JPH06292189A - Method and device for coding image - Google Patents

Abstract

PURPOSE:To provide a method and device to code an image improving coding efficiency and providing satisfactory picture quality. CONSTITUTION:This method and device are provided with memory 11 to store an image signal, a subtractor 12 which takes difference between a predictive value and the next inputted image signal with a stored image signal A as the predictive value, a quantizer 32 which quantizes a differential image signal, a decoder circuit 4 to decode a coded differential coding image signal and to return it to an image signal A' in correspondence to the image signal A stored in the memory 11, a subtractor 22 which calculates difference between a restored image signal A' and the image signal A, a weighting circuit 33 which weights on the difference calculated by the subtractor 22, and an adder 31 which adds weighted difference to the differential image signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding method and apparatus used for compressing, transmitting and recording digital images.

[0002]

2. Description of the Related Art The transfer rate of digital images is from several 100M
It reaches Gbps, and there are restrictions on communication costs during transmission and recording capacity during recording. Therefore, image encoding methods and apparatuses have been developed that minimize image quality deterioration and reduce the transfer rate. The CCIR H.261 standard, which is an example of the above-described conventional image encoding method, that is, the motion-compensated interframe difference two-dimensional DCT will be described below.

FIG. 11 shows a two-dimensional difference between motion compensation frames.
It is a block diagram of DCT. In FIG. 11, an image signal input terminal has a motion detection circuit 107 for calculating a motion vector of an image, and a two-dimensional converter for converting image data into a conversion coefficient.
The DCT circuit 101 is connected, and the motion detection circuit 107 generates a predicted image data.
6, the two-dimensional DCT circuit 101 is connected to a quantizer 102 that quantizes data. Quantizer 102
Is connected to the output terminal 109 and the inverse quantizer 103 for returning the quantized data to the original data, and the inverse quantizer 103 is further connected to the inverse two-dimensional DCT circuit 104 for inverse transforming the data. ing. The inverse two-dimensional DCT circuit 104 is connected to a frame memory 105 that stores the restored data via an adder, and the frame memory 105 is connected to a motion-compensated interframe prediction circuit 106. or,
A subtractor is connected between the motion compensation inter-frame prediction circuit 106 and the image signal input terminal, and the subtractor and the two-dimensional DCT are connected.
Between the circuit 101 and the motion compensation inter-frame prediction circuit 1
A switch is provided between 06 and the adder, and the switch is provided with an intra-frame and inter-frame switching signal input terminal 108 for switching the switch.

The operation of the image coding apparatus configured as described above will be described below. The first frame of encoding, that is, the first frame, is intraframe-encoded by the switching signal input to the switching signal input terminal 108 and indicating the intraframe encoding without taking a difference. That is, the input image data is converted into transform coefficient data by the two-dimensional DCT circuit 101 in a certain block unit, and the transform coefficient data is quantized by the quantizer 103 and sent to the transmission line through the output terminal 109. In general, images have a high correlation, so when DCT is performed, energy concentrates on transform coefficient data corresponding to low frequency components. Therefore, it is possible to minimize the deterioration of the image quality and reduce the amount of data by displaying high frequency components that are visually inconspicuous and finely quantizing the low frequency components that are important components. The transform coefficient data sent to the transmission line is simultaneously sent to the inverse quantizer 103 and the inverse two-dimensional DCT circuit 10.
4 returns to real-time data, and the frame memory 10
Stored in 5.

Next, when the second frame is input, interframe differential encoding is performed. In this case, first, in the motion detection circuit 107, for example, the well-known full search method is used to obtain the inter-frame motion vector in block units.
The motion-compensated inter-frame prediction circuit 106 uses the detected motion vector to generate a motion-compensated prediction value of the previous frame to the next frame in block units (that is, here, from the first frame to the second frame). Calculate the predicted value). Next, the difference between the input image of the second frame and the predicted value generated from the decoded image of the first frame is calculated, and the difference data is encoded by the same method as in the first frame. After the third frame, the prediction value is coded by the same method as the second frame. According to the above method, since the difference from the predicted value is encoded, the energy is reduced as compared with the case where the prediction is not performed, and thus more efficient encoding is possible (for example, CCIT
T Recommendation H.261, "Codec for audiovisual services" at px64 kbit /
s ", Geneva, 1990).

