Moving image coder and moving image decoder

The present invention relates to a moving image coding device and a moving image decoding device having a motion compensation function.

As a coding method for moving image data, motion compensation prediction inter-frame coding is known. In motion compensation prediction inter-frame coding, a motion vector representing the "motion" of an image element between frames is detected in the coding apparatus. In addition, the detected motion vector is used to predict the image of the original frame from the past frame (or from the past and future frames), and the difference between the actual image and the predicted image is detected. Then, the motion vector information and the difference value information thereof are transmitted. The decoding device reproduces the moving image by using the motion vector information and the difference value information.

In motion compensation prediction inter-frame coding, once an error occurs, the error propagates to subsequent frames. Therefore, in motion compensation prediction inter-frame coding, intra-frame coded pictures are usually inserted periodically. Here, the in-frame encoded picture is encoded independently of other frames. Therefore, even if an error occurs, the error does not propagate to the frame after the in-frame coded picture.

However, the in-frame coded picture has a significantly larger amount of information than the inter-frame coded picture. Therefore, if the in-frame coded picture is inserted periodically, the peak value of the traffic becomes high. And, in order to guarantee this traffic, it is necessary to increase the buffer size.

As one of the techniques for solving this problem, a coding method called sequential refresh has been proposed. The sequential refresh method will be described with reference to FIG. Note that "refresh" means performing in-frame coding. Further, in the following description, it is assumed that each frame is composed of four regions 501 to 504.

As shown in FIG. 1, in the nth frame, the image of the region 501 is encoded by the intra-frame coding, and the images of the regions 502 to 504 are encoded by the inter-frame coding. Subsequently, in the n + 1th frame, the image of the region 502 is encoded by the intra-frame coding, and the image of the regions 501, 503, 504 is encoded by the inter-frame coding. Similarly, in the n + 2 frame, the image in region 503 is encoded by in-frame coding, and in the n + 3 frame, the image in region 504 is encoded by in-frame coding. In this way, in the example shown in FIG. 1, the entire area is refreshed with 4 frames as a cycle. The sequential refresh method is described in, for example, Patent Documents 1 to 3.<patcit num="1"><text>Japanese Unexamined Patent Publication No. 2003-179938</text></patcit><patcit num="2"><text>Japanese Patent Application Laid-Open No. 6-113286</text></patcit><patcit num="3"><text>Japanese Patent Application Laid-Open No. 2005-260936</text></patcit>

By the way, in motion compensation prediction inter-frame coding that introduces a sequential refresh method, in order to suppress error propagation or realize a "heading playback function", the area that can be referred to for motion compensation is limited. There is a need. Hereinafter, the limitation of the reference area will be described with reference to FIGS. 2 to 4. In FIGS. 2 to 4, regions 501, 502, 503, and 504 are in order in the nth frame, the n + 1th frame, the n + 2th frame, and the n + 3th frame, respectively, as in FIG. It shall be refreshed to. In this case, in the nth frame, the areas 502 to 504 have not been refreshed. That is, areas 502 to 504 are areas where refreshing has not been completed. In the n + 1th frame, the refresh of the area 501 is completed, and the refresh of the areas 503 to 504 is not completed. Similarly, in the n + 2th frame, the areas 501 to 502 are the refresh completed areas, and the areas 504 are the refresh uncompleted areas. In the n + 3rd frame, areas 501 to 503 are refresh completion areas.

When encoding an image by inter-frame coding, for example, the image of the previous frame is referred to. Here, when encoding the image of the unrefreshed region, the image of any region can be referred to. That is, in FIG. 2, references 511 and 512 are allowed. However, in order to suppress the propagation of errors, it is not possible to refer to the image in the refresh incomplete region when encoding the image in the refresh complete region. That is, in FIG. 2, references 513 and 514 are allowed, but reference 515 is not.

