JPH0630392A - Picture coding device - Google Patents

Abstract

PURPOSE:To provide the picture coding device in which optimum coding for an entire picture is attained. CONSTITUTION:A luminance signal and a color difference signal are inputted to a field memory 11 or 12. A block division.circuit 14 applies block processing to an inputted signal and outputs the result to a DCT circuit 15. A control circuit 13 controls an address of the field memory 11 or 12 to allow the block division circuit 14 to output prescribed block data in a pattern. The DCT circuit 15 applies DCT processing to picture data inputted and outputs the result to a quantization circuit 16, in which a DC term and an AC term are quantized. A zigzag scan circuit 17 reads a quantization output in the order of zigzag scanning and inputs the result to a coding circuit 18. The coding circuit 18 uses a Huffman table 19 for the DC term to apply Huffman coding and to output the result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding device relating to a compression encoding system for still images or moving images such as videophones, image transmission devices, electronic still cameras and the like.

[0002]

2. Description of the Related Art In recent years, with the widespread use of videophones and electronic still cameras, there has been remarkable progress in digital processing of images, particularly in high-efficiency coding techniques for compressing image data.

When an image is compression-coded, generally, the code amount tends to increase for a fine pattern, and the code amount tends to decrease for a monotonous pattern. For this reason, the transmission code amount changes for each image, so that real-time moving images cannot be transmitted on a transmission path having a constant speed, or the capacity of a recording medium is insufficient, resulting in a lack of images.

Therefore, there is also one in which a buffer memory is provided in the output section so that the code of the image is constant regardless of the pattern.
FIG. 9 shows a block diagram of such a conventional image coding apparatus.

The image data read from the frame memory 1 is input to a DCT (discrete cosine transform) circuit 2. The DCT coefficient DCT processed by the DCT circuit 2 is
Input to the quantization circuit 3. After being quantized in the quantizing circuit 3, the tunable coding circuit 4 performs Huffman coding, for example. The Huffman code from the variable length coding circuit 4 is output via the buffer memory 5. Quantization circuit 3
Changes the quantization characteristic according to the usage state of the buffer memory 5 and controls the output rate constant.

However, in the above-mentioned conventional image coding apparatus, the entire screen cannot be adjusted because the coding characteristics are successively changed according to the usage state of the buffer memory. For example, if image data with a monotonous pattern in the upper half of the screen and fine patterns in the lower half is input, the number of bits used for encoding the image data in the upper half of the screen is too large, so the image in the lower half is used. There is a problem that a sufficient number of bits cannot be assigned for data encoding, resulting in deterioration of image quality.

[0007]

As described above, in the conventional image coding apparatus, the output rate is made constant by successively changing the coding characteristics according to the usage amount of the output buffer, and the output rate is constant over the entire screen. However, there is a drawback that the image quality is deteriorated because the optimum encoding cannot be performed.

Therefore, the present invention eliminates the above drawbacks,
(EN) It is possible to provide an image coding apparatus that can determine the coding characteristic by previously obtaining the code amount according to the coding characteristic used in the immediately preceding screen and enable the optimum coding of the entire screen. To aim.

[0009]

In order to achieve the above-mentioned object, according to the present invention, a digital image signal is coded by a coding means with a predetermined coding characteristic, and when the code amount is estimated, the digital image signal is recorded. A part of the image signal is given to the encoding means to divide the image, and at the time of estimating the code amount, it includes an image encoding characteristic obtained by performing encoding immediately before the code amount obtained by the encoding means. While estimating the code amount of each encoded output obtained by multiple times of encoding,
The optimum coding characteristic is determined from the estimated code amount and the target code amount.

[0010]

The coding characteristic determining means configured as described above is
By performing the encoding with the encoding characteristic immediately before the encoding, which is necessary before the original encoding, the time required for the encoding can be shortened. This makes it possible to perform optimum encoding on the entire screen.

[0011]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention. The luminance signal and the color difference signal are input to the field memory 11 or the field memory 12. The field memory 11 or the field memory 12 is controlled by the control circuit 13 for writing and reading.
Next, the field memory 11 or the field memory 1
The signal from 2 enters the block division circuit 14.

FIG. 2 is a diagram for explaining the output of image data from the block division circuit 14. Block division circuit 14
Then, the input signal is input horizontally 8 as shown in FIG.
Pixels and 8 vertical pixels are divided into blocks in units of 64 pixels and D
Output to the CT circuit 15. The control circuit 13 controls the address of the field memory 11 or 12 to
Predetermined block data in the screen is divided into block dividing circuits 14
Output from.

