JP2009081622A - Moving image compression encoder - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a moving image compression encoder that requires only a small amount of compression encoded data even when the entire picture undergoes a transition in a moving image. <P>SOLUTION: The moving image compression encoder includes a register for storing a transition vector representing a transition direction and a transition amount of the entire picture, and past transition image data generating means for shifting a past image in a direction indicated by the transition vector to obtain a past transition image, the moving image compression encoder generates a motion vector on the basis of the past transition image and an input image, and performs motion compensation processing on the past transition image on the basis of the motion vector to generate a prediction image. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a moving image compression encoding apparatus that compresses and encodes moving image data.

  MPEG-4 (Moving Picture Experts Group phase-4) is known as a moving image compression coding technique. MPEG-4 is generally used for video distribution with a low image quality and a high compression rate through a line having a low communication speed such as a mobile phone or a telephone line. There are the following two main compression encoding methods for images in MPEG-4. One is intra coding in which compression coding is performed using only information of the current target image. The other is inter coding that compresses and encodes a prediction difference image obtained by a difference between a past image (or future image) and a current target image. Of the image groups constituting a moving image, the difference between images that are continuous with each other on the time axis is often small, and the amount of data can be reduced by compressing the difference. Normally, when compressing image data by inter coding, an image is divided into a plurality of blocks, and motion vectors are generated by motion detection and motion compensation based on the motion vectors is performed for each block.

For example, Patent Document 1 discloses a video signal encoding apparatus using inter encoding. Here, the video signal of the target frame is divided into a plurality of blocks, and the motion vector of each block is detected using the video signals of the frame and the previous frame. Based on the motion vector in the neighboring block of the detection target block, the motion vector of the detection target block is predicted. In the block in which the motion vector for prediction is obtained, a motion vector in the vicinity of the prediction region is searched in detail, and in the block in which the motion vector for prediction is not obtained, a motion vector in a predetermined range is roughly searched. Thereby, the detection accuracy of the motion vector is improved, and the number of operations is reduced.
JP-A-6-225289

  As described above, in inter coding, a motion vector and motion compensation are usually performed for each block. In this case, when a part of the screen changes in the moving image, only the difference between the current target image and the past image in the block corresponding to the part is large, and the difference in the other blocks is small. As a result, the data amount of the entire screen obtained by compressing and encoding the difference data of all blocks is also small. On the other hand, when the entire screen is changed, the difference becomes large in all blocks, so that there is a problem that the data amount of the entire screen obtained by compressing and encoding the difference data of all blocks becomes enormous.

  In normal network distribution, for example, it is necessary to maintain a CBR (Constant Bit Rate) such as 1 Mbit. When the entire screen changes as described above, it becomes difficult to maintain the set bit rate. When the amount of compressed data exceeds the set bit rate, frames are thinned out, so that important frames may be missed. In addition, when frames are thinned out, the interval between frames increases, and the amount of compression-encoded data further increases.

  Since the video signal encoding device disclosed in Patent Literature 1 predicts the motion vector of the detection target block based on the motion vector in the neighboring block of the detection target block, when the entire screen changes, The difference between both the detection target and the neighboring blocks becomes large, and the data amount of the entire screen obtained by compressing and coding the difference data of all the blocks becomes enormous.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a moving image compression encoding apparatus that can reduce the amount of compression encoded data even when the entire screen changes in a moving image. And

  A moving image compression coding apparatus according to the present invention includes a buffer that holds input image data, a subtractor that obtains a difference between the input image data and predicted image data, and obtains predicted difference image data, and the predicted difference image data. A compression unit that compresses to obtain compressed prediction difference image data; a decoding unit that decodes the compressed prediction difference image data to obtain decoded prediction difference image data; and adds the decoded prediction difference image data and the prediction image data An adder that obtains past image data, a frame memory that stores the past image data, a motion detection unit that generates a motion vector based on the past image data and the input image data, and the past image data A motion compensation unit that performs motion compensation processing based on a motion vector to generate the predicted image data, the compressed predicted difference image data, and the motion vector A variable-length encoding unit that variable-encodes the video data, a register that stores a transition vector representing a transition direction and a transition amount of the entire screen, and the past image data Transition image data generating means for obtaining transition image data by converting a transition image obtained by shifting a past image represented based on the transition amount in a direction indicated by the transition vector, and including the frame The memory stores the transition image data instead of the past image data, and the motion detection unit generates the motion vector based on the transition image data and the input image data instead of the past image data. The motion compensation unit performs motion compensation processing based on the motion vector on the transition image data instead of the past image data to generate the predicted image data. The features.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.
<First embodiment>
FIG. 1 is a block diagram showing a moving picture compression encoding apparatus according to the present invention.

