JPH0884344A - Image signal coding method and image signal coder - Google Patents

Abstract

PURPOSE: To improve considerably coding efficiency when a characteristic point is detected from an input image signal, the signal is coded, a difference between the decoded image signal reconfigured from the coded characteristic point signal and an original image signal is taken and the difference signal is coded separately from the characteristic point in the image signal coding method and the image signal coder. CONSTITUTION: A parameter used to evaluate the effect of a difference K11 given onto the image quality is obtained and the importance K21 of the difference signal K11 is calculated from the parameter to decide an important area and an unimportant area and the coding is controlled matching the importance K21 of the difference signal K11 to evaluate the area important of the coding of the difference signal K11 and area not importance and much more information quantity is allocated to the more important area according to the evaluation.

Description

Detailed Description of the Invention

[0001]

[Table of Contents] The present invention will be described in the following order. Industrial Application Conventional Technology (FIGS. 6 to 13) Problems to be Solved by the Invention Means for Solving the Problems (FIG. 1) Action (FIG. 1) Embodiments (FIGS. 1 to 5) Effects of the Invention

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal encoding method and an image signal encoding apparatus, and more particularly, to a remote location of an image by a transmission medium having a limited transmission capacity when transmitting or recording and reproducing the image signal. An image transmission device or a tape recorder / disc that uses a high-efficiency encoding method for efficiently compressing an image signal with a smaller amount of information, which is a device for transmission and recording / reproduction on a storage medium having a limited storage capacity. It is suitable to be applied to a recorder or the like.

[0003]

2. Description of the Related Art Conventionally, for coding an image signal, a coding method such as predictive coding for handling an image in pixel units, orthogonal transform coding such as discrete cosine transform (DCT), subband coding, There are band-division coding represented by the wavelet transform and block-coding methods in which the band-divided data is divided into blocks and handled. Although predictive coding is easy to implement, deterioration is easily perceived at high compression rates. Further, in a block coding method such as a subband code by DCT or block division or a wavelet transform, a relatively high image quality can be obtained even at a high compression rate by performing appropriate quantization for each block. However, when the compression rate is further increased, block distortion begins,
The visually unfavorable effects will become noticeable.

Therefore, as a coding method capable of reducing distortion that causes visual interference even at a high compression rate, image characteristic point detection is performed by extracting characteristic points of the image structure and efficiently encoding the points. The structure extraction coding method according to FIG. 6 shows an image signal coding apparatus for detecting feature points of an image and coding the difference image thereof. In this image signal encoding device,
The input image signal A11 becomes a filter image signal A12 by the preprocessing filter circuit 101. The pre-processing filter circuit 101 performs pre-processing for subsequent feature point detection processing such as smoothing filtering for noise removal.

The filter image signal A12 is input to the two-dimensional feature point detection circuit 102. The two-dimensional feature point detection circuit 102 detects feature points, traces a series of feature points in an appropriate order, extracts them as a continuous chain, and outputs a chain signal A13. Chain signal A
13 is a chain coding circuit 103 which performs chain coding in the direction traced by the feature point chain and quantizes and codes the amplitude value around the feature point chain to obtain a coded feature point signal. It becomes A14. The structure of the chain signal at this time is shown in FIG. 8, and the structure of the chain coding is shown in FIG.

Next, the encoded feature point signal A14 is internally decoded by the local decoder. First, the encoded feature point signal A1 input to the internal chain decoding circuit 106.
4 is an internal reconstructed feature point signal A17 which is chain-decoded and is the amplitude value information around the position of the feature point. The internally reconstructed feature point signal A17 is reconstructed into an internally reconstructed feature point image signal A18 by the internal feature point image reconstruction circuit 107. The internally reconstructed feature point image signal A18 is the same signal as the reconstructed feature point image that will be reconstructed from only the feature point information on the image signal decoding device side.