Thus, for example, block data of every four blocks in each row (hereinafter referred to as a block line) in the horizontal direction and every four block lines in the vertical direction can be sequentially output from the block division circuit 14. The DCT circuit 15 converts the input image data into DC
T processing is performed, the frequency components are converted, and 64 coefficients are output for each block.

FIG. 3 shows the DCT of the output from the DCT circuit 15.
It is a figure for explaining a coefficient. The DCT coefficients are sequentially arranged from the low-frequency component in the horizontal and vertical directions to the high-frequency component.
In FIG. 3, the portion of 0 indicates a DC term, and the value is an average value of all pixels. The other part is the AC term.

The DCT coefficient from the DCT circuit 15 is input to the quantization circuit 16. In the quantization circuit 16, the DC term,
And quantizing the AC term. The resulting quantized output is input to the zigzag scan circuit 17.

The quantized output in the zigzag scan circuit 17 is read in the zigzag scan order of 0, 1, 3, ... In FIG. In the encoding circuit 18, the DC term is Huffman-encoded using the Huffman table 19. For the AC term, the number of consecutive 0s in the quantized output (0 run) and D immediately after that.
Create a combination with the number of bits of the CT coefficient. The Huffman table 19 is used to perform Huffman coding on this combination and output it.

FIG. 4 shows an example of the Huffman code.
For example, after two consecutive 0s as quantized output,
Is input, the encoding circuit 18 outputs 1101 shown in FIG.
The Huffman code of 1 is output.

As described above, the code length of one block and the code length of one screen are changed by Huffman coding. On the other hand, the code length also changes with quantization. The quantizing circuit 16 quantizes based on the quantized value from the multiplier 22. For example, when the quantized value is increased, the probability that 0 appears as a quantized output increases, and the code length (total number of bits) is increased. ) Becomes shorter.

In a general image, the code length sharply decreases as the quantization value increases. The multiplier 22 adds α to the basic quantization value stored in the basic quantization table 24.
The quantized value given to the quantized circuit 16 is changed by multiplying the quantized table scale factor α inputted from the calculation circuit 21 or the outside. However, if the quantized value is changed, the total number of bits will not be fixed.

In this embodiment, prior to encoding, the quantization table scale factor α used until then is used.
In, the predetermined block data is encoded, the code amount at that time is obtained, and the quantized value for the actual encoding is obtained from this value.

The scale factor α given to the multiplier 22 before encoding is the value α p used up to that point by the switching circuit 23, or α1 to n given from the outside. The code amount accumulating circuit 20 calculates a scale factor αt for actual encoding from the obtained code amount.

The total number of bits of the target encoded output is NT
, The predetermined block data is encoded with α p, the total number of bits of the encoded output estimated by the code amount at that time is NP, the allowable error is ε bit, and NP is

[0024]

[Equation 1] , The predetermined block data is encoded again with a scale factor α1 such that α1 <αP, and the total number of bits of the encoded output estimated from the code amount at that time is N.
Set to 1, and obtain αT from equation (1).

[0025]

[Equation 2] If so, the predetermined block data is encoded with a scale factor α2 such that α2> αP, and the total number of bits of the encoded output estimated from the code amount at that time is set to N2, and αT is calculated by the following equation (2). .

[0026]

[Equation 3] In this way, the code amount accumulating circuit 20 calculates the scale factor αT.

FIG. 5 is for explaining the block data to be encoded at the time of estimating the code amount (prescan) for obtaining the scale factor α T, FIG. 6 is for explaining the quantized value, and FIG. 7 is for explaining the operation sequence.

The input luminance and color difference signals are alternately input and recorded in the field memories 11 and 12 for each screen. After the signal of the Nth field is written in the field memory 11, the blanking time of the input signal is used to output data of 1/4 block in both the horizontal and vertical directions in the field memory 11 as shown by the hatched portion in FIG. Read out (first prescan), and the block division circuit 14
The DCT coefficient is given to the zigzag scan circuit 17 in block units.

The zigzag scan circuit 17 gives the DCT coefficient to the quantization circuit 16 in order from the low frequency component to the high frequency component. The quantization circuit 16 performs quantization based on the scale factor α. The switching circuit 23 first supplies the multiplier 22 with the scale factor αP used in the immediately preceding encoding.
The multiplier 22 multiplies the basic quantization table 24 and α P to create a quantized value and supplies it to the quantization circuit 16. In the basic quantization table 24, for example, as shown in FIG. 6, each quantized value corresponds to each frequency component of the quantized output shown in FIG.