  The moving image compression encoding apparatus 100 includes a buffer 110, a subtractor 120, a compression unit 130, a decoding unit 140, an adder 150, a register 160, a VOP motion detection unit 161, an adder 170, a frame, A memory 171, a motion detector 180, a motion compensator 181, and a variable length encoder 190 are included.

  The buffer 110 has a function of holding input image data. The buffer 110 receives, for example, image data obtained by imaging by an imaging unit (not shown) via a video interface (not shown), and holds this as input image data.

  The subtractor 120 obtains predicted difference image data by taking the difference between the input image data held in the buffer 110 and the predicted image data from the motion compensation unit 181, and gives this to the compression unit 130. Normally, when performing motion compensation processing, input image data is handled in units of macro blocks of 16 × 16 pixels. The subtracter 120 generates prediction difference image data for each 8 × 8 pixel block obtained by dividing a 16 × 16 pixel macroblock into four in accordance with the MPEG-4 standard.

  The compression unit 130 includes a DCT unit 131 and a quantization unit 132, and compresses the prediction difference image data from the subtracter 120 to obtain compressed prediction difference image data. A DCT (Discrete Cosine Transform) unit 131 performs a normal discrete cosine transform process on the prediction difference image data from the subtractor 120 for each block to obtain a DCT transform coefficient. The quantization unit 132 performs normal quantization processing for each obtained block on the obtained DCT transform coefficient to obtain a quantized DCT transform coefficient. Note that the above-described compression is a general term for discrete cosine transform processing and quantization processing, and the above-described compressed prediction difference image data corresponds to a quantized DCT transform coefficient.

  The processing of the compression unit 130 described above is a case of inter coding, but in the case of intra coding, the following processing is performed. The DCT unit 131 takes out the input image data held in the buffer 110 and performs normal discrete cosine transform processing on each block to obtain DCT transform coefficients. The quantization unit 132 performs normal quantization processing for each obtained block on the obtained DCT transform coefficient to obtain a quantized DCT transform coefficient.

  The decoding unit 140 includes an inverse quantization unit 141 and an inverse DCT unit 142, and decodes compressed prediction difference image data, that is, quantized DCT transform coefficients, to obtain prediction difference image data. The inverse quantization unit 141 performs inverse quantization processing on the quantized DCT transform coefficient to obtain a DCT transform coefficient. The inverse DCT unit 142 performs inverse discrete cosine transform processing on the obtained DCT transform coefficient to obtain predicted difference image data. Note that the above-described decoding is a general term for inverse quantization processing and inverse discrete cosine transform processing.

  The adder 150 adds the prediction difference image data from the decoding unit 140 and the prediction image data from the motion compensation unit 181 to obtain past image data.

  The register 160 stores in advance a transition vector representing the transition direction and transition amount of the entire screen and rotation angle information representing the rotation angle of the entire screen. The register 160 receives, for example, a transition vector and rotation angle information from an MCU (Micro Controller Unit) (not shown) via an MCU interface (not shown) and stores them. The transition vector is a vector that represents, for example, the transition direction and the transition amount of the entire imaging screen of the imaging unit (not shown). The rotation angle information is information representing the rotation angle of the entire imaging screen of the imaging unit, for example.

  Here, for example, it is assumed that the imaging unit is a surveillance camera. Usually, the surveillance camera pans (ie, shakes the camera left and right) or tilts (ie, shakes the camera up and down) by a preset amount in a preset direction. The register 160 can store these directions and quantities as transition vectors. Further, the surveillance camera may rotate by a preset angle. In this case as well, the register 160 can store these angles as rotation angle information.

  The VOP motion detection unit 161 causes the adder 170 to generate transition image data. The VOP motion detection unit 161 causes the adder 170 to indicate the past image represented based on the past image data from the adder 150 in the direction indicated by the transition vector stored in the register 160. The shift is performed by the amount of transition, and the rotation is performed by the rotation angle represented by the rotation angle information stored in the register 160. The adder 170 gives the transition image data obtained by these processes to the frame memory 171.