Here, a difference image signal A19 is obtained by taking the difference between the filter image signal A12 and the internally reconstructed feature point image signal A18. This difference image signal A19 is encoded by the difference image encoding circuit 108 and becomes an encoded difference image signal A20. 10 (A) and FIG.
It shows in (B) and FIG. 10 (C). Multiplexing circuit (MUX)
At 104, the coded feature point signal A14 and the coded differential image signal A20 are multiplexed, and a multiplexed signal A15 is output.
The multiplexed signal A15 has its information amount smoothed by the buffer memory 105 and becomes the output signal A16 of the image signal encoding device.

FIG. 11 shows a differential image coding circuit for coding a differential image signal. This differential image coding circuit can be replaced with the differential image coding circuit 108 in FIG. 6, and here the differential image signal is block-coded and then transform-coded by DCT. An input differential image signal C11 corresponding to the differential image signal A19 in FIG.
Is a block differential image signal C in the block circuit 131.
Converted to 12. This block formation is, for example, as shown in FIG. 13, dividing the entire screen into 8 × 8 square blocks and rearranging them in an appropriate order. The blocked differential image signal C12 is converted by the DCT conversion circuit (DCT) 132 to become a converted differential image signal C13.

On the other hand, the quantization control circuit 136 refers to the amount of information C17 stored in the buffer memory 135 arranged in the subsequent stage, calculates the quantization step size C18, and inputs it to the quantization circuit (Q) 133. The converted difference image signal C13 is
In the quantizing circuit 133, the quantizing step size C18
Are quantized to form a quantized difference image signal C14. The quantized difference image signal C14 is a variable length coding circuit (VLC).
The variable length coding is performed by inputting to H.134 and the Huffman code and the run length code of zero coefficient are combined, and the variable length coded difference image signal C15 is obtained.

The buffer memory 135 smoothes the information amount of the variable length coded differential image signal C15, outputs the smoothed coded differential image signal C16, and outputs the accumulated information amount C17 of the buffer memory 135 to the quantization control circuit 136. The smoothed encoded difference image signal C16 is a multiplexing circuit (MUX).
It is input to 137, is multiplexed with the quantization step size C18, and becomes the output C19 of the differential image coding circuit. This output signal corresponds to the encoded difference image signal A20 in FIG.

Next, FIG. 7 shows an image signal decoding apparatus for detecting feature points of an image and encoding the difference image thereof. In this image signal decoding device, after the input signal B11 is stored in the buffer memory 121, the shunt circuit (DMU)
X) 122 separates the encoded feature point signal B13 and the encoded difference image signal B16. Chain decoding circuit 1
23 performs chain decoding from the coded feature point signal B13, reproduces the position of a series of feature points and amplitude value information around it, and outputs a restored feature point signal B14. The restored feature point signal B14 is reconstructed into a restored feature point image signal B15 by the feature point image restoration circuit 124.

The differential image decoding circuit 125 also decodes the coded differential image signal B16 and restores the differential image signal B1.
7 is output. Here, the restored feature point image signal B15 and the restored difference image signal B17 are added to reconstruct the final restored image signal B18. The restored image signal B18 becomes an output signal B19 of the image signal decoding device after being subjected to appropriate post-filter processing by the post-processing filter circuit 126. The restored difference image signal B17 and the restored image signal B18 at this time are made to correspond to the one-dimensional signals, respectively, as shown in FIG.
(D) and FIG. 10 (E).

Further, FIG. 12 shows a differential image decoding circuit for decoding the differential image signal. This difference image decoding circuit can be replaced with the difference image decoding circuit 125 of FIG. 7, and is a decoding circuit corresponding to the difference image encoding circuit shown in FIG. The encoded difference image signal B16 of FIG.
The input signal D11 corresponding to is the shunt circuit (DMUX) 1
At 41, the smoothed coded difference image signal D12 and the quantization step size D15 are separated. The smoothed coded difference image signal D12 is immediately input to the buffer memory 142 to become the variable length coded difference image signal D13, which is further input to the variable length decoding circuit (IVLC) 143, which is variable length decoded and quantized. The difference image signal D14 is obtained.

