JPH09200769A - Inter-motion compensation frame encoding system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent predictive slave block distortion from being conspicuous when the frame correlation between preceding and following frames in a scene change, etc., is low. SOLUTION: Predictive error power provided from a motion vector detection part 9 is inputted to a scene change detecting means 1 and it is discriminated whether the relevant frame is the frame of scene change or not. When that frame is discriminated as the scene change, in the case of encoding a macro block later, data correction is performed to the relevant frame for suppressing a high frequency component in the data outputted from a quantizing means.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion-compensated interframe coding system, and more particularly to a coding system for improving image quality by changing a quantization matrix when a scene change occurs in a moving image.

[0002]

2. Description of the Related Art Interframe coding methods are relatively often used as data compression means for image signals. Although various interframe compression coding methods have been proposed, the frame difference of the image signal is coded, and the orthogonal transformation which efficiently uses the correlation in the two-dimensional space of the image with respect to this difference signal. The transform coding used is often used,
In particular, an encoding method based on DCT as orthogonal transformation is ISO / IEC which is a moving image encoding standard for storage media.
It is adopted in an international standard encoding system such as 11172-2 (hereinafter, commonly referred to as MPEG1).

[0003] When performing compression encoding based on the MPEG1 standard, the compression encoding unit is divided into hierarchies. That is, an input signal is separated into a luminance signal and a chrominance signal, and an intra-frame coded frame (I frame), a forward predictive coded frame (P frame) or a bidirectional predictive coded frame (B frame) for each frame. Prediction method. Further, the frame is divided into units of grid-like macroblocks, the motion compensation and the quantization step are determined for each macroblock, and the luminance signal is further divided into four in the DCT process, the quantization process and the variable length coding process. In blocks, luminance signal blocks Y0, Y1, Y
The encoding process is performed in the order of 2, Y3, and the color signal blocks Cb and Cr.

[0004] Encoding of an image by a conventional DCT-based encoding method will be described by taking MPEG1 as an example. FIG. 8 is a block diagram of an example of a conventional image encoding device. In this conventional image coding apparatus, the input image data is divided into two-dimensional macroblocks, and a motion vector detection unit 87 detects a motion vector for each macroblock between one and two reference frames. . Based on the motion vector information of the macroblock, the motion compensation interframe difference data is obtained, the difference data is divided into 8 × 8 blocks, and then the DCT unit 81 performs orthogonal transformation to quantize the transformed data. The unit 82 performs sampling, and the variable-length coding unit 83 creates compression-coded data. Based on the code amount of the encoded data output from the variable-length encoding unit 83, the code amount control unit 88 adjusts the quantization coefficient so as to match the target code amount of the compressed data.

In the inter-frame coding method, in order to perform inter-frame prediction, it is necessary to create reference frame data for inter-frame coding of a succeeding frame. When the frame to be encoded is an I frame or a P frame, the dequantization unit 84 and the inverse DCT unit 85 perform decompression processing and store the image data in the frame memory 86.
The image data in the frame buffer 86 becomes reference frame data for inter-frame encoding of a subsequent frame.

In such an interframe coding method, when the target frame is an interframe coded frame,
A difference between frames of the image signal is obtained, and this difference value is converted into a variable length code such as Huffman code. Due to the characteristics of moving images,
The frame correlation after motion compensation becomes high, and the difference value becomes 0 or close to 0. By assigning a short variable length code to them, the compression efficiency is improved. As a conventional example, when the data after motion compensation prediction of a divided block becomes 0, it can be an uncoded block, and therefore the compression efficiency can be further improved.

As a conventional method for improving compression efficiency by creating non-encoded blocks, there is an image encoding control method disclosed in Japanese Patent Laid-Open No. 2-241289. Hereinafter, a conventional image encoding control method will be described with reference to FIG.
The subtraction circuit 91, the quantizer 94, the addition circuit 99, and the one-frame delay circuit 98 constitute an inter-frame encoding unit. The valid / invalid block determination circuit 92 accumulates the inter-frame difference signal in block units and determines whether the block is valid or invalid. The difference signal is forcibly set to 0. In this case, when it is determined that the block of the luminance signal is invalid, the block of the color signal corresponding to the block is forcibly processed as invalid, and the luminance is determined. When it is determined that the signal block is valid, the color signal block corresponding to the block is also processed as valid. Therefore, the luminance signal and the color signal are paired and transmitted as the inter-frame coded signal, and only one of the luminance signal and the color signal is not transmitted.