[0006]

However, in the above configuration, since the decoded image of the previous frame is used as the prediction value, the difference signal between the original image and the prediction value becomes the difference information + quantization error and is actually transmitted. I needed to send more information than I should have. FIG. 12 is an example of a one-dimensional DPCM decoded image, where the horizontal axis represents time and the vertical axis represents amplitude. In the figure, the solid line shows the original image and the dotted line shows the decoded image. As the dotted line in the figure shows, although the actual difference ei between adjacent pixels is small, the decoded image immediately before it is used as a prediction value contains the quantization error eq, so the data ei + eq above the actual information is encoded. Need to be
There is a problem that the coding efficiency becomes poor. Further, especially when the transfer rate is low, the quantization error eq becomes large, so that there is a drawback that a high frequency component which is not in an actual image is generated as shown in FIG. 12 and the image quality is deteriorated.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide an image coding method and apparatus which have high coding efficiency and high image quality.

[0008]

According to the present invention of claim 1, a predetermined image signal B is input, a predicted value for the image signal B is generated from another input image signal A, and the predicted value and the image signal B are generated. And the difference image signal is encoded,
The encoded difference encoded image signal is decoded, it is returned to the image signal A'corresponding to the other input image signal A, and the difference between the restored image signal A'and the other input image signal A. Is a quantization error, and the difference image signal or the encoding method is changed according to the result of the calculated quantization error or the result of the quantization error and the difference image signal.

According to a tenth aspect of the present invention, an image signal storage means for storing the input image signal A, and a predicted value for the input predetermined image signal B are generated from the image signal A stored in the image signal storage means. Predicting value generating means, image signal error calculating means for calculating the difference between the predicted value and the image signal B, encoding means for encoding the difference image signal, and the difference encoded by the encoding means. Restoration means for decoding the coded image signal and returning it to the image signal A ′ corresponding to the image signal A, and quantization error calculation for calculating the difference between the restored image signal A ′ and the image signal A Means and
Depending on the result of the quantization error calculated by the quantization error calculating means or the result of the difference image signal calculated by the quantization error and the image signal error calculating means, the difference image signal or the encoding of the encoding means An image coding apparatus including a differential signal coding changing unit that changes the method.

[0010]

According to the present invention, the predicted value of the input image signal B is generated from another input image signal A, and the predicted value and the image signal B are generated.
Calculating a differential image signal from, and encoding the differential image signal, and decoding the encoded differential encoded image signal,
It is returned to the image signal A ′ corresponding to the other input image signal A, and the restored image signal A ′ and the other input image signal A
And the difference is calculated as a quantization error, and the difference image signal or the encoding method is changed according to the result of the calculated quantization error or the result of the quantization error and the difference image signal.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings showing its embodiments.

In the following first to fifth embodiments, the past image signal itself is treated as a predicted value.

FIG. 1 is a schematic block diagram of an image coding apparatus according to the first embodiment of the present invention, FIG. 2 is a block diagram showing details of the image coding apparatus, and FIG. 3 is its coding. It is a figure explaining a method. That is, at the input terminal of the image signal of the image encoding device, the original image prediction error calculation circuit 1 for calculating the difference between the previously input pixel and the later input pixel, and the previously input pixel and its pixel are encoded. The quantization error calculation circuit 2 for calculating the difference between the decoded pixel decoded and decoded again is connected. Original picture prediction error calculation circuit 1
The quantization error calculation circuit 2 is connected to a coding circuit 3 that codes a pixel signal, and the output of the coding circuit 3 is an output terminal and a decoding unit that is a restoring unit that restores the coded signal into a pixel signal. The decoding circuit 4 is connected to the quantization error calculating circuit 2.

