 11. A zero coefficient counting means (410) for counting the number of zero coefficients (or non-zero coefficients) for each frequency component for a coefficient A (or coefficient B) inside a scan order determining means (401). And the number of zero coefficients for each frequency component counted by the zero coefficient counting means (410) are compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is given priority. 5. The signal processing system according to claim 4, further comprising a priority order determining means (411) for ranking the data. 12. A zero coefficient counting means (410) for counting the number of zero coefficients (or non-zero coefficients) for each frequency component for a coefficient A (or coefficient B) inside a scanning order determining means (401). And the number of zero coefficients for each frequency component totaled by the zero coefficient counting means (410) are compared, and the scanning order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized. 10. The signal processing system according to claim 9, further comprising a priority order determining means (411) for ranking the data. 13. A coefficient strength calculating means (412) for obtaining the strength of a coefficient for each frequency component for a coefficient A (or coefficient B) inside a scan order determining means (401).
And at least the coefficient strength value for each frequency component as input, compare the coefficient strength values for each frequency component, and prioritize the scanning order of frequency components with the larger coefficient strength values When the number of zero coefficients for each frequency component is input to the priority determining means (411), the values of the coefficient intensities are compared first, and the Prioritize the scanning order of frequency components with large intensity values, and compare the number of zero coefficients (or non-zero coefficients) when there are multiple frequency components with the same magnitude. do it,
11. The apparatus according to claim 1, wherein a scan order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized.
The signal processing system according to claim 1. 14. A coefficient strength calculating means (412) for obtaining a coefficient strength for each frequency component for a coefficient A (or coefficient B) inside a scan order determining means (401).
And at least the coefficient strength value for each frequency component as input, compare the coefficient strength values for each frequency component, and prioritize the scanning order of frequency components with the larger coefficient strength values When the number of zero coefficients for each frequency component is input to the priority determining means (411), the values of the coefficient intensities are first compared, Prioritize the scanning order of frequency components with large intensity values, and compare the number of zero coefficients (or non-zero coefficients) when there are multiple frequency components with the same magnitude. do it,
12. The apparatus according to claim 4, wherein a scan order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized.
The signal processing system according to claim 1. 15. A coefficient strength calculation means (412) for obtaining the strength of a coefficient for each frequency component for a coefficient A (or coefficient B) inside a scan order determination means (401).
And at least the coefficient strength value for each frequency component as input, compare the coefficient strength values for each frequency component, and prioritize the scanning order of frequency components with the larger coefficient strength values When the number of zero coefficients for each frequency component is input to the priority determining means (411), the values of the coefficient intensities are first compared, Prioritize the scanning order of frequency components with large intensity values, and compare the number of zero coefficients (or non-zero coefficients) when there are multiple frequency components with the same magnitude. do it,
13. The method according to claim 9, wherein a scan order of frequency components having a small number of zero coefficients (or a large number of non-zero coefficients) is prioritized.
The signal processing system according to claim 1. 16. The read control information generating means (4)
The configuration is such that complexity information F 2 of image data (or audio data) is inputted, and each frequency in the frequency component storage means (2) is read from the inputted complexity information F inside the read control information generation means (4). A complexity group classifier (40) for classifying components into a plurality of complexity groups
4) a complexity group information storage means (405) for storing the complexity group information G classified by the complexity group classification means (404); and a complexity group information encoding for encoding the complexity group information G. Means (40
6), wherein at least the complexity group information G is used as read control information C, and the coefficient read control means (3) generates the address signal D based on at least the read control information C. Claim 1, Claim 10, or Claim 13 characterized by the above-mentioned.