  The frame memory 171 stores the transition image data from the adder 170 in units of frames. When transition image data is not generated by the VOP motion detector 161 and the adder 170, the past image data from the adder 150 is stored.

  The motion detection unit 180 generates a motion vector according to normal motion vector generation processing based on the input image data from the buffer 110 acquired via the VOP motion detection unit 161 and the transition image data extracted from the frame memory 171. To do. When transition image data is not generated by the VOP motion detection unit 161 and the adder 170, the motion detection unit 180 generates a motion vector based on the input image data and past image data. The motion detection unit 180 gives the motion vector to the motion compensation unit 181.

  The motion compensation unit 181 performs normal motion compensation processing based on the motion vector acquired from the motion detection unit 180 on the transition image data from the frame memory 171 to generate predicted image data. When transition image data is not generated by the VOP motion detection unit 161 and the adder 170, the motion compensation unit 181 performs normal motion compensation processing on the past image data from the frame memory 171 to generate a predicted image. Generate data. The motion compensation unit 181 gives the predicted image data to the subtracter 120 and the adder 150.

  The variable length coding unit 190 performs variable length coding on the quantized DCT transform coefficient (compressed prediction difference image data), the motion vector, and the rotation angle information. The variable-length encoded data obtained by the variable-length encoding is VOP (Video Object Plane). VOP is a basic unit of video data handled by MPEG-4. When the image has a rectangular shape, the VOP corresponds to a frame or a field. Hereinafter, a VOP obtained by intra coding is referred to as I-VOP (intra-VOP), and a P-VOP (Predicted VOP) obtained by inter coding. The variable length encoding unit 190 supplies variable length encoded data to an external circuit such as an MCU (not shown) via an MCU interface (not shown), for example.

  FIG. 2 is a block diagram illustrating the moving image compression encoding apparatus 100 together with the imaging unit 210 and the like. The moving image compression encoding apparatus 100 is connected to the imaging unit 210 via a video I / F 220. The moving image compression encoding apparatus 100 receives image data obtained by imaging by the imaging unit 210 via the video I / F 220. The image data is held as input image data in a buffer (buffer 110 shown in FIG. 1) in the moving image compression encoding apparatus 100. The moving image compression encoding apparatus 100 starts compression encoding of the input image data when the input image data of one whole image is stored in the buffer.

  The moving image compression encoding apparatus 100 is connected to the MUC 240 via the MUCI / F 230. The moving image compression encoding apparatus 100 receives the transition vector and the rotation angle information from the MUC 240, and stores them in a register (the register 160 illustrated in FIG. 1). The moving image compression encoding apparatus 100 receives an instruction from the MUC 240 via the MUCI / F 230, performs an operation according to the instruction, issues an interrupt signal to the MUC 240 at the end of the image compression encoding process, and performs the MUCI / F 230. The variable-length encoded data is transmitted to the MUC 240 via.

  FIG. 3 is a flowchart showing an encoding processing routine by the moving image compression encoding apparatus 100. First, the buffer 110 receives an input image from the imaging unit 210 via the video I / F 220 and stores it (step S301). Subsequently, the register 160 receives the transition vector and the rotation angle information from the MUC 240 via the MUCI / F 230 and stores them (S302). When the input image data of the entire image is stored in the buffer 110, the moving image compression encoding apparatus 100 starts compression encoding of the input image data (S303).

  FIG. 4 is a flowchart showing an image compression processing routine. A control unit (not shown) makes the following determination. First, it is determined whether or not the input image data held in the buffer 110 should be intra-coded (S401). When intra coding is to be performed, the compression unit 130 and the variable length coding unit 190 perform normal intra coding processing on the input image data to generate an I-VOP (S402). When the input image data held in the buffer 110 is not intra-encoded, that is, inter-encoded, the control unit (not shown) determines whether the input vector is (0, 0) (S403) and the input rotation angle. Is 0 degree (S404). When the input vector is (0, 0) (S403) and the input rotation angle is 0 degree (S404), the compression unit 130 and the variable length coding unit 190 perform normal inter coding on the input image data. Processing is performed to generate a P-VOP (S405). Otherwise, the process proceeds to the VOP motion detection process routine (S406).

  FIG. 5 is a flowchart showing a VOP motion detection processing routine. FIG. 6 is a flowchart showing an inter encoding processing routine after the VOP motion detection processing. FIG. 7 is a diagram illustrating various images and motion vectors. Hereinafter, a description will be given with reference to FIGS.