Details of the above-mentioned circuits are as shown in FIG.
The original image prediction error calculation circuit 1 is a memory 1 for storing pixel signals.
1 and a subtracter 12 that calculates the difference between the pixel signal stored in the memory 11 and the pixel signal that is input next, and the quantization error calculation circuit 2 stores the pixel signal in the memory 21 and Subtractor 2 for calculating the difference between the pixel signal stored in the memory 11 and the pixel signal encoded by the encoding circuit 3 and restored by the decoding circuit 4.
It is composed of two. Also, the encoding circuit 3 is a subtractor 2
2, a weighting circuit 33 for weighting the difference signal output, an adder 31 for adding the weighted difference signal and the difference signal output from the subtracter 12, and a quantum for quantizing the added signal. The decoding circuit 4 includes a dequantizer 41 that returns the quantized signal to a signal before being quantized, and an adder 42 that restores a pixel signal using the returned signal. , And a memory 43 that stores the restored pixel signal.
The memories 11 and 21 described above constitute image signal storage means,
The subtractor 12 constitutes an image signal error calculating means, and the subtractor 2
2 constitutes a quantization error calculating means, the adder 31 and the quantizer 32 of the encoding circuit 3 constitute an encoding means, and the weighting circuit 33 constitutes a difference signal encoding changing means.
In this case, the memory 11 also serves as a predicted value generating means.

The operation of the image coding apparatus configured as described above will be described with reference to FIGS.

The original image prediction error calculation circuit 1 calculates the input pixel and the past input pixel as a prediction value, and calculates the difference between them as a prediction error. On the other hand, the quantization error calculation circuit 2 outputs the difference between the past input pixel and the decoded pixel in which the pixel is encoded and further decoded, as the quantization error. The encoding circuit 3 changes the prediction error according to the output quantization error information, and encodes the changed signal. The encoded signal is sent to the transmission path via the output terminal and is sent to the decoding circuit 4. The decoding circuit 4 locally decodes the input signal.

The above process will be described in detail with reference to FIGS. In FIG. 2, one pixel is stored in the memory 11, the difference between that pixel and the pixel to be encoded, that is, the pixel immediately after (image signal B) is calculated by the subtractor 12, and the original image prediction error is output. This original picture prediction error is ei in FIG. 3, that is, true difference information that must be transmitted. On the other hand, the quantization error calculation circuit 2 stores in the memory 21 the pixel one pixel before the pixel to be encoded (another input image signal A or image signal A), and outputs it from the decoding circuit in the subtractor 22. (Image signal A ′), that is, the difference from the locally decoded output one pixel before is calculated. This difference corresponds to eq in FIG. 3, that is, the quantization error. In the conventional example, ei + eq was encoded, but in the present invention, for example, ei + eq is quantized only when eq is equal to or greater than a predetermined threshold value, and in other cases, only ei is quantized. For example, in FIG. 3, since eq is small at points a to d, only ei is encoded. At point e, eq is above the threshold, so ei +
encode eq. The above operation can be regarded as weighting on the quantization error, and can be realized by the weighting circuit 33, the adder 31, and the quantizer 32. The encoded output is decoded by the dequantizer 41 and the adder 42 and stored in the memory 43.

According to the above method, when the true information ei is small, it was conventionally necessary to send eq + ei, but in this case e
Only i is required, and the encoded output can be reduced. At this time, if only ei is always encoded, if the quantization error eq has the same polarity, the quantization error accumulates and causes deterioration in image quality. In this embodiment, since the quantization error of a certain value or more is encoded, eq is not accumulated and effective encoding is possible.

FIG. 4 is an example of a configuration diagram of the weighting circuit 33. The weighting circuit 33 can be realized simply by the CPU. An example of an algorithm for weighting the quantization error described in the above embodiment by the CPU of FIG. 4 is shown in FIG.
Shown in.

FIG. 6 is a block diagram of an image coding apparatus according to the second embodiment of the present invention, in which the same operation as in FIG. 1 is performed. The difference from the first embodiment is that the memories 11 and 21 are shared. According to this embodiment, the same effect as that of the first embodiment can be realized with less hardware as compared with the first embodiment.