The quantized difference image signal D14 is input to the inverse quantization circuit (IQ) 144 and is inversely quantized by the quantization step size D15 separated by the shunt circuit 141,
The restored conversion difference image signal D16 is obtained. The restored conversion difference image signal D16 is inversely transformed by the inverse DCT conversion circuit (IDCT) 145 to be the restored block difference image signal D17.
Becomes The restored block differential image signal D17 is converted into a restored differential image signal D18 by the deblocking circuit 146.

[0015]

By the way, the feature points of the image are detected and coded as described above, the difference between the restored image reconstructed from only the feature points and the original image is calculated, and the DCT is applied to the difference signal. In the image signal coding method for performing the block coding as described above, the amount of information stored in the buffer memory and the strength of the input signal used in the conventional DCT coding method as described above are used in the coding block. The algorithm that determines the coding parameters cannot judge a significant difference signal, so the difference signal in the important area is cut off, or the difference signal in the unimportant area is coded to occupy a wasteful amount of information. There is a drawback that it ends up.

The present invention has been made in consideration of the above points, and detects a characteristic point from an input image signal and encodes it, and a reconstructed image signal and an original image signal reconstructed from the encoded characteristic point signal. It is intended to propose an image signal coding method capable of significantly improving the coding efficiency when the difference signal is taken and the difference signal is coded separately from the feature points.

[0017]

In order to solve the above problems, the present invention detects a feature point from an input image signal and encodes it, and a reconstructed image signal reconstructed from the encoded feature point signal. In the image signal coding method, in which the difference between the original image signal and the original image signal is obtained and the difference signal is encoded separately from the feature points, the image is divided into a plurality of small areas, and the importance of the difference signal (K11) with respect to the restored image signal (K21)
The encoding process of the difference signal (K11) is controlled by using the importance evaluation step that is digitized for each small region and the importance (K21) of the difference signal (K11) that is digitized by the importance evaluation step. An encoding control step is provided.

Further, in the present invention, feature points are detected from the input image signal and coded, the difference between the reconstructed image signal reconstructed from the coded feature point signal and the original image signal is calculated, and the difference is calculated. In an image signal encoding device that encodes a signal separately from feature points, an image is divided into a plurality of small areas, and the importance (K21) of the difference signal (K11) with respect to the restored image signal is digitized for each small area. Importance evaluation means (11, 16, 17) and its importance evaluation means (11,
An encoding control means (12, 13, 18) for controlling the encoding process of the difference signal (K11) is provided by using the importance (K21) of the difference signal (K11) digitized in (16, 17). I did it.

[0019]

The parameter for evaluating the influence of the difference signal (K11) on the image quality is obtained, and the difference signal (K
By calculating the importance (K21) of 11) to determine the important region and the non-important region and controlling the encoding according to the importance (K21) of the difference signal (K11), the difference It is possible to evaluate areas where the coding of the signal (K11) is important and areas where the coding is not important, and according to the evaluation, it is possible to allocate a larger amount of information to the more important areas, thus improving the coding efficiency, It is possible to obtain a restored image of higher quality.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 shows a differential image coding circuit applied to the image signal coding method and the image signal coding apparatus according to the present invention. This difference image encoding circuit is the difference image encoding circuit 108 of the image signal encoding device described above with reference to FIG.
, Which corresponds to the differential image encoding circuit in FIG. The input differential image signal K11 corresponding to the differential image signal A19 in FIG. 6 is converted into the blocked differential image signal K12 by the block circuit 11. For example, this block division divides the entire screen into 8 × 8 square blocks,
This is a process of rearranging in an appropriate order. The block difference image signal K12 is converted into a converted difference image signal K13 by the DCT conversion circuit (DCT) 12.

The quantization control circuit 18 refers to the block difference evaluation value K21 of the differential signal evaluation circuit 17 for calculating the importance of the differential signal, which will be described later, and the accumulated information amount K17 of the buffer memory 15 arranged in the subsequent stage. Then, the quantization step size K18 is calculated and input to the quantization circuit (Q) 13. The transformed difference image signal K13 is quantized in the quantization circuit 13 with the quantization step size K18 to become a quantized difference image signal K14. Quantized difference image signal K14
Is input to the variable length coding circuit (VLC) 14 and is subjected to variable length coding in which, for example, a Huffman code and a run length code of zero coefficient are combined to form a variable length coded difference image signal K15.