The signal processing system according to claim 1. 17. A read control information generating means (4) for determining which frequency component the energy is biased from based on the value of the coefficient A (or coefficient B) to obtain information indicating the bias status (hereinafter referred to as “bias status”). Coefficient distribution determining means (407) for outputting coefficient distribution information), coefficient distribution information storing means (408) for storing coefficient distribution information, and coefficient distribution information encoding means (409) for encoding coefficient distribution information. The scan order determining means (401) determines the scan order for each coefficient distribution information, and stores the scan order information E for each coefficient distribution information in the scan order table (402). The signal processing system according to claim 1, claim 10, or claim 13, or claim 16. 18. A read control information generating means (4) determines which frequency component the energy is biased from the value of the coefficient A (or the coefficient B), and obtains information indicating the bias status (hereinafter, referred to as bias information). Coefficient distribution determining means (407) for outputting coefficient distribution information), coefficient distribution information storing means (408) for storing coefficient distribution information, and coefficient distribution information encoding means (409) for encoding coefficient distribution information. The scan order determining means (401) determines the scan order for each coefficient distribution information, and stores the scan order information E for each coefficient distribution information in the scan order table (402). The signal processing system according to claim 4, claim 9, claim 11, claim 12, claim 14, or claim 15. 19. Inside the sine information processing means (7), among the coefficients of each frequency component in the frequency component storage means (2), the energy intensity of each coefficient in the unit of frequency decomposition (hereinafter referred to as "the intensity of energy"). , Strength), and a coefficient strength rank determining means (704) for determining the rank of the strength of each coefficient (hereinafter referred to as a coefficient strength rank); a coefficient strength rank table (705) for storing the coefficient strength rank; Sign information is preferentially estimated from the coefficient with the highest rank, and from the estimation result,
It has a signature information estimating means (706) for generating and outputting new signature information. The non-zero determination signature information extracting means (701) determines whether each coefficient is non-zero. Signature information estimating means (706)
Or a configuration for executing the processing in the above. Claim 3, Claim 4, Claim 5, Claim 9, Claim 11, Claim 12, Claim 12, Claim 14, or Claim Item 19. The signal processing system according to Item 15 or 18. 20. A signature information processing means (7),
In the sign information estimating means (706), for each coefficient of the frequency component in the frequency component storing means (2), temporary sign information (hereinafter, temporary sign information) and a true sign The signature information setting means (707) for setting information (hereinafter, true signature information) and the signature information setting means (707) provide temporary signature information and each coefficient (hereinafter, referred to as true signature information) for which true signature information is set as necessary. Coefficient storage means with signature information (708)
A frequency component synthesizing unit (709) that synthesizes each frequency component of the coefficient with signature information and outputs a frequency synthesis result with signature information; and a frequency synthesis result storage unit (710) that stores the frequency synthesis result with signature information. A first signature information estimation parameter calculating means (711) for calculating a first signature information estimation parameter (hereinafter, a first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. And the first
First signature information estimation parameter storage means (712) for storing signature information estimation parameters; and estimating which temporary signature information is true signature information from the value of the first signature information estimation parameter corresponding to each temporary signature information. A first sign information generating means (713) for outputting match / mismatch information indicating whether or not the estimated sign information and the true sign information match as new sign information; and a sign information setting means (707). ) Has a function of setting true signature information with respect to a coefficient for which estimation of whether or not the temporary signature information is true signature information is performed, and a coefficient in which true signature information is set (hereinafter, a coefficient with true signature information) ) (Hereinafter referred to as true sign frequency synthesis result)
The sum of the absolute value of the difference between the frequency and the coefficient (hereinafter referred to as the provisional signature frequency combined coefficient) in which the provisional signature information is set (hereinafter referred to as the provisional signature frequency combination result) is defined as the first variation, and the true signature is obtained. Regarding the frequency synthesis result in the mixed state of the information-added coefficient and the temporary signature information-added coefficient (hereinafter, the mixed frequency synthesis result), the waveform of the mixed frequency synthesis result in the center axis of the waveform and the region surrounded by the waveform The sum of the absolute value of the difference between the waveform and the central axis of the waveform is defined as a first similarity,
The signature information estimation parameter calculation means (711) calculates the first change amount or the first similarity as the first signature information estimation parameter, and the first signature information generation means (713) calculates the first signature information corresponding to each temporary signature information. First
If the signature information estimation parameter is the first change amount, the first
The temporary signature information when the value of the signature information estimation parameter is small is estimated as true signature information, and when the first signature information estimation parameter corresponding to each temporary signature information is the first similarity, the first signature information estimation is performed. 20. The signal processing system according to claim 19, wherein the temporary signature information when the value of the parameter is large is estimated to be true signature information. 21. Sign information processing means (7)