  First, in the VOP motion detection process shown in FIG. 5, the VOP motion detection unit 161 causes the adder 170 to change the past image 601 from the adder 150 into the input vector and rotation angle stored in the register 160. The transition image 602 is generated by shifting in the indicated direction (S501). Here, it is assumed that the transition vector is a vector representing a shift in the right direction such as (+10, 0). Here, in order to simplify the explanation of the operation, the rotation angle is set to 0 degree. The transition vector and the rotation angle are obtained from the rotation direction and speed of the imaging unit 210, and are determined to match the transition direction and transition amount of the entire screen from the past image 601 to the input image 603. . As a result of the shift, as shown in FIG. 6, the transition image 602 is an image obtained by shifting the past image 601 in the right direction as a whole. The adder 170 gives the transition image 602 to the frame memory 171.

  Next, the process proceeds to the inter encoding process shown in FIG. In the processing, the motion detection unit 180 is based on the input image 603 from the buffer 110 acquired via the VOP motion detection unit 161 and the transition image 602 extracted from the frame memory 171 according to a normal motion vector generation process. A motion vector 604 is generated for each macroblock (S601). As shown in FIG. 6, the input image 603 is an image in which the past image 601 has shifted to the right as a whole. Here, the transition amount of the entire screen from the past image 601 to the input image 603 matches the transition amount of the entire screen from the past image 601 to the transition image 602 (that is, the transition amount of the transition vector). To do. The motion vector 604 obtained at this time is (0, 0) in each macroblock.

  The motion compensation unit 181 performs normal motion compensation processing based on the motion vector 604 acquired from the motion detection unit 180 on the transition image 602 from the frame memory 171 to generate a predicted image 605 (S602). As described above, since the motion vector 604 is (0, 0) in each macroblock, the screen of the transition image 602 and the screen of the prediction image 605 are the same.

  The subtractor 120 obtains a difference between the input image 603 held in the buffer 110 and the predicted image 605 from the motion compensation unit 181 for each macroblock, and obtains a predicted difference image 606, which is supplied to the compression unit 130. Give (S603). The compression unit 130 performs normal discrete cosine transform processing and quantization processing on the prediction difference image 606 for each macroblock to obtain quantized DCT transform coefficients (S604). Finally, the variable length coding unit 190 performs variable length coding including the transition vector and the rotation angle information in addition to the quantized DCT transform coefficient and the motion vector to obtain a P-VOP (S605). As described above, since the motion vector 604 is (0, 0) in each macroblock, the amount of compressed data of the motion vector in each macroblock is reduced during the encoding.

  As described above, the moving image compression coding apparatus according to the present embodiment generates a transition image by shifting a past image in the direction of a transition vector and a rotation angle stored in advance, and based on the transition image and the input image. To generate a motion vector. The transition direction and amount of the transition vector (the transition direction and amount of the entire screen from the past image to the transition image) are set in advance to match the transition direction and amount of the entire screen from the past image to the input image. Thus, the motion vector obtained for each macroblock is (0, 0). Alternatively, the numerical value representing the motion vector can be made small even when they do not completely match. Therefore, the amount of compressed data of the motion vector of each macroblock can be reduced.

The moving image compression encoding apparatus according to the present embodiment encodes a transition vector and a rotation angle representing a transition direction and amount of the entire screen together with a quantized DCT transform coefficient. Rather than compressing and encoding all motion vectors for each macroblock (each motion vector having a relatively large value), the motion vector value of each macroblock is reduced as a whole as in this embodiment, and the screen The amount of data after the compression encoding can be reduced by compressing and encoding the transition vector and the rotation angle representing the entire transition direction and amount. Thus, according to the moving image compression encoding apparatus of the present embodiment, the amount of compression encoded data can be reduced even when the entire screen changes in the moving image.
<Second Embodiment>
The moving image compression encoding apparatus 100 according to the present embodiment is represented by the block diagram of FIG. 1 as in the first embodiment.

  FIG. 8 is a flowchart showing a VOP motion detection processing routine by the VOP motion detector 161. FIG. 9 is a diagram illustrating an example of an image generated by the VOP motion detection process. Hereinafter, the VOP motion detection process will be described with reference to FIGS.