FIG. 7 is a block diagram of an image coding apparatus according to the third embodiment of the present invention. The difference from the first and second embodiments is the configuration of the encoding circuit 3. In this embodiment, in order to change the encoding method by controlling the quantization characteristic of the quantizer 32, the encoding circuit 3 controls the quantizer 32 and the quantization characteristic control for controlling the quantization characteristic. It is composed of a circuit 34. In the present embodiment, instead of adding the quantization error, the quantization of the original image prediction error is controlled so that the quantization error does not accumulate when the quantization error is a certain value or more. For example, the polarity of the original image prediction error is reversed so as to be eq-ei to prevent accumulation. According to the above embodiment, the same effects as those of the first and second embodiments can be obtained.

FIG. 8 is a block diagram of an image coding apparatus according to the fourth embodiment of the present invention. The difference from the third embodiment is that the quantization characteristic control circuit 34 is controlled so that the encoding characteristic is not controlled only by the quantization error but is controlled by using the information of both the quantization error and the original picture prediction error. This is the point of composition. In this embodiment, the quantization characteristic is controlled as follows. If the distance | ei-eq | between the original image prediction error ei and the quantization error eq is less than a fixed value, only ei is coded, and if it is more than the fixed value, ei + eq is coded. The same effects as those of the first and second embodiments can be obtained by the above embodiments.

FIG. 9 is a block diagram of an image coding apparatus according to the fifth embodiment of the present invention. In this embodiment,
Unlike the fourth embodiment, an image in frame units is input and the inter-frame difference is encoded. Therefore, in FIG. 9, the memory 11 in the original image prediction error calculation circuit 1 and the memory 43 in the decoding circuit are frame memories. In addition, a two-dimensional DCT circuit 35 is added to the encoding circuit 3 and an inverse two-dimensional DCT circuit 44 is added to the decoding circuit, assuming that the inter-frame difference is subjected to two-dimensional DCT.

The image coding apparatus configured as described above will be described below with reference to FIG.

The basic operation is the same as in the first to third embodiments. In the present embodiment, as in the conventional example, the image is divided into small blocks and the inter-frame difference is calculated on a block-by-block basis using a two-dimensional DC
T, quantize and compress and code. At this time, the quantization error is weighted as in the first embodiment, added to the true information ei, and then encoded. Weighting is performed by a two-dimensional FIR filter. Various weighting is possible by changing the characteristics of the two-dimensional filter. Examples of weighting include human visual characteristics, frequency characteristics of original image (image signal), frequency characteristics of original image prediction error (difference image signal), and the like. Make each quantization error a characteristic that is inconspicuous to the human eye, make it the same as the frequency characteristic of the original image to make noise inconspicuous, make it the same as the frequency characteristic of the original image prediction error, and make noise as an error It has the feature of making it inconspicuous, and can be changed depending on the image or the purpose of use.

FIG. 10 is a block diagram of an image coding apparatus according to the sixth embodiment of the present invention. In the present embodiment, the configuration of the encoding circuit 3 is changed from that of the fifth embodiment, the two-dimensional DCT circuit 37 is placed in front of the weighting circuit 33, and the position of the adder 31 is set next to that of the two-dimensional DCT circuit 35. Have moved to. With the above configuration, two two-dimensional DCT circuits are required, but weighting can be performed on the frequency and the same effect as that of the fifth embodiment can be obtained.

FIG. 11 is a block diagram of an image coding apparatus according to the seventh embodiment of the present invention. In the present embodiment, the present invention is combined with the motion compensation and inter-frame difference two-dimensional DCT similar to the conventional example. Motion detection circuit 10 of FIG.
7 and the motion-compensated inter-frame prediction circuit 13 constitute a predicted image generation means, and the motion detection circuit 107 and the motion-compensated inter-frame prediction circuit 45 prediction restored image generation means. That is, the motion detection circuit 107 detects a motion vector from the first frame and the second frame, and the motion-compensated inter-frame prediction circuit 13 uses the motion vector.
A prediction image corresponding to the first frame to the second frame is generated, while the motion compensation inter-frame prediction circuit 45 uses the motion vector to predict the first frame to the second frame which is restored after being encoded and then restored. Generate a restored image.
Next, the difference between the second frame and the predicted image is calculated as the original image prediction error, and the difference between the predicted image and the predicted restored image is calculated as the quantization error. Perform the same processing as.