The buffer memory 15 smoothes the information amount of the variable length coded differential image signal K15, outputs the smoothed coded differential image signal K16, and outputs the accumulated information amount K17 of the buffer memory 15 to the quantization control circuit 18. To do. The smoothed coded difference image signal K16 is a multiplexing circuit (MUX).
It is input to 19 and is multiplexed with the quantization step size K18 to become the output K22 of the differential image coding circuit. This output signal corresponds to the encoded difference image signal A20 in FIG.

On the other hand, the internally reconstructed feature point signal A1 in FIG.
The input edge image signal K19 corresponding to 7 is converted into a block edge image signal K20 by the block circuit 16. This block formation is similar to the processing for the differential image signal K11. The differential signal evaluation circuit 17 calculates an evaluation value indicating the importance of the differential signal from the block differential image signal K12 and the block edge image signal K20 for each area of the corresponding 8 × 8 square block and calculates the block difference. The evaluation value K21 is output.

A first embodiment of the differential signal evaluation circuit 17 is shown in FIG. Here, as shown in FIG. 2A, processing is performed in units of 8 × 8 square blocks. FIG.
The differential image signal L11 corresponding to the blocked differential image signal K12 of is a square signal L12 by the multiplier 31,
The cumulative adder 32 outputs the block difference square sum signal L13.
Becomes Further, the differential image signal L14 corresponding to the blocked edge image signal K20 in FIG. 1 is binary data representing whether a pixel is an edge or not is represented by "1" or "0", and is thus added by the cumulative adder 33, The number of edge points existing in the block is counted and becomes a block edge density signal L15.

The block difference sum of squares signal L13 and the block edge density signal L15 are used for the comparator 3 composed of a ROM.
4 and outputs the difference evaluation value L16 of the block. This output signal corresponds to the block difference evaluation value K21 in FIG. FIG. 3 shows a first configuration showing the comparison standard of the comparator 34 for evaluating the importance of this block difference. For example, the combination of the two parameters of the sum of squared differences and the edge density is the area A shown in the graph of FIG.
If the difference signal is included in the region B, the importance of the difference signal is high, and if it is included in the region B, the importance is determined to be low. As the ROM configuration, the sum of squared differences is assigned to the lower address and the edge density is assigned to the upper address and input to the ROM, and the data "1" or "0" of the corresponding address is used as the numerical value of the importance.

The graph of FIG. 3B shows the correspondence of the weighting coefficient for the differential signal coding according to the importance evaluation value.
When the importance level is area A, the weighting coefficient for encoding is "1", and when the importance level is area B, the weighting coefficient for encoding is "0". In the quantization control circuit 18 of FIG. 1, when the weighting coefficient is “1”, the signal is coded as it is, but when the weighting coefficient is “0”, the signal in the corresponding small area is not coded. Quantization control is performed.

Next, FIG. 4 shows a second configuration showing the comparison standard of the comparator 34 for evaluating the importance of the block difference. For example, if the combination of the two parameters of the sum of squared differences and the edge density is included in the area A shown in the graph of FIG. 4A, the difference signal has high importance and is included in the area B. If it is included in the region C, the importance is moderate, if it is included in the region D, the importance is relatively low, and if it is included in the region E, the importance is almost zero. To judge.

Further, the graph of FIG. 4B shows the correspondence of the weighting coefficient of the differential signal coding according to the importance evaluation value.
The weighting coefficient for coding is "1" when the importance level is in the area A, and the weighting coefficient for coding is "0.75" when the importance level is in the area B. Also, the weighting coefficient becomes small. In the quantization control circuit 18 of FIG. 1, the weighting coefficient is used to divide the quantization step size in the corresponding small area, and in the case of a small weighting coefficient, the DCT-transformed coefficient is quantized with a larger step size. The amount of generated information is reduced by performing various controls.