The data quantized by the quantizing circuit 16 is Huffman coded by the coding circuit 18, and then the code amount accumulating circuit 20.
Calculate the code amount with. This code amount is multiplied by 16 to estimate the code amount of all blocks. In a natural image, the correlation between the images is high, and the code amount when the code amount is estimated by using the partial images as in this embodiment and the code amount when the entire 1/16 block is encoded are The code amount obtained by multiplying 16 times is within an error of 10% although it depends on the sampling position of the block. The code amount NP thus obtained and the target code amount NT are constant values ε.

[0031]

[Equation 4] If so, the scale factor is αP. After the prescan is completed, all the image signals of the Nth frame are read again from the field memory 11 and are quantized and encoded. Also,

[0032]

[Equation 5] In that case, the horizontal and vertical 1/4 block data is read again from the field memory 11 (second prescan), quantized by the quantization circuit 16 with the scale factor α1, and Huffman coded by the coding circuit 18. After that, the code amount accumulating circuit 20 obtains the code amount.

This code amount is multiplied by 16 to estimate the code amount of all blocks. With the code amount N1 thus obtained, the scale factor αT at the time of encoding is obtained by the equation (1),
All the image signals of the Nth field are read again from the field memory 11 and are quantized and encoded.

[0034]

[Equation 6] The same is true for.

On the other hand, the input signal of the (N + 1) th field is written in the field memory 12 of the (N + 1) th field to obtain the scale factor for the image signal of the (N + 1) th field. In this case, the scale factor used in the Nth field is used first. After the scale factor α T is obtained, the image signal of the (N + 1) th field is encoded in the (N + 2) th field period. In this embodiment, the scale factor is α during the second prescan.
Although it was set to 1 or α2, in order to estimate a more accurate scale factor,

[0036]

[Equation 7] As shown in FIG.
Change the scale factor during prescan.

Here, the number of effective pixels in one field of the luminance signal is 768 × 240 pixels, and the writing / reading clock of the field memory is 4 fsc (fsc: NTS).
C signal subcarrier frequency = 3.5795 ... MH
z), the read time from the field memory is about 12.9 ms.

On the other hand, one field period is about 16.6 ms.
Therefore, 3.7 ms can be used for estimating the code amount. Since it takes about 0.8 ms to read a 1/16 image from the field memory, the prescan for estimating the code amount can be performed up to 4 times. In the above example, once if all fields are the same, 2 if they are different.
The pre-scan was performed once to determine the scale factor for encoding. When the pre-scan can be performed up to four times, under the conditions shown in FIG. 8, for example, the scale factor at the time of encoding is determined by the following equation (3).

[0039]

[Equation 8]

By using the scale factor thus obtained, it is possible to control the code amount more accurately. Furthermore, the number of prescans may be fixed,
For example, the allowable error from the target code amount is defined as δ, and δ is

[0041]

[Equation 9] If, the prescan may be stopped and the scale factor may be calculated.

In the above description, the system uses the field memory, but instead of this, the same applies to the case where the frame memory is used and the interlaced field signal is framed and then frame-encoded. Is.

As described above, in this embodiment, since only 1/16 block data of the entire screen is read at the time of estimating the code amount, the scale factor can be obtained in the blanking period, and the code of the moving image can be obtained. Is possible.

[0044]

As described above, according to the present invention, the code amount is estimated by using only a part of the block of the entire screen before the image signal is encoded, and the blanking of the input signal is performed. Since the quantization characteristic can be determined in time, the moving image can be encoded in real time with high image quality.

[Brief description of drawings]

FIG. 1 is a block diagram of an image coding apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a pixel array of an input image.

FIG. 3 is a diagram for explaining the position of each coefficient after orthogonal transform.

FIG. 4 is a diagram showing an example of a Huffman code.

FIG. 5 is a diagram for explaining how to take blocks used for code amount estimation.

FIG. 6 is a diagram showing an example of a basic quantization table.

FIG. 7 is a diagram for explaining an encoding operation sequence.

FIG. 8 is a diagram illustrating the relationship between prescan and estimated code amount.

FIG. 9 is a diagram showing an example of a conventional image encoding device.