As in the present embodiment, if the image coding system has a so-called predictive coding system that takes a difference between frames or a difference between pixels, the DPC is used.
Regardless of M, ADPCM, DCPM with motion compensation, etc., the present invention can be easily combined with existing coding schemes,
It is possible to realize more efficient encoding than the conventional encoding method. In the seventh embodiment, the frame memory 11 is added as compared with the conventional example. However, since a frame memory is normally used as an input buffer and an input buffer is always required for motion detection, in an actual encoding device, when adding a frame memory to obtain the effect of the present invention, Almost never.

As described above, according to the present invention, since the true original picture prediction error and the quantization error are separately encoded, the true information is correct, and the quantization error can be encoded in consideration of visual characteristics. Encoding with high efficiency and good image quality is possible. Also,
All are related to the encoding unit, and the decoder can be used as it is without changing the conventional configuration.

In the fifth to seventh embodiments, an example of weighting the quantization error has been described, but the present invention is not limited to this, and a quantization characteristic control circuit is added as in the third embodiment. However, the same effect can be obtained even if the quantization characteristic of the original image prediction error is controlled by the quantization error.

Further, in the sixth and seventh embodiments, the two-dimensional DCT has been described as an example, but the present invention is not limited to this and can be applied in combination with block coding such as vector quantization or subband coding. Is.

Further, a filter or the like may be used to generate the prediction value by the motion compensation inter-frame prediction circuit 13 or 45 which is the prediction value generation means of the above embodiment.

[0033]

As is apparent from the above description, the present invention has the advantages that the coding efficiency is improved and the image quality is improved.

[Brief description of drawings]

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

FIG. 2 is a block diagram of an image coding apparatus according to the same embodiment.

FIG. 3 is a diagram illustrating an encoding method according to the same embodiment.

FIG. 4 is a configuration diagram of a weighting circuit of the embodiment.

FIG. 5 is a flowchart of an algorithm of the weighting circuit according to the embodiment.

FIG. 6 is a block diagram of an image coding apparatus according to a second embodiment of the present invention.

FIG. 7 is a block diagram of an image coding apparatus according to a third embodiment of the present invention.

FIG. 8 is a block diagram of an image coding apparatus according to a fourth embodiment of the present invention.

FIG. 9 is a block diagram of an image coding apparatus according to a fifth embodiment of the present invention.

FIG. 10 is a block diagram of an image encoding device according to a sixth embodiment of the present invention.

FIG. 11 is a block diagram of an image coding apparatus according to a seventh embodiment of the present invention.

FIG. 12 is a block diagram of a conventional image encoding device.

FIG. 13 is a diagram illustrating a conventional encoding method.

[Explanation of symbols]

 1 Original Image Prediction Error Calculation Circuit 2 Quantization Error Calculation Circuit 3 Encoding Circuit 4 Decoding Circuit 11, 21 Memory 12, 22 Subtractor 13 Motion Compensation Interframe Prediction Circuit (Original Data) 31 Adder 32 Quantizer 33 Weighting Circuit 34 Quantization Characteristic Control Circuit 45 Motion Compensation Interframe Prediction Circuit (Restored Data) 107 Motion Detection Circuit

Claims (12)