Further, in order to realize "head-to-head reproduction" of a moving image, it is not possible to refer to an image in a refresh-unfinished region when encoding an image in a refresh-completed region. For example, in the example shown in FIG. 3, the image of the refresh area or the refresh completed area is referred to (references 521 to 523). Therefore, in this case, the reproduction of the moving image can be started from the n + 3rd frame. On the other hand, in the example shown in FIG. 4, the image of the refresh-unfinished area is referred to when encoding the image of the refresh-completed area (see 524). In this case, since the n + 1th frame cannot be decoded, the n + 2nd frame and the n + 3rd frame cannot be reproduced as a result. That is, the reproduction of the moving image cannot be started from the n + 3rd frame.

FIG. 5 is a diagram illustrating problems to be solved in the sequential refresh method. Here, it is assumed that the image of block A in the n + 1 frame is encoded with reference to the image in the nth frame. Further, it is assumed that block B and block C are detected as image candidates to be referred to by block A. Here, block B does not include an image of the area where refreshing has not been completed. On the other hand, block C contains an image of an unrefreshed area.

In such a situation, according to the prior art (for example, the technique described in Patent Document 1), the reference of the block C is prohibited to suppress the propagation of the error. However, if the image of block C is preferable to the image of block B as the image to be referred to for motion compensation of block A, the prior art uses the image of block B for encoding / decoding. If the image is changed, the image will be deteriorated.

As described above, in the conventional motion compensation prediction frame-to-frame coding, when the sequential refresh is introduced, the image may be deteriorated. That is, in the conventional motion compensation prediction inter-frame coding, it is difficult to realize both suppression of the peak of the amount of information and good image quality.

An object of the present invention is to obtain good image quality while suppressing a peak of the amount of information in motion compensation prediction inter-frame coding. The moving image coding apparatus of the present invention has a configuration in which each of a plurality of blocks obtained by dividing an input image is coded, and an effective region that can be referred to in a reference image for interframe coding is defined. The definition means to be defined, the detection means for detecting the reference area to be referred to for interframe coding of the target block in the reference image, and the image of the reference area for the reference area belonging to the effective area are output. A predictive image generation means that generates a predictive image for inter-frame coding by outputting a complementary image for the reference region that does not belong to the effective region, and an input image is encoded by using the predictive image. It has a coding means, which is to be used.

In the moving image coding device having the above configuration, when the reference area for inter-frame coding includes pixels outside the effective area, when the prediction image is created, the image outside the effective area is replaced. A complementary image is used for. Therefore, error propagation is suppressed.

The predictive image generation means may include complementary means for generating a complementary image based on the pixels in the effective region. In this case, the complementing means may generate the complementing image by a complementing method selected from a plurality of complementing methods prepared in advance.

The moving image decoding device of the present invention has a configuration for decoding coded data obtained by a moving image coding device that encodes each of a plurality of blocks obtained by dividing an input image, and is capable of decoding between frames. An acquisition means for acquiring effective area information that defines an effective area that can be referred to in a coded reference image, and a detection means for detecting a reference area to be referred to for interframe coding of a target block in the reference image. By outputting an image of the reference area for the reference area belonging to the effective area and a complementary image for the reference area not belonging to the effective area, a predicted image for inter-frame coding can be obtained. It has a predictive image generating means for generating and a decoding means for decoding the coded data using the predicted image.

According to the present invention, in motion compensation prediction inter-frame coding, good image quality can be obtained while suppressing the peak of the amount of information.

<figref num="1">It is a figure explaining the sequential refresh method.</figref><figref num="2">It is a figure explaining the limitation of a reference area.</figref><figref num="3">It is a figure explaining the heading of a moving image.</figref><figref num="4">It is a figure explaining the problem about the heading of a moving image.</figref><figref num="5">It is a figure explaining the problem to be solved in the sequential refresh method.</figref><figref num="6">It is a figure which shows the structure of the moving image coding apparatus of embodiment of this invention.</figref><figref num="7">It is a figure which shows the structure of the prediction image generation part provided in the moving image coding apparatus.</figref><figref num="8">It is a figure explaining the detection of a reference area.</figref><figref num="9">It is a figure explaining the effective domain information.</figref><figref num="10">It is a figure (the 1) explaining the operation of the prediction image generation part.</figref><figref num="11">It is a figure (2) explaining the operation of the prediction image generation part.</figref><figref num="12">It is a figure explaining the 1st complementation method.</figref><figref num="13">It is a figure explaining the 2nd complement method.</figref><figref num="14">It is a figure explaining the 3rd complement method.</figref><figref num="15">It is a figure which shows the structure of the prediction image generation part which has the function of selecting a complementation method.</figref><figref num="16">It is a figure which shows the structure of the complement part which has the function of selecting a complement method.</figref><figref num="17">It is a figure explaining the notification method of the effective area / non-effective area.</figref><figref num="18">It is a figure which shows the structure of the moving image decoding apparatus of embodiment of this invention.</figref><figref num="19">It is a figure explaining the structure and operation of the prediction image generation part of the Embodiment provided in the moving image decoding apparatus.</figref>