In particular, in the MPEG compression encoding method, the definition of a block (skip macroblock) that does not require generation of variable-length code data is defined in a macroblock hierarchy higher than the block. The skip macroblock is a block to which the motion-compensated macroblock data of the reference frame is pasted (memory copy is performed) when the difference macroblock data after the motion-compensated inter-frame prediction becomes 0. As a result, the amount of code assigned to other macroblocks can be increased,
By reducing the quantization coefficient, block noise generated due to a quantization error is suppressed, and as a result, image quality is improved.

Further, as a method for appropriately allocating the code amount by correcting the data after the quantization processing for the purpose of improving the image quality of the compression-encoded moving image, for example, Japanese Patent Application No. 7-184170, there is a skip for a macroblock having a high frame correlation by correcting the quantized data as shown in FIG. 10 with respect to the coded block in the general MPEG shown in FIG. It has been proposed to forcibly generate macroblocks to improve image quality. This method corrects the quantized data under the condition of the motion vector information, the frame type of the target frame, and the component value of the quantized data, thereby forcibly generating the skip macroblock. It is possible to allocate a larger code amount to other blocks. This makes it possible to reduce the quantization scale of the macroblock as a whole. That is, the quantization error can be reduced, and image quality degradation represented by block distortion can be suppressed. When the skip macroblock occurs, the variable length coding process in the macroblock can be omitted and the compression speed can be improved.

The generation of the skip macroblock is MPE
In the G1 standard, it is permitted in P frame and B frame.

The conditions for the MPEG1 standard skip macroblock in a P frame are defined as follows.

First, all the elements of the 8 × 8 quantized intra-block data are 0, and secondly, the value of the motion vector of the target macroblock is 0 in both the x and y directions. Third, the coding type of the macroblock is determined to be the interframe coding block.

In Japanese Patent Application No. 7-184170, the conditions for forcibly correcting data in P frames are as follows:
4) I am doing so. That is, the processing according to the flowchart shown in FIG. 11 is performed.

1) When α in the block data after 8 × 8 quantization, (α is an integer of 0 or more).

2) (β is an integer of 0 or more).

3) Both the x and y components of the motion vector of the encoding target macroblock are 0.

4) The coding type of the macroblock is determined to be an interframe coding block.

Next, the conditions for the MPEG1 standard skip macroblock in the B frame are defined as follows.

First, all the elements of the 8 × 8 quantized intra-block data are 0, and second, the motion vector value of the target macroblock is the most recent in both the x and y directions. Match the motion vector of the encoded macroblock, thirdly, the coding type of the macroblock is the block for interframe coding, and is the same as the coding type of the most recently coded macroblock It has been determined by.

In the specification of Japanese Patent Application No. 7-184170, B
In the frame, the conditions for forcibly correcting the data are as follows 1) to 4). That is, the processing according to the flowchart shown in FIG. 12 is performed.

1) When α is 8 × 8 quantized in-block data, α is an integer of 0 or more.

2) (β is an integer of 0 or more).

3) The absolute value of the difference between the motion vector component of the most recently encoded macroblock and the x and y components of the motion vector of the encoding target macroblock is 0 or more and an integer γ or less.

4) It has been determined that the coding type of the macroblock uses the same prediction method as that of the macroblock which has been most recently variable length coded.

In the motion compensation interframe coding according to the above method, it is considered that the image quality is improved by directly using the reference frame data as the skip macro block, rather than by coding the differential block information. By generating a skip macroblock forcibly by data correction of block data based on vector information and quantized block data, the generated code amount of the block is set to 0 and the code amount allocated to other blocks is increased. By doing so, a method for improving the image quality is described.

[0026]

However, even if the method described in the conventional example is used, when a scene change occurs in a moving image, the frame correlation becomes extremely low, so that the skip macro is forced. The block itself cannot be generated. In such a case, it is not possible to solve the problem that the quantization scale becomes large over the entire screen and the image quality is extremely deteriorated due to the occurrence of block noise or the like.