Next, a second embodiment of the differential signal evaluation circuit is shown in FIG.
Shown in The difference image signal M11 corresponding to the block difference image signal K12 in FIG. 1 becomes the absolute value signal M12 by the absolute value calculation circuit 41, and becomes the block difference absolute value sum signal M13 by the cumulative adder 42. Also, a differential image signal M14 corresponding to the block edge image signal K20 in FIG.
Is binary data that represents whether a pixel is an edge or not by "1" or "0", and therefore is added by the cumulative adder 43, and the number of edge points existing in the block is counted to obtain the block edge. It becomes the density signal M15.

The block difference absolute value sum signal M13 and the block edge density signal M15 are input to a comparator 44 composed of a ROM, and a difference evaluation value M16 of the block is output. This output signal corresponds to the block difference evaluation value K21 in FIG. Also in the case of the differential signal evaluation circuit of the second embodiment, the configuration of the comparator 44 for evaluating the importance of the block difference is the same as that of the differential signal evaluation circuit of the first embodiment described above in FIG. .

According to the above configuration, the influence of the differential signal on the image quality is evaluated, the region where the coding is important is distinguished from the region where the coding is not important, and the appropriate coding weighting is performed. Based on the amount of information stored in the buffer memory and the strength of the input signal, compared to the algorithm that determines the coding parameters, the difference signal in the important area is saved, and the difference signal in the unimportant area is not coded. The amount can be saved. Further, as compared with the conventional difference signal encoding method, since the information amount can be assigned to the area where the influence of the difference component visually is large, a compressed image with less visual deterioration can be obtained.

In the above embodiment, the input image is divided into 8 × 8 square blocks to make a small area for importance evaluation. However, this small area may be a block having an appropriate arbitrary size. Also, even if the shape is not limited to the square block and the shape of the image is taken into consideration in consideration of the characteristics of the image,
The same effect as that of the above-described embodiment can be realized.

Further, in the above embodiment, the number of feature points is counted as a method for obtaining the strength of the feature point signal in the small area. However, in addition to this, for example, the sum of the variation amounts at the feature points may be used. . Further, as a method of obtaining the intensity of the difference signal in the small area, the sum of absolute values or the sum of squares of the difference signals is calculated. However, for example, a variance value may be calculated to quantify the degree of change. In addition, a two-level or five-level importance evaluation
Although the case of evaluation in stages has been described, the evaluation may be performed in more stages, or may be performed in an arbitrary real number functional relationship.

Further, in the above-mentioned embodiment, the case of using the block-based DCT as the method of encoding the differential signal has been described. For example, Haar transform, vector encoding, predictive encoding not divided into blocks, and wavelet. Band division coding such as conversion or baseband coding may be used. Also, the case where the value determined according to the importance of the difference signal is used as a parameter for changing the quantization step size has been described. However, for example, the signal value in the small area may be multiplied by the value, or D
Weighting may be performed according to the position of the coefficient converted by CT.

[0036]

As described above, according to the present invention, a parameter for evaluating the influence of the differential signal on the image quality is obtained, and the importance of the differential signal is calculated from the value to determine the important region and the non-important region. By determining and, and by controlling the encoding in accordance with the importance of the differential signal, it is possible to evaluate the region where the encoding of the differential signal is important and the region where the encoding of the differential signal is not important, and more importantly according to the evaluation. It is possible to realize an image signal encoding method and an image signal encoding device that can allocate a larger amount of information to different areas, improve the encoding efficiency, and obtain a restored image with higher image quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing a differential image coding circuit to which an image signal coding method and an image signal coding device according to the present invention are applied.

FIG. 2 is a block diagram showing a first embodiment of a differential signal evaluation circuit in the differential image encoding circuit of FIG.

FIG. 3 is a schematic diagram showing a first configuration of a criterion of importance evaluation of a differential signal evaluated by the differential signal evaluation circuit of FIG.

4 is a schematic diagram showing a second configuration of a criterion for evaluating the importance of a differential signal evaluated by the differential signal evaluation circuit of FIG.

5 is a block diagram showing a second embodiment of the differential signal evaluation circuit in the differential image coding circuit of FIG.

FIG. 6 is a block diagram showing an image signal coding apparatus using coding by feature point extraction and coding of a difference signal thereof.