FIG. 6 is a diagram showing a configuration of a moving image coding device according to an embodiment of the present invention. The moving image coding device 100 encodes moving image data for each of a plurality of blocks obtained by dividing the image by using motion compensation prediction. In the inter-frame coding, the sequential refresh method described with reference to FIG. 1 is adopted. The inter-frame coding may be forward prediction or bidirectional prediction.

The prediction error signal generation unit 1 calculates the difference between the original image and the prediction image for each frame (or for each block constituting the frame) and outputs it as a prediction error signal. The predicted image will be described later. The orthogonal conversion unit 2 performs orthogonal conversion on the prediction error signal. The orthogonal transform is, for example, the DCT transform in this example. In the DCT transform, each pixel value is converted into a frequency component, and coefficient data representing each frequency component is generated. The quantization unit 3 quantizes the output signal (coefficient data in this embodiment) of the orthogonal transformation unit 3. The coefficient coding unit 4 performs entropy coding on the quantized coefficient data. The multiplexing unit 5 multiplexes and transmits the encoded coefficient data, the encoded motion vector information, and the encoded control data. The control data will be described later.

The data output from the multiplexing unit 5 is transmitted to the decoding device via the network, for example. Alternatively, the data output from the multiplexing unit 5 is written to the recording device. The multiplexing method is not particularly limited, but is, for example, TDM.

The inverse quantization unit 6 and the inverse orthogonal conversion unit 7 execute conversion processing corresponding to the quantization unit 3 and the orthogonal conversion unit 2, respectively, and reproduce the prediction error signal. The decoded image generation unit 8 generates a decoded image based on the reproduced prediction error signal and the predicted image. This decoded image is an image that would be obtained in a decoding device.

The decoded image storage unit 11 is, for example, a semiconductor memory, and temporarily stores the decoded image generated by the decoded image generation unit 8. At this time, the decoded image may be stored in the decoded image storage unit 11 after the block distortion is removed by the digital blocking filter 12.

The motion vector calculation unit 13 calculates the motion vector of the target block based on the original image and the decoded image stored in the decoded image storage unit 11. The calculation of the motion vector is not particularly limited, but can be realized by a known technique. It should be noted that calculating the motion vector of the target block is substantially the same as detecting the reference area to be referred to by the target block.

The predicted image generation unit 14 generates a predicted image based on the decoded image stored in the decoded image storage unit 11 and the motion vector obtained by the motion vector calculation unit 13. The configuration and operation of the predicted image generation unit 14 will be described in detail later.

The refresh control unit 15 generates a refresh control signal for realizing the sequential refresh described with reference to FIG. The selection unit 16 selects the predicted image or "zero" generated by the predicted image generation unit 14 according to the refresh control signal. At this time, in the region where refreshing is not executed, the predicted image generated by the predicted image generation unit 14 is selected. In this case, the prediction error signal generation unit 1 outputs a prediction error signal representing the difference between the original image and the predicted image. That is, inter-frame coding is performed. On the other hand, "zero" is selected in the area where refresh is executed. In this case, the prediction error signal generation unit 1 outputs the prediction error signal as it is from the original image. That is, in-frame coding is performed.