Further, by performing encoding once, etc.,
If you know the frame in which the scene change occurs, that is, the frame where the generated code amount becomes extremely high,
It is also possible to set parameters such as quantization for the corresponding frame in advance, but in this case, since encoding work is required twice, especially when considering moving image compression with software, The time required for the compression process significantly increases. Real-time compression with hardware is also impossible unless there is a means to store the entire image scene on a storage medium such as a video tape or a hard disk.

According to the present invention, when the skip macroblock is forcibly generated, the image quality is severely deteriorated, and it is expected that the transition of the scene and the generated code amount will be extremely generated by one compression coding process. It is an object of the present invention to solve the above problems and improve the image quality of a compression-encoded image by detecting a frame to be encoded during the encoding process and performing band compression on a high frequency component in the quantized data. To do.

[0029]

In order to achieve the above object, the motion-compensated interframe coding system of the present invention divides an input image signal into blocks by a predetermined number of samples and collectively codes the blocks. In an inter-frame coding processing method for performing a coding process, in order to perform motion-compensated inter-frame prediction, a frame memory that stores one or two images as reference frames, and between the frame memory and the block, Motion vector detection means for detecting a block in a reference frame that minimizes the difference between the block and the reference frame, and discrete cosine transform of the inter-frame difference signal for which motion compensation prediction has been performed for each divided block. Transforming means, a quantizer for quantizing the output obtained by the discrete cosine transform, and encoding the quantized data. A variable length coder for encoding and a power compensation component of a motion compensation inter-frame prediction block output from the motion vector detecting means to determine whether the frame is a scene switch. Characteristic scene change detection means, a first register that stores the presence or absence of a scene change or the size of the area where the scene change occurs, obtained from the scene change detection means, and a band limit for the block. In order to perform, a second register storing a plurality of types of low-pass characteristics, a value of block data after quantization with respect to an output of the quantizer before the variable length coding, Depending on the presence or absence of the scene change of the encoding target frame obtained from the first register, the value stored in the second register is used for quantization. It is obtained so as to have a data correcting means and correcting the block data.

The motion-compensated interframe coding method of the present invention is an interframe coding processing method in which an input image signal is divided into blocks for each predetermined number of samples, and the coding processing is collectively performed for each block. , A frame memory that accumulates one or two images as a reference frame for performing motion compensation inter-frame prediction, and the difference between the block and the reference frame between the frame memory and the block is the smallest. Motion vector detecting means for detecting such a block in the reference frame, converting means for performing discrete cosine transform on the inter-frame difference signal for which motion compensation prediction has been performed for each divided block, and obtained by the discrete cosine transform. A quantizer for quantizing the output, a variable length encoder for coding the quantized data, and a coding pair A motion compensation inter-frame difference block is obtained between the block and the motion compensation inter-frame prediction block output from the frame memory, and the frame is calculated based on the power component of the prediction error of the motion compensation inter-frame difference block. Scene change detecting means for determining whether a scene change has occurred, a register storing a plurality of types of low-pass characteristics for performing band limitation on the block, and the variable length code. With respect to the output of the quantizer before encoding, it is stored in the register depending on the value of the block data after quantization and the presence or absence of the scene change of the encoding target frame obtained from the scene change detecting means. The data correction means is characterized in that the block data after quantization is corrected using a value. Than it is.

The motion-compensated interframe coding method of the present invention has an effect of a low-pass filter (LPF) on the quantized data when it is determined that a frame in which a scene change occurs in an image scene. By intentionally performing data correction, that is, a process of slightly blurring the image, block noise that is very visually obtrusive is reduced.

The above-mentioned means for judging whether or not the frame has a low frame correlation is such that when the number of macroblocks having a large power of the frame difference after the motion prediction exceeds the threshold value A continuously, a scene change occurs. Is determined to occur, and at this time, band limitation is performed on the quantized block. As a method, (1) the high frequency component of the quantized coefficient is set to zero. (2) Each component aij is 0 ≦ αij ≦
Multiply by a coefficient matrix α of 1. However, the coefficient matrix α has a characteristic that the coefficient value related to the low frequency component is closer to 1, and the coefficient value related to the high frequency component is closer to 0.