In the sign information estimating means (706), for each coefficient of the frequency component in the frequency component storing means (2), temporary sign information (hereinafter, temporary sign information) and a true sign The signature information setting means (707) for setting information (hereinafter, true signature information) and the signature information setting means (707) provide temporary signature information and each coefficient (hereinafter, referred to as true signature information) for which true signature information is set as necessary. Coefficient storage means with signature information (708)
A frequency component synthesizing unit (709) that synthesizes each frequency component of the coefficient with signature information and outputs a frequency synthesis result with signature information; and a frequency synthesis result storage unit (710) that stores the frequency synthesis result with signature information. A first signature information estimation parameter calculating means (711) for calculating a first signature information estimation parameter (hereinafter, a first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. And the first
A first signature information estimation parameter storage means (712) for storing signature information estimation parameters; and a second signature information estimation parameter (hereinafter, referred to as a second signature information parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. A second signature information estimation parameter calculating means (714) for calculating the signature information estimation parameter), a second signature information estimation parameter storage means (715) for storing the second signature information estimation parameter, and a first signature information estimation parameter. Using the second signature information estimation parameter, calculate a third signature information estimation parameter (hereinafter, a third signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. A signature information estimation parameter calculating means (716), and a third signature information estimation parameter. From the third signature information estimation parameter storage means (717) to remember and the value of the third signature information estimation parameter corresponding to each temporary signature information, it is estimated which temporary signature information is true signature information, and the estimated signature information is estimated. And second signature information generating means (718) for outputting, as new signature information, coincidence / non-coincidence information indicating whether or not the signature information matches the true signature information. It has a function of setting true signature information for a coefficient for which estimation of whether or not the information is true signature information is completed, and a frequency synthesis result based on a coefficient for which true signature information is set (hereinafter, a coefficient with true signature information). (Hereinafter, the result of the true signature frequency synthesis) and the frequency in which the temporary signature information is set (hereinafter, the coefficient with the temporary signature information) (hereinafter, the temporary signature information). The sum of the absolute value of the difference from the in-frequency synthesis result) is defined as a first variation, and the frequency synthesis result in the mixed state of the coefficient with the true sign information and the coefficient with the temporary sign information (hereinafter, the mixed frequency synthesis result) The sum of the absolute values of the differences between the center axis of the waveform and the center axis of the waveform in the region surrounded by the center axis of the waveform of the mixed frequency synthesis result is defined as the first similarity. The difference between the waveform and the value range in a region outside the predetermined value range is defined as the degree of difference, and the first sign information estimation parameter calculation means (711)
In (2), the first change amount or the first similarity is calculated as a first signature information estimation parameter, and the second signature information estimation parameter calculation means (714) calculates the difference as a second signature information estimation parameter. One change or difference is normalized by a normalization coefficient, and a value obtained by adding the normalized difference to the normalized first change is defined as a second change. A value obtained by subtracting the normalized dissimilarity degree from the normalized first similarity degree by the normalization coefficient and the normalized similarity degree is defined as a second similarity degree. The change amount or the second similarity is calculated as a third signature information estimation parameter, and the second signature information generation means (71
In 8), when the third signature information estimation parameter corresponding to each temporary signature information is the second variation, the temporary signature information when the value of the third signature information estimation parameter is small is estimated as true signature information. If the third signature information estimation parameter corresponding to each temporary signature information is the second similarity,
20. The signal processing system according to claim 19, wherein the temporary signature information when the value of the third signature information estimation parameter is large is estimated to be true signature information. 22. A signature information processing means (7)
In the sign information estimating means (706), for each coefficient of the frequency component in the frequency component storing means (2), temporary sign information (hereinafter, temporary sign information) and a true sign The signature information setting means (707) for setting information (hereinafter, true signature information) and the signature information setting means (707) provide temporary signature information and each coefficient (hereinafter, referred to as true signature information) for which true signature information is set as necessary. Coefficient storage means with signature information (708)
A frequency component synthesizing unit (709) that synthesizes each frequency component of the coefficient with signature information and outputs a frequency synthesis result with signature information; and a frequency synthesis result storage unit (710) that stores the frequency synthesis result with signature information. A first signature information estimation parameter calculating means (711) for calculating a first signature information estimation parameter (hereinafter, a first signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. And the first