The motion vector data coding unit 21 encodes the motion vector information representing the motion vector obtained by the motion vector calculation unit 13. The method of encoding the motion vector information is, for example, entropy coding. The control data coding unit 22 encodes the control data generated by the prediction image generation unit 14. The method of encoding the control data is, for example, entropy coding. The control data will be described in detail later.

FIG. 7 is a diagram showing the configuration of the prediction image generation unit 14. The motion vector information and the refresh control signal are given to the prediction image generation unit 14 as described with reference to FIG.

Motion vector information is given to the reference area detection unit 31. The motion vector information is generated by the motion vector calculation unit 13, and represents the motion vector of the target block as shown in FIG. Then, the reference area detection unit 31 detects the position (that is, the coordinates) of the reference area in the reference image to be referred to by the coded symmetric block based on the motion vector information. For example, the coordinates of the four corners of the target block are "(89, 121) (96, 121) (89, 128) (96, 128)", and the motion vector of the target block is "(7, 9)". If so, "(82, 112) (89, 112) (82, 119) (89, 119)" is obtained as the position of the reference area. The reference image is not particularly limited, but is, for example, an image of the frame immediately before the original image. Then, the reference area detection unit 31 gives the extraction unit 32 and the determination unit 34 the reference area information representing the detected reference area.

The extraction unit 32 extracts the reference image from the decoded image storage unit 11, and further extracts the pixel data of the reference area in the reference image based on the reference area information. When the size of the symmetric block is 8 × 8, 64 pixel data are extracted.

A refresh control signal is given to the refresh management unit 33. The refresh control signal is generated by the refresh control unit 15 in order to realize sequential refresh, and indicates an area to be coded in the frame within each frame. Then, the refresh management unit 33 generates effective area information representing an effective area that is allowed to be referred to in other images according to the refresh control signal.

FIG. 9 is a diagram illustrating effective domain information. Here, for the sake of simplicity, as shown in FIG. 9A, the image area of each frame is composed of five areas 41a to 41e, and each area 41a to 41e is refreshed in order. It shall be. That is, in the nth frame, the n + 1th frame, the n + 2nd frame, the n + 3rd frame, and the n + 4th frame, the area 41a, the area 41b, the area 41c, the area 41d, and the area 41e are refreshed, respectively. It shall be.

In the embodiment, the effective area is defined as "the refresh area and the area above the refresh area" as shown in FIG. 9 (b). That is, in the nth frame, the area 41a is the effective area. In the n + 1th frame, the regions 41a and 41b are effective regions. In the n + 2nd frame, the regions 41a to 41c are effective regions. In the n + 3rd frame, the regions 41a to 41d are effective regions. In the n + 4th frame, the regions 41a to 41e are effective regions.

Alternatively, the effective area may be defined as "a refresh area and an area where refresh is performed within a predetermined period" as shown in FIG. 9 (c). In FIG. 9 (c), the predetermined period is 2 frame times. That is, for example, focusing on the n + 2 frame, the region 41b is refreshed in the n + 1 frame, and the region 41a is refreshed in the nth frame. Therefore, the effective region of the n + 2th frame is the regions 41a to 41c. Effective domains are defined for other frames as well. In the following description, an area other than the effective area in the frame may be referred to as an "ineffective area".

The determination unit 34 checks whether or not the reference area detected by the reference area detection unit 31 belongs only to the effective area. That is, the determination unit 34 checks whether the reference region includes only the pixels in the effective region or the pixels in the non-effective region. As described above, the effective area information representing the effective area (and the non-effective area) is generated by the refresh management unit 33 for each frame.

Complementary unit 35 generates a complementary image according to an algorithm described later. The complementary image may be generated by using the decoded image (that is, the reference image) stored in the decoded image storage unit 11, or may be generated independently of the decoded image. The selection unit 36 selects the pixels of the reference region extracted by the extraction unit 32 or the pixels of the complementary image generated by the complement unit 35 according to the determination result by the determination unit 34. That is, the image of the reference area is output for the reference area belonging to the effective area, and the complementary image is output for the reference area belonging to the non-effective area.

The operation of the prediction image generation unit 14 will be described with reference to FIGS. 10 and 11. Here, it is assumed that the coded block is 8 × 8 pixels. In this case, the reference area is also 8 × 8 pixels.