  First, the VOP motion detection unit 161 shifts the past image 900 from the adder 150 in the direction indicated by the transition vector and the rotation angle stored in the register 160 to generate the center image 901 (S801). Subsequently, the VOP motion detection unit 161 evaluates the image quality of the central image 901. Examples of the evaluation method include PSNR (Peak Signal to Noise Ratio), SSIM (Structural Similarity), and VQM (Video Quality Metric). Here, evaluation is performed using PSNR.

  PSNR is one of indexes generally used in moving image evaluation, and is an index indicating how close an image subjected to processing such as compression is to the original image. Usually, the unit is db (decibel). It is said that the closer it is to 50 db, the closer it is to the original image, and generally it is a beautiful image if it is about 40 db. The VOP motion detection unit 161 calculates the PSNR of the central image and stores it (S802).

  Subsequently, the VOP motion detection unit 161 shifts the central image 901 by predetermined pixels in the vertical and horizontal directions, and generates four peripheral images 902, 903, 904, and 905 (S803). The VOP motion detection unit 161 calculates the PSNR of each of the surrounding images 902, 903, 904, and 905 and stores it (S804).

  The VOP motion detection unit 161 compares the PSNRs of the central image 901 and the peripheral images 902, 903, 904, and 905 to determine whether the PSNR value of the central image 901 is the largest (S805). Here, it is assumed that the PSNR value of the peripheral image 903 is the largest. In this case, the VOP motion detection unit 161 sets the peripheral image 903 as the central image 903 (S806).

  Subsequently, the VOP motion detection unit 161 shifts the central image 903 by a predetermined number of pixels in the vertical and horizontal directions to generate four peripheral images 906, 907, 908, and 909 (S803). The VOP motion detection unit 161 calculates the PSNR of each of the surrounding images 906, 907, 908, and 909 and stores it (S804).

  The VOP motion detection unit 161 compares the PSNRs of the central image 903 and the peripheral images 906, 907, 908, and 909 to determine whether the PSNR value of the central image 903 is the largest (S805). Here, it is assumed that the PSNR value of the central image 903 is the largest. In this case, the VOP motion detection unit 161 sets the central image 903 as the transition image 903 (S807).

  Finally, the VOP motion detection unit 161 generates a VOP transition vector 910 based on the past image 900 and the transition image 903 (S808). The VOP transition vector 910 is a vector that represents the shift direction and amount of the entire screen from the past image 900 to the transition image 903. The VOP motion detection unit 161 sets the VOP transition vector 910 in the register 160 and supplies it to the adder 170. The adder 170 shifts the past image 900 from the adder 150 in the direction indicated by the VOP transition vector 910 by the amount of transition indicated by the VOP transition vector 910 to generate a transition image 903, which is generated by the frame memory 171. To give.

  The compression process after the transition image 903 is generated is the same as in the first embodiment. After the compression processing, the variable length coding unit 190 performs variable length coding including the VOP transition vector and the rotation angle information in addition to the quantized DCT transform coefficient and the motion vector to obtain a P-VOP.

  As described above, the moving image compression encoding apparatus according to the present embodiment shifts the past image in the direction indicated by the transition vector and the rotation angle, generates a central image, and further creates a plurality of peripheral images. Subsequently, the moving image compression coding apparatus evaluates the image quality of the central image and the peripheral image. If the image quality of the peripheral image is better than that of the central image, the peripheral image is set as a new central image. The moving image compression encoding apparatus repeats the above processing until the image quality of the central image is the best, and the best central image is used as the transition image. Thereby, the consistency between the input image and the transition image is further increased, and the amount of compressed data of the motion vector of each macroblock can be further reduced as compared with the first embodiment.

  In this embodiment, a VOP transition vector representing the shift direction and amount of the entire screen from the past image to the transition image is generated based on the transition image and the past image generated in this way. The moving image compression encoding apparatus receives a P-VOP by generating a P-VOP by performing variable length encoding including a VOP transition vector and a rotation angle information in addition to a quantized DCT transform coefficient and a motion vector. This device can correctly decode this.

  The first and second embodiments are examples in which compressed data is reduced by using past images, but if a future image is used instead of a past image, B-VOP (Bi-directional predicted VOP) compression is performed. Data can be reduced.