A first signature information estimation parameter storage means (712) for storing signature information estimation parameters; and a second signature information estimation parameter (hereinafter, referred to as a second signature information parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. A second signature information estimation parameter calculating means (714) for calculating the signature information estimation parameter), a second signature information estimation parameter storage means (715) for storing the second signature information estimation parameter, and a first signature information estimation parameter. Using the second signature information estimation parameter, calculate a third signature information estimation parameter (hereinafter, a third signature information estimation parameter) corresponding to the frequency synthesis result with signature information in which each temporary signature information is set. A signature information estimation parameter calculating means (716), and a third signature information estimation parameter. Based on the third signature information estimation parameter storage means (717) and the value of the third signature information estimation parameter corresponding to each temporary signature information, the combination of the temporary signature information that matches the combination of the true signature information is obtained. A third signature information generating unit (719) for adding new signature information and outputting new signature information; and in the signature information setting unit (707), estimating whether the temporary signature information is true signature information. Has the function of setting the true sign information for the coefficient for which the signature has been completed, and the frequency synthesis result (hereinafter, the true sine frequency synthesis result) based on the coefficient for which the true signature information is set (hereinafter, the coefficient with true signature information). , The difference between the result of the frequency synthesis (hereinafter, the result of the temporary signature frequency synthesis) using the coefficient for which the temporary signature information is set (hereinafter, the coefficient with the temporary signature information). The sum of the logarithmic values is defined as a first change amount, and regarding the frequency synthesis result in a mixed state of the coefficient with the true sign information and the coefficient with the temporary sign information (hereinafter, the mixed frequency synthesis result), the waveform of the waveform of the mixed frequency synthesis result And the sum of the absolute values of the differences between the waveform and the central axis of the waveform in the region surrounded by the waveform is defined as the first similarity. , The sum of the absolute values of the differences between the waveform and the value range is regarded as the degree of difference, and the first sign information estimation parameter calculating means (711) calculates the first change amount or the first similarity as the first sign information estimation parameter. And the second
The sign information estimation parameter calculating means (714) calculates the degree of difference as a second sign information estimation parameter, normalizes the first amount of change or the degree of difference by a normalization coefficient, and obtains the normalized first amount of change. A value obtained by adding the normalized dissimilarity is defined as a second variation, and the first similarity or the dissimilarity is normalized by multiplying by a normalization coefficient, and the normalized dissimilarity is calculated as the normalized first similarity. The subtracted value is used as the second similarity, and the third signature information estimation parameter calculation means (716)
In (3), the second change amount or the second similarity is calculated as a third signature information estimation parameter, and the third signature information generation means (719) calculates the third signature information estimation parameter corresponding to each temporary signature information into the second signature information estimation parameter. In the case of the change amount, a combination of temporary sign information having a small value of the third sign information estimation parameter is preferentially set as a code including a large number of the same symbols (hereinafter, referred to as a first code), or
A combination of temporary signature information with a small value of the signature information estimation parameter is preferentially set to a short code (hereinafter, a second code), and a first code corresponding to a combination of temporary signature information that matches a combination of true signature information; Alternatively, the second code is output as new signature information, and the third signature information estimation parameter corresponding to each provisional signature information is the second signature information.
In the case of the similarity, a combination of temporary sign information having a large value of the third sign information estimation parameter is preferentially set as a code including a large number of identical symbols (hereinafter, a first code),
Alternatively, a combination of temporary signature information having a large value of the third signature information estimation parameter is preferentially set to a short code (hereinafter, a second code), and corresponds to a combination of temporary signature information that matches a combination of true signature information. The signal processing system according to claim 19, wherein the first code or the second code is configured to be output as new signature information. 23. Inside the sign information estimating means (706), there is provided an inverse quantizing means (720) for performing inverse quantization on the coefficient B or the coefficient with sign information, and at least the sign of the inversely quantized sign is provided. 23. The signal processing system according to claim 20, wherein the information-added coefficient is input to a frequency component synthesizing unit (709) to synthesize each frequency component. . 24. The frequency component synthesizing means (709) calculates with high bit precision.
Or claim 21, or claim 22, or claim 2
4. The signal processing system according to 3. 25. The coefficient encoding means (5), wherein the inputted coefficients are classified into a plurality of groups, and statistical processing and encoding processing are performed for each group. Claim 2, claim 3, or claim 4, or claim 5, or claim 6, or claim 9, or claim 10, or claim 11, or claim 12, or claim 13, or claim 3, or claim 14, or claim 15, or claim 16, or claim 17, or claim 18, or claim 19, or claim 2
25. The signal processing system according to claim 21, or claim 21, or 22, or 23, or 24. 26. The signature information encoding means (703) classifies each input coefficient into a plurality of groups, and performs statistical processing and encoding processing for each group. Claim 4, or Claim 5, or Claim 9, or Claim 11, or Claim 12, or Claim 14, or Claim 15, or Claim 18,
Or claim 19, or claim 20, or claim 2
25. The signal processing system according to claim 22, or claim 23, claim 23, or claim 24. 27. Quantizing means (1) which quantizes coefficient A output from frequency decomposition means (9) and outputs the result as coefficient B.
The signal processing system according to claim 6, characterized in that the coefficient B is stored in the frequency component storage means (2). 28. The basic encoding means (1),
Or claim 2, or claim 3, or claim 4, or claim 5, or claim 9, or claim 10, or claim 11, or claim 12, or claim 1
3, or claim 14, or claim 15, or claim 16, or claim 17, or claim 18, or claim 19, or claim 20, or claim 21, or claim 22, or claim 23, or claim 2
The signal processing system according to claim 6, wherein the signal processing system has the same configuration as that according to claim 4, 25, 26, or 27.