FIG. 10 shows the operation when the reference area includes only the pixels in the effective area. In this case, the prediction image generation unit 14 outputs the image of the reference region in the decoded image as it is as the prediction image for inter-frame coding. That is, the selection unit 35 selects the pixel data in the reference region of the decoded image as the pixel data of the predicted image.

FIG. 11 shows the operation when the reference area includes pixels in the non-effective area. In this example, it is assumed that the 1st to 7th lines of the reference area belong to the effective area and the 8th line belongs to the non-effective area. In this case, the prediction image generation unit 14 outputs an image of the reference area as the prediction image for the reference area belonging to the effective area and a complementary image for the reference area belonging to the non-effective area. That is, the pixel data in the reference area of the decoded image is selected as the pixel data of the 1st to 7th lines of the 8 × 8 block, and the pixel data of the complementary image is selected as the pixel data of the 8th line.

The predicted image generated as described above is sent to the prediction error signal generation unit 1. Then, the difference (that is, the error) between the original image and the predicted image is calculated, and the error is encoded. However, in the refresh area, since intra-frame coding is performed, "zero" is selected by the selection unit 16 instead of the predicted image.

The effective domain information generated by the refresh management unit 33 is encoded by the control data coding unit 22. That is, the effective domain information is transmitted as control data. Next, an example of the method of generating the complementary image will be described. In the following description, as shown in FIG. 12A, the coded block (or the unit of motion prediction) is 16 × 16 pixels. Then, it is assumed that (0,0) to (15, j) belong to the effective domain.

In the first complement method, the pixel data in the non-effective region (that is, the pixel data in the complement image) is obtained by copying the pixel data of the closest pixel in the effective region, as shown in FIG. 12 (b). Will be generated. This process is expressed by the following equation (1). pred (x, y) = pred (x, j) (1) Here, "pred (x, y)" is pixel data of pixels in the non-effective region. Further, "pred (x, j)" is the pixel data of each pixel on the line adjacent to the ineffective region.

The first complement method is realized by the configuration shown in FIG. 12 (c). That is, in order to realize the first complement method, the complement section 35 includes a copy section 41. The copy unit 41 executes the calculation of the above equation (1) for the pixels belonging to the non-effective area among the pixels constituting the reference area.

In the second complement method, the pixel data of each pixel in the ineffective region is generated by averaging the pixel data of the pixels adjacent to the ineffective region in the effective region, as shown in FIG. 13 (a). Will be done. This process is expressed by the following equation (2).

<maths num="1"><img file="JP5437807B2_D0001.tif" /></maths>

The second complement method is realized by the configuration shown in FIG. 13 (b). That is, in order to realize the second complement method, the complement unit 35 includes an effective reference pixel storage unit 42, a boundary pixel selection means 43, and an averaging unit 44. The effective reference pixel storage unit 42 stores the pixel data of the pixels belonging to the effective area among the pixels constituting the reference area based on the effective area information. The boundary pixel selection means 43 selects the pixel data of the pixels on the line adjacent to the non-effective area among the pixels stored in the effective reference pixel storage unit 42. The averaging unit 44 averages the pixel data selected by the boundary pixel selection means 43. That is, the operation of the above equation (2) is executed.

In the third complement method, the pixel data in the non-effective region is generated by filtering the pixel data of the pixels adjacent to the non-effective region in the effective region, as shown in FIG. 14 (a). This process is expressed by the following equation (3).

<maths num="2"><img file="JP5437807B2_D0002.tif" /></maths>

It is assumed that the filter coefficient is "1" in Eq. (3). That is, for example, when complementing in the diagonal 45 degree direction, filtering is performed under the following conditions. w_{x, y} (x-y + j-1) = 0.25 w_{x, y} (x-y + j) = 0.5 w_{x, y} (x-y + j + 1) = 0.25 When "i" does not correspond to any of "x-y + j-1", "x-y + j", and "x-y + j +", "w"_{x, y} (i) = 0 " The third complement method is realized by the configuration shown in FIG. 14 (b). That is, in order to realize the fourth complement method, the complement unit 35 includes an effective reference pixel storage unit 42, a boundary pixel selection means 43, and a filter 45. The effective reference pixel storage unit 42 and the boundary pixel selection means 43 are the same as the second complement method. The filter 45 executes the filtering according to the above equation (3) for the pixel data selected by the boundary pixel selection means 43.