[Claims] 1. A predetermined image signal B is input, a predicted value for the image signal B is generated from another input image signal A, a difference between the predicted value and the image signal B is calculated, and the difference image signal is calculated. Is encoded, the encoded differential encoded image signal is decoded, it is returned to the image signal A ′ corresponding to the other input image signal A, and the restored image signal A ′ and the other input are input. A difference from the image signal A is calculated as a quantization error, and the encoding method is changed according to the result of the calculated quantization error or the result of the quantization error and the difference image signal. Image coding method. 2. A predetermined image signal B is input, a prediction value for the image signal B is generated from another input image signal A, a difference between the prediction value and the image signal B is calculated, and the difference image signal is calculated. Is encoded, the encoded differential encoded image signal is decoded, it is returned to the image signal A ′ corresponding to the other input image signal A, and the image signal B ′ is restored from the restored image signal A ′. Reconstructed predicted value is generated, and the difference between the reconstructed predicted value and the predicted value is calculated as a quantization error, and the result of the calculated quantization error or the result of the quantization error and the difference image signal An image coding method, characterized in that the coding method is changed according to. 3. The input image signal A is stored, the difference between the stored image signal A and the next input image signal B is calculated, the difference image signal is encoded, and the encoded image signal is encoded. The difference-encoded image signal, restore it to an image signal A'corresponding to the stored image signal A, and restore the restored image signal A'and the stored image signal A.
An image characterized by changing the encoding method according to the result of the calculated quantization error or the result of the calculated quantization error and the difference image signal. Encoding method. 4. The image coding method according to claim 1, wherein the image signal is a signal for each pixel or each frame. 5. The encoding method is changed by adding the quantization error to the difference image signal when the quantization error exceeds a predetermined value. Or the image coding method described in 3. 6. The encoding method is changed by adding a value after performing a predetermined weighting to the quantization error to the difference image signal. Image coding method. 7. The image coding method according to claim 6, wherein the predetermined weighting is a frequency characteristic that is substantially the same as a human visual characteristic. 8. The image coding method according to claim 6, wherein the predetermined weighting has substantially the same frequency characteristic as the frequency characteristic of the input image signal. 9. The image coding method according to claim 6, wherein the predetermined weighting has substantially the same frequency characteristic as the frequency characteristic of the differential image signal. 10. An image signal storage means for storing the input image signal A, and a predicted value generation means for generating a predicted value for the input predetermined image signal B from the image signal A stored in the image signal storage means. An image signal error calculating means for calculating a difference between the predicted value and the image signal B, an encoding means for encoding the difference image signal, and a difference encoded image signal encoded by the encoding means. Decrypt the
Restoration means for returning it to the image signal A'corresponding to the image signal A, quantization error calculation means for calculating the difference between the restored image signal A'and the image signal A, and its quantization A difference for changing the encoding method of the encoding means according to the result of the quantization error calculated by the error calculating means or the result of the quantization error and the difference image signal calculated by the image signal error calculating means. An image coding apparatus comprising a signal coding changing unit. 11. An image signal storage means for storing the input image signal A, an image signal error calculation means for calculating a difference between the stored image signal A and the next input image signal B, and Encoding means for encoding the differential image signal,
Restoration means for decoding the differentially encoded image signal encoded by the encoding means and returning it to the image signal A ′ corresponding to the stored image signal A, and the restored image signal A ′. Error calculation means for calculating the difference between the image signal A and the image signal A, and the result of the quantization error calculated by the quantization error calculation means, or the quantization error and the difference calculated by the image signal error calculation means. An image coding apparatus comprising: a differential signal coding changing unit that changes a coding method of the coding unit according to a result of an image signal. 12. An image signal storage means for storing an input image signal A, an image signal error calculation means for calculating a difference between the stored image signal A and an image signal B input next, and the image signal error calculation means. Encoding means for encoding the differential image signal,
Restoration means for decoding the differentially encoded image signal encoded by the encoding means and returning it to the image signal A ′ corresponding to the stored image signal A, and the restored image signal A ′. Error calculation means for calculating the difference between the image signal A and the image signal A, and the result of the quantization error calculated by the quantization error calculation means, or the quantization error and the difference calculated by the image signal error calculation means. An image coding apparatus comprising: a differential signal coding changing unit that changes a coding method of the coding unit according to a result of an image signal.