When the above-mentioned method is used, since the frame in which the scene change has occurred is determined by using the data generated by the conventional compressor, the scale of hardware and software is almost the same. Does not increase. Also, since a scene change is detected by one-time encoding processing, it is possible to minimize the decrease in image encoding speed.

[0034]

BEST MODE FOR CARRYING OUT THE INVENTION (Description of Configuration) Next, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram of a first embodiment of a motion-compensated interframe coding system according to the present invention.

Referring to FIG. 1, in the first embodiment of the present invention, a scene change detecting section 1 for judging a scene change from a detection result of a motion vector detector 9 and a register for storing information about a scene change regarding the frame. 11, a DCT unit 2 that orthogonally transforms the horizontal and vertical coordinate axis components of the pixel intensity supplied with the prediction error to the horizontal and vertical frequency axis components, a quantizer 3 that quantizes the output from the DCT unit, and a quantized data scene Data correction means 4 that corrects based on change information, variable length coding unit 5 that performs Huffman coding on the corrected data, reference frame data for motion prediction when the target frame is an I or P frame Inverse quantization unit 6 and inverse DCT unit 7 for creating a frame, a frame memory 8 for storing reference frame data for prediction, and a block The motion vector for each macroblock is detected by the matching method, supplied to the frame memory 8 and the variable length coding unit 5, and the power of the prediction error regarding the detected block is output to the scene change unit 9, and the generated code amount. , A code amount control unit 10 for controlling a quantization scale for adjusting so as to match the target code amount, and appropriate parameters required by the data correction unit 4 based on the determination contents stored in the register 11,
The coefficient matrix section 12 stores data as a coefficient matrix.

(Explanation of Operation) In FIG. 1, when the compression target frame is an I frame, a constant value 0 is output from the frame memory 8 and the input luminance information divided into blocks,
The color information is directly input to the DCT unit 2 and subjected to orthogonal transformation. The output of the DCT unit is sent to the quantization unit, and division is performed on the quantization scale. In the intra-coded frame, data subjected to Huffman coding by the variable length coding processing unit 5 is output without performing data correction. The Huffman-encoded data is also input to the code amount control unit 10. The code amount control unit 10 controls the quantization scale to be larger when the code amount is larger than the target code amount in order to match the average output code amount of the encoder with the target code amount. However, if the generated code amount is smaller than the target code amount, control is performed to reduce the quantization scale.

Further, the inverse quantizer 6 multiplies the quantized data by the quantization scale of the macroblock, and then performs the inverse DCT operation, and this data is used as reference frame data for interframe coding of the subsequent image. To the frame memory 8.

Next, when the frame to be compressed is a P frame, the motion amount of the macroblock is first extracted by the motion vector detecting section 9 using the frame data written in the frame memory 8 as a reference frame, and the detected motion amount is calculated. The predicted block data to be referenced is output from the frame memory 8. The difference between the predicted value and the encoding target macroblock is input to the DCT unit 2, and orthogonal transformation processing is performed.
Performs quantization processing. If the data correction unit 4 determines based on the quantized data that scene switching has occurred in the frame, the coefficient matrix 1
Based on 2, the data correction 4 limits the band for the quantized output, and once the value of the high-frequency component of the block data is forcibly set to 0 for the quantized output, the censoring process is once performed.

The scene change detection means 1 determines whether or not a scene change has occurred in the frame. In the first embodiment, the scene change detecting means 1 acquires data for determining whether a scene change has occurred from the motion vector detecting means 9 as shown in FIG. For the determination, the error power between the coding target macroblock and the prediction block of the reference frame, which is calculated at the time of block matching at the time of motion vector detection, for example, the square of the difference between each pixel of both blocks is calculated. The sum or the sum of the absolute values of the differences is used.

The scene change detection means 1 determines whether or not a scene change has occurred in the frame and in what size area according to the flowchart shown in FIG. The operation of the scene change detection means will be described with reference to FIG.