In the fourth complement method, data representing a predetermined color and brightness is generated as pixel data of the complement image regardless of the pixels in the effective region. The prediction image generation unit 14 of the embodiment generates a complementary image by, for example, any of the above-mentioned first to fourth complementary methods. In addition, the prediction image generation unit 14 dynamically selects the optimum method (for example, in block units) from two or more of the first to fourth complement methods, and uses the selected method. A complementary image may be generated.

FIG. 15 is a diagram showing a configuration of a prediction image generation unit having a function of selecting a complement method. In FIG. 15, the reference area detection unit 31, the extraction unit 32, the refresh management unit 33, and the determination unit 34 are as described with reference to FIG. 7.

The complement unit 37 can execute a plurality of complement methods (for example, the first to fourth complement methods described above). The calculation unit 38 selects the optimum method from a plurality of complement methods. Then, the complement unit 37 outputs the complement image generated by the selected complement method. The calculation unit 38 outputs complement method information indicating the selected complement method. The complement method information is encoded by the control data coding unit 23 and sent to the decoding device.

FIG. 16 is a diagram showing a configuration of a complement unit 37 having a function of selecting a complement method. In this embodiment, the complement unit 37 includes first to fourth complement processing units 41a to 41d that generate complementary pixel data by the first to fourth complement methods. The error calculation units 42a to 42d calculate the error between the pixel data of the pixels belonging to the ineffective region in the reference region and the complementary pixel data generated by the complement processing units 41a to 41d, respectively. The pixels belonging to the non-effective region in the reference region are pixels that are replaced by the complementary pixels in the complementary unit 37. The determination unit 43 selects a complement processing unit that generates pixel data having the minimum error from the complement processing units 41a to 41d. The selection unit 44 selects the pixel data generated by the corresponding complement processing unit according to the determination result by the determination unit 43.

The method of selecting the complement method is not limited to the above method. For example, the complementation method may be determined based on the motion vector of the coded block or the motion vector of the peripheral block. In this case, for example, if the motion vector of the target block is small, the complementary image may be generated according to the first complement method, and if the motion vector of the target block is large, the complementary image may be generated according to the second complement method.

Further, as described above, the complementary image is generated when the reference area includes pixels in the ineffective area. In other words, when the reference area does not include the pixels in the non-effective area, it is not necessary to generate the complementary image and it is not necessary to transmit the complementary method information to the decoding device. Therefore, the prediction image generation unit may include a switch 39 that guides the complement method information to the control data coding unit 23 only when the reference region includes pixels in the ineffective region. By introducing this configuration, the amount of information of control data transmitted to the decoding device can be reduced. Whether or not the reference area includes pixels in the non-effective area is determined by the effective area information generated by the refresh management unit 33.

FIG. 17 is a diagram illustrating a method of notifying the effective area / non-effective area. Here, it is assumed that each frame is composed of regions 501 to 505. Further, the original image (encoded frame) refers to the image of the immediately preceding frame. Further, it is assumed that the area 503 is refreshed in the reference image and the area 504 is refreshed in the original image.

In the example shown in FIG. 17, the effective area for each block belonging to the areas 501 to 503 of the original image is only the areas 501 to 503 of the reference image. On the other hand, the effective area for each block belonging to the area 505 of the original image is the entire area of the reference image (that is, areas 501 to 505). That is, the effective area is different for each block.

As described above, the effective domain information indicating the effective domain is transmitted to the decoding device via the control data coding unit 23. At this time, the effective domain information may be generated in block units and transmitted to the decoding device. In addition, a plurality of blocks constituting the original image are grouped into a block belonging to an area (areas 501 to 503) in which the reference area is restricted and a block belonging to the area (area 505) in which the reference area is not restricted, and each group is grouped. The effective domain information may be transmitted to the decoding device. In H.264, processing can be performed in units of slices composed of a plurality of blocks. In this case, the effective domain information may be added to the slice header and transmitted to the decoding device.