It is a block diagram showing a moving image compression encoding apparatus. It is a block diagram showing a moving image compression encoding apparatus with an imaging part etc. It is a flowchart showing an encoding process routine. It is a flowchart showing an image compression process routine. It is a flowchart showing a VOP motion detection processing routine. It is a flowchart showing the inter coding process routine after a VOP motion detection process. It is a figure showing various images and motion vectors. It is a flowchart showing a VOP motion detection processing routine. It is a figure showing an example of the image produced | generated by the VOP motion detection process routine.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Moving image compression encoding apparatus 110 Buffer 120 Subtractor 130 Compression part 131 DCT part 132 Quantization part 140 Decoding part 141 Inverse quantization part 142 Inverse DCT part 150 Adder 160 Register 161 VOP motion detection part 170 Adder 171 Frame memory 180 Motion Detection Unit 181 Motion Compensation Unit 190 Variable Length Coding Unit 210 Imaging Unit 220 Video I / F
230 MUCI / F
240 MUC

Claims (10)

A buffer to hold the input image data;
A subtractor that obtains predicted difference image data by taking the difference between the input image data and the predicted image data;
A compression unit that compresses the prediction difference image data to obtain compressed prediction difference image data;
A decoding unit that decodes the compressed prediction difference image data to obtain decoded prediction difference image data;
An adder that adds the decoded prediction difference image data and the prediction image data to obtain past image data;
A frame memory for storing the past image data;
A motion detector that generates a motion vector based on the past image data and the input image data;
A motion compensation unit that generates the predicted image data by performing a motion compensation process based on the motion vector on the past image data;
A variable-length encoding unit that variable-encodes the compressed prediction difference image data and the motion vector;
A register for storing a transition vector representing a transition direction and a transition amount of the entire screen;
Transition image data generating means for obtaining transition image data by converting a transition image obtained by shifting the past image represented based on the past image data in the direction indicated by the transition vector by the amount of transition; Including
The frame memory stores the transition image data instead of the past image data,
The motion detection unit generates the motion vector based on the transition image data and the input image data instead of the past image data,
The moving image compressing and coding apparatus, wherein the motion compensation unit generates the predicted image data by performing a motion compensation process based on the motion vector on the transition image data instead of the past image data. The register further stores rotation angle information representing the rotation angle of the entire screen,
2. The transition image data generation unit obtains the transition image data by converting a past image obtained by further rotating the past image by a rotation angle represented by the rotation angle information into data. The moving image compression encoding apparatus described in 1. It further includes an image capturing unit that captures the input image data,
2. The moving image compression encoding apparatus according to claim 1, wherein the transition vector represents a transition direction and a transition amount of the entire imaging screen of the imaging unit. It further includes an image capturing unit that captures the input image data,
The moving image compression coding apparatus according to claim 2, wherein the rotation angle information represents a rotation angle of an entire imaging screen of the imaging unit. Input image data dividing means for dividing the input image data into a plurality of blocks,
The transition image data generation means generates the transition image data for each block,
2. The moving image compression encoding apparatus according to claim 1, wherein the motion detection unit generates the motion vector for each block. A central image generating means for obtaining a central image by shifting the past image represented based on the past image data by the amount of the transition in the direction indicated by the transition vector;
Peripheral image generation means for obtaining at least one peripheral image by shifting the central image by a predetermined pixel in at least one of the upper, lower, left and right directions;
Image quality evaluation means for evaluating image quality for each of the central image and the peripheral image;
2. The image selection unit according to claim 1, further comprising: an image selection unit that sets the image having the best image quality among the central image and the peripheral image evaluated by the image quality evaluation unit as the transition image. A moving image compression coding apparatus.   7. The moving image compression encoding apparatus according to claim 6, wherein the central image generation unit generates the central image by further rotating the past image by a predetermined rotation angle. The image selecting means sets the peripheral image as a new central image only when the image having the best image quality among the central image and the peripheral image evaluated by the image quality evaluating means is the peripheral image. age,
The peripheral image generating means generates the at least one new peripheral image by shifting the new central image by a predetermined pixel in at least one of the upper, lower, left and right directions,
7. The moving image compression coding apparatus according to claim 6, wherein the image quality evaluation unit evaluates image quality for each of the new central image and peripheral image. New transition vector generation means for generating a new transition vector based on the past image represented based on the past image data and the transition image obtained by the image quality evaluation means;
7. The moving picture compression encoding apparatus according to claim 6, further comprising new transition vector variable length encoding means for variable length encoding the new transition vector.   The moving image compression coding apparatus according to claim 6, wherein the image quality evaluation unit evaluates the image quality by any one of PSNR evaluation, SSIM evaluation, and VQM evaluation.

Images