First, at the beginning of the processing of each frame, the values of the scene change determination flag, the number of occurrence points, and the number of blocks with a large prediction error (counter 1) are initialized (step 31). Next, the prediction error power P is obtained from the prediction error component used at the time of vector detection by the motion vector detection unit 9 by square error calculation or the like (step 32). When the power P is larger than the preset threshold Pth, 1 is added to the number of blocks (counter 1) having a large prediction error (step 33, step 34). Steps 32 to 34 are repeated, and if the value of the counter 1 exceeds the set value A, it is determined that a scene change has occurred in this frame (step 35), the scene change occurrence flag is turned on, and the scene change occurs. 1 is added to the number of occurrence points (counter 2) (step 36). At this time, as the setting value A, for example, the number of macroblocks included in one line of the image is selected. Next, the counter 1 is cleared (step 37), and steps 32 to 37 are repeated until the final macroblock of the frame (step 38). As a result, the determination result as to whether a scene change has occurred is stored in the scene change occurrence flag, and the number of areas determined to be a scene change is stored in the counter 2.

Generally, in the motion compensation interframe coding method, when the frame is a P frame or a B frame, the motion vector detection process is completed before the coding process of each macroblock as shown in FIG. To do. Therefore, in the case of the configuration shown in FIG. 1 in which the scene change is determined using the prediction error component at the time of detecting the motion vector, the information about the scene change regarding each frame is registered in the register 11
To be stored.

On the basis of the information about the scene change stored in the register 11, the data correction unit 4 performs band limitation processing on the quantized data.

The quantized data is subjected to the following correction, for example.

(1) The data in the block of 64 points is rearranged in the order of zigzag scanning so that the lower frequency component is closer to the beginning of the data in the order of the arrows shown in FIG. Set to 0.

(2) Each of the 64-point block data is multiplied by the coefficient matrix as shown in the example of FIG.

When performing the processing of (2), when the value of the counter 2 is large, that is, when the frame correlation of the image is very low, the band pass characteristic is set to a narrower band, and when the value of the counter 2 is small. That is, when the frame correlation is low in only a part of the frame, the coefficient matrix is switched such that the bandpass characteristic is slightly widened.

The quantized coefficient after data correction by the above method is subjected to variable length coding 5 and coded and output.

The above-described data correction function is a kind of LP.
Although it corresponds to F (low-pass filter), the number of effective coefficients of the quantized data of each macroblock is reduced by performing this process, so that it is possible to prevent the quantization scale from rapidly increasing even at a scene change. it can. Therefore, the quantization error is reduced particularly in the low frequency component, and the distortion deterioration can be made inconspicuous. (Other Embodiments) Next, a second embodiment of the present invention will be described with reference to FIG.

The difference from the first embodiment is that the scene change occurrence detection is not performed at the time of motion vector detection as in the first embodiment, but the power component after motion prediction, that is, the power input to the DCT unit 2 is calculated. This is done using the ingredients. In this case, since the calculation is performed using the block data after the actual motion estimation, the accuracy of the scene change determination can be improved.

In the case of the second embodiment, D of the frame
Occurrence of a scene change is indefinite when the motion prediction error is input to the CT unit. Therefore, the determination procedure is as shown in FIG. Therefore, the scene change flag is turned on only when the processing of a certain number of macroblocks is completed.

A scene change determination procedure will be described with reference to the flowchart of FIG. First, at the beginning of processing each frame, the scene change determination flag, the number of occurrence points,
The value of the number of flocs (counter 1) having a large prediction error is initialized (step 41).

Next, the difference between the prediction block output from the motion vector detection 9 and the block of the encoding target frame is obtained, and this value is input to the DCT section 2 and at the same time input to the scene change detection means 1. The power of this prediction error block is calculated by calculating the sum of squares (step 42). If this power P is larger than a preset threshold Pth, 1 is added to the number of blocks (counter 1) having a large prediction error (step 43,
Step 44). Steps 42 to 44 are repeated. If the counter value exceeds the set value A, it is determined that a scene change has occurred in this frame (step 45), the scene change occurrence flag is turned on, and The scene change determination process ends. At this time, as A, for example, the number of macroblocks included in one line of the image is selected.

When the scene change occurrence flag is turned on, the information that the scene change has occurred in the frame is immediately sent to the data correction means 4, and all the subsequent macroblocks in the frame,
The data is corrected based on the information stored in the coefficient matrix 12. While the count value does not exceed A, the processes of steps 42 to 45 are repeated until the final macroblock (step 46). As described above, the determination result as to whether a scene change has occurred is stored in the scene change occurrence flag. Information on the above scene change is stored in the GIRESTA 11.