[Explanation of symbols]

11, 12 ... Field memory, 13 ... Control circuit, 14
... Block division circuit, 15 ... DCT circuit, 16 ... Quantization circuit, 17 ... Zigzag scan circuit, 18 ... Encoding circuit, 19 ... Huffman table, 20 ... Code amount accumulating circuit,
21 ... α calculation circuit, 24 ... Basic quantization table.

Claims (5)

[Claims] 1. A coding means for coding a digital image signal with a predetermined coding characteristic, and a part of the digital image signal when estimating the code amount,
Means for dividing the image, which is given to the encoding means, and a plurality of times of encoding including an image encoding characteristic in which the encoding is performed immediately before the encoding amount obtained by the encoding means at the time of estimating the encoding amount. An image coding apparatus, comprising: a means for estimating a code amount of each of the obtained coded outputs, and for determining an optimum coding characteristic from the estimated code amount and a target code amount. 2. A unit for obtaining a transform coefficient by orthogonally transforming block data obtained by dividing a digital image signal into blocks for each of a plurality of pixels, and a transform coefficient obtained by the unit, which is input, and a predetermined quantization characteristic. Quantizing means for quantizing and outputting by the quantizing means, first coding means for variable-length coding the transform coefficient by the quantizing means, and obtaining a transform coefficient for a part of the block data at the time of code amount estimation. Means for dividing the image given to the quantizing means, estimating the total code amount from the code amount of the variable-length coded output, and changing the quantization characteristic from the estimated code amount and the target code amount to obtain a predetermined value. A second encoding means for performing the encoding once, and a predetermined number of encodings including the image encoding characteristic of the encoding amount immediately before the code amount obtained by the second encoding means. For each encoded output It estimates the total code amount from issue amount, the image encoding device characterized by comprising a means for determining the optimal coding properties from the estimated code amount and the target to code amount. 3. A unit for obtaining a transform coefficient by orthogonally transforming block data obtained by dividing a digital image signal into a plurality of pixels into blocks, and a transform coefficient obtained by the unit is inputted and a predetermined quantization characteristic is inputted. Quantizing means for quantizing and outputting by the quantizing means, coding means for variable-length coding the transform coefficient by the quantizing means, and an image coded immediately before the code amount obtained by the coding means. A code amount estimating means for estimating a total code amount from a code amount of a variable length coded output obtained for a part of the block data according to the encoding characteristic, and a code amount estimated by the code amount estimating means and a target. An image coding apparatus, comprising: means for determining the coding characteristic of an image coded immediately before as the optimum coding characteristic if the difference from the code amount is within an allowable value. 4. A unit for obtaining a transform coefficient by orthogonally transforming block data obtained by dividing a digital image signal into blocks for each of a plurality of pixels, and a transform coefficient obtained by the unit is input and a predetermined quantization characteristic is input. Quantizing means for quantizing and outputting by the quantizing means, coding means for variable-length coding the transform coefficient by the quantizing means, and an image coded immediately before the code amount obtained by the coding means. A code amount estimating means for estimating a total code amount from a code amount of a variable length coded output obtained for a part of the block data according to the encoding characteristic, and a code amount estimated by the code amount estimating means and a target. If the difference from the code amount is equal to or more than the allowable value, a unit for changing the encoding characteristic again, the total code amount estimated by the code amount estimating unit, and the encoding characteristic of the image encoded immediately before Mark Image encoding apparatus characterized by comprising a means for determining the optimal coding property from the encoded characteristic when changing the properties. 5. A unit for obtaining a transform coefficient by orthogonally transforming block data obtained by dividing a digital image signal into a plurality of pixels into blocks, and a transform coefficient obtained by the unit is input and a predetermined quantization characteristic is input. Quantizing means for quantizing and outputting by the quantizing means, coding means for variable-length coding the transform coefficient by the quantizing means, and coding immediately before the code quantity obtained by the coding means at the time of code quantity estimation Means for estimating the total code amount from the code amount of the variable-length coded output obtained for a part of the block data by the image coding characteristic, and the code amount estimated by the means and the target code amount. Means for changing the amount of change in the coding characteristic, and a code of the variable-length coded output when the variable-length coding is performed again for a part of the block data of the block data. Means for estimating the total code amount, and means for determining the optimum coding characteristic from the coding characteristic of the image coded immediately before and the coding characteristic when the coding characteristic is changed. An image encoding device characterized by the above.