FIG. 7 is a block diagram showing an image signal decoding apparatus using encoding by feature point extraction and encoding of a difference signal thereof.

FIG. 8 is a schematic diagram for explaining an internal representation format of a feature point chain.

FIG. 9 is a schematic diagram for explaining an internal representation format when a feature point chain is chain-encoded.

FIG. 10 is a signal waveform diagram for explaining encoding and decoding of feature points and encoding of a difference signal.

FIG. 11 is a block diagram showing a conventional differential image encoding circuit.

FIG. 12 is a block diagram showing a conventional differential image decoding circuit.

[Fig. 13] Fig. 13 is a schematic diagram for explaining a block processing of differential image coding.

[Explanation of symbols]

11, 16 ... Blocking circuit, 12 ... DCT conversion circuit (DCT), 13 ... Quantization circuit (Q), 14 ... Variable length coding circuit (VLC), 15 ... Buffer memory,
17 ... Differential signal evaluation circuit, 18 ... Quantization control circuit,
19 ... Multiplexing circuit, 31 ... Multiplier, 32, 33, 4
2, 43 ... Cumulative adder, 34 ... Comparator, 41 ... Absolute value calculation circuit.

Claims (10)

[Claims] 1. A feature point is detected from an input image signal and coded, a difference between a restored image signal reconstructed from the coded feature point signal and an original image signal is calculated, and the difference signal is used as the feature. In the image signal encoding method for encoding separately from the points, the image is divided into a plurality of small areas, and the importance of the difference signal with respect to the restored image signal, an importance evaluation step of digitizing each of the small areas, An image signal encoding method, comprising: an encoding control step for controlling the encoding process of the difference signal by using the importance of the difference signal digitized by the importance evaluation step. 2. The importance evaluation step comprises a feature point strength calculation step for obtaining the strength of the feature point signal existing in the small area, and a difference signal strength calculation for obtaining the strength of the difference signal in the small area. 1 of the intensity of the feature point signal and the intensity of the difference signal
The image signal coding method according to claim 1, further comprising: an evaluation value determining unit that determines an evaluation value for each of the small areas from a plurality of calculation results. 3. The image signal coding method according to claim 2, wherein said feature point strength calculation step counts the number of said feature points existing in said small area. 4. The image signal coding method according to claim 2, wherein the difference signal strength calculation step is adapted to obtain a sum of absolute values or a sum of squares of the difference signals in the small region. . 5. The encoding control step uses a level detection step for the importance of the difference signal digitized for each of the small regions and a level for the importance of the detected difference signal as a quantization parameter. The image signal encoding method according to claim 1, further comprising a corresponding operation step. 6. A feature point is detected from an input image signal and coded, a difference between a reconstructed image signal reconstructed from the coded feature point signal and an original image signal is calculated, and the difference signal is used as the feature. In the image signal encoding device for encoding separately from the points, the image is divided into a plurality of small areas, and the importance of the difference signal with respect to the restored image signal, importance evaluation means for digitizing each of the small areas, An image signal encoding apparatus, comprising: an encoding control unit that controls the encoding process of the difference signal by using the importance of the difference signal that is digitized by the importance evaluation unit. 7. The importance level evaluation means calculates a feature point strength calculating means for calculating the strength of the feature point signal existing in the small area, and a difference signal strength calculation for calculating the strength of the difference signal in the small area. 1 of the intensity of the feature point signal and the intensity of the difference signal
7. The image signal coding apparatus according to claim 6, further comprising: an evaluation value determining unit that determines an evaluation value for each of the small areas from a plurality of calculation results. 8. The image signal coding apparatus according to claim 7, wherein the feature point strength calculation means counts the number of the feature points existing in the small area. 9. The image signal coding apparatus according to claim 7, wherein the difference signal strength calculation means is adapted to obtain a sum of absolute values or a sum of squares of the difference signals in the small area. . 10. The encoding control means uses, as a quantization parameter, a level detection means for the importance of the difference signal digitized for each of the small areas, and a level for the importance of the detected difference signal as a quantization parameter. 7. The image signal coding apparatus according to claim 6, further comprising a calculating means for making the correspondence.