[0055]

As described above, in the motion-compensated interframe coding method of the present invention, when a scene change occurs in a moving image to be coded, calculation is performed in the coding process of a conventional encoder. Since the scene change is detected by using the stored data, the amount of calculation is hardly increased from the conventional encoding process. Further, when a scene change determination is made, band limitation is performed on the block, so that block noise can be reduced even in a frame that causes extreme image quality deterioration such as block distortion in normal encoding, so that the visual image quality can be reduced. It has the effect of making the deterioration of inconspicuous.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a first motion compensation interframe coding method used in the present invention.

FIG. 2 is a block diagram illustrating a second motion-compensated interframe coding method used in the present invention.

FIG. 3 is a flowchart showing the operation of an embodiment of the scene change detection means of the present invention.

FIG. 4 is a flowchart showing the operation of an embodiment of the scene change detection means of the present invention.

FIG. 5 is a characteristic diagram of an example of a coefficient matrix having a low-pass characteristic according to the present invention.

FIG. 6 is an explanatory diagram of zigzag scanning for explaining the band limiting characteristic of the present invention.

FIG. 7 is a flowchart of conventional motion compensation interframe coding processing.

[Fig. 8] Fig. 8 is a block diagram of conventional motion compensation interframe coding processing.

[Fig. 9] Fig. 9 is a block diagram of conventional motion compensation interframe coding processing.

FIG. 10 is a block diagram of a conventional motion-compensated interframe coding process.

FIG. 11 is a flowchart of a conventional data correction method after quantization.

FIG. 12 is a flowchart of a conventional data correction method after quantization.

[Explanation of symbols]

 1 scene change detection unit 2 DCT unit 3 quantization unit 4 data correction unit 5 variable length coding unit 6 inverse quantization unit 7 inverse DCT unit 8 frame memory 9 motion vector detection unit 10 code amount control unit 11 register 12 coefficient matrix unit

Claims (2)

[Claims] 1. An inter-frame coding processing method in which an input image signal is divided into blocks for each predetermined number of samples, and coding processing is performed collectively for each block, in order to perform motion-compensated inter-frame prediction, A frame memory that stores one or two images as a reference frame, and a motion that detects a block in a reference frame such that the difference between the block and the reference frame is the smallest between the frame memory and the block Vector detecting means, transforming means for performing discrete cosine transform of the inter-frame difference signal subjected to motion compensation prediction in each divided block, and quantizer for quantizing the output obtained by the discrete cosine transform. A variable length encoder for encoding the quantized data, and a motion compensation frame output from the motion vector detecting means. Scene change detection means for determining whether or not the frame is a scene change based on a power component of a prediction error of an inter-frame prediction block; and a scene change detection means obtained from the scene change detection means. The first register storing the size of the region where the occurrence of the scene change or the scene change occurs, and the second register storing the plurality of types of low-pass characteristics for band limiting the block. With respect to the output of the quantizer before the variable-length coding, the value of the block data after quantization and the presence or absence of the scene change of the coding target frame obtained from the first register, And a data correction means for correcting the quantized block data using the value stored in the second register. Moving image motion compensation inter-frame coding scheme to. 2. An inter-frame coding processing method in which an input image signal is divided into blocks for each predetermined number of samples, and coding processing is performed collectively for each block, in order to perform motion-compensated inter-frame prediction, A frame memory that stores one or two images as a reference frame, and a motion that detects a block in a reference frame such that the difference between the block and the reference frame is the smallest between the frame memory and the block Vector detecting means, transforming means for performing discrete cosine transform of the inter-frame difference signal subjected to motion compensation prediction in each divided block, and quantizer for quantizing the output obtained by the discrete cosine transform. A variable-length encoder for encoding the quantized data, an encoding target block, and an output from the frame memory. A motion-compensated inter-frame difference block is obtained between the motion-compensated inter-frame prediction block and the motion-compensated inter-frame prediction block. A scene change detecting means for determining the following: a register storing a plurality of types of low-pass characteristics for band limiting the block; and a quantizer before the variable-length coding. Depending on the value of the quantized block data and the presence or absence of the scene change of the encoding target frame obtained from the scene change detection means, the value stored in the register is used for the output. A motion compensation inter-frame code of a moving image, comprising: a data correction unit characterized by correcting the block data Method.