The moving image data encoded by the moving image coding device having the above configuration is transmitted to the moving image decoding device and decoded. Alternatively, the encoded moving image data is once recorded on a recording medium, and then read out by a moving image decoding device and decoded.

FIG. 18 is a diagram showing a configuration of a moving image decoding device according to an embodiment of the present invention. This moving image decoding device decodes the coded data generated by the moving image coding device described above and reproduces the moving image. In the following description, it is assumed that the prediction image generation unit 14 included in the moving image coding device generates a complementary image by a predetermined complementary method.

The separation unit 51 separates the received coded data into coefficient data, motion vector data, and control data. The coefficient decoding unit 52 entropically decodes the coefficient data. The entropy decoding by the coefficient decoding unit 52 corresponds to the entropy decoding by the coefficient coding unit 4 of the moving image coding apparatus. The dequantization unit 53 dequantizes the entropy-decoded coefficient data. The inverse quantization by the inverse quantization unit 53 corresponds to the quantization by the quantization unit 3. The inverse orthogonal conversion unit 54 performs inverse orthogonal transformation of the inverse quantized coefficient data. The inverse orthogonal transformation by the inverse orthogonal transforming unit 54 corresponds to the orthogonal transform by the orthogonal transforming unit 2.

The decoded image generation unit 55 reproduces the original image using the predicted image. The reproduced original image is temporarily stored in the decoded image storage unit 56 as a decoded image. At this time, the decoded image may be stored in the decoded image storage unit 56 after the block distortion is removed by the digital blocking filter 57.

The motion vector data decoding unit 58 decodes the received motion vector data. The decoding by the motion vector data decoding unit 58 corresponds to the coding by the motion vector data coding unit 21 of the motion image coding device. This makes it possible to obtain a motion vector for each block. The control data decoding unit 59 decodes the received control data. The decoding by the control data decoding unit 59 corresponds to the coding by the control data coding unit 22. The control data here is effective domain information representing a region that can be referred to in inter-frame coding. The prediction image generation unit 60 generates a prediction image based on the decoding image, the motion vector, and the effective domain information stored in the decoding image storage unit 56. Then, the original image is reproduced in the decoded image generation unit 55 using this predicted image.

Although omitted in FIG. 18, when the decoding target block belongs to the refresh area, the prediction image generation unit 60 outputs zero. In this case, the decoded image generation unit 55 stores the image composed of the pixel data obtained by the inverse orthogonal conversion unit 54 in the decoded image storage unit 56 as the decoded image. That is, in-frame decoding is performed for this region.

FIG. 19 is a diagram illustrating the configuration and operation of the prediction image generation unit 60 of the embodiment. The operation of the prediction image generation unit 60 is basically the same as that of the prediction image generation unit 14 included in the moving image coding device. However, the prediction image generation unit 60 is provided in the moving image decoding device, and the effective domain information is given by the coding device.

The complementary pixel generation unit 71 acquires the motion vector of the block to be decoded, and extracts the image of the reference area pointed to by the motion vector from the decoded image storage unit 56. The determination unit 72 uses the effective area information to determine whether or not the reference area acquired by the complementary pixel generation unit 71 includes pixels in the non-effective area. When the pixel in the reference area belongs to the effective area, the selection unit 73 selects the pixel data read from the decoded image storage unit 56. On the other hand, when the pixel in the reference region belongs to the ineffective region, the selection unit 73 selects the pixel data of the complementary image generated by the complementary pixel generation unit 71.

The complementary pixel generation unit 71 generates pixel data of the complementary image by the same method as the complementary method in the coding apparatus. Further, when the complement method to be executed for each block is dynamically selected from the plurality of complement methods in the coding apparatus, the complement method information indicating the selected complement method is given to the prediction image generation unit 60. Be done. Then, the prediction image generation unit 60 generates a complementary image by the corresponding complementary method according to the given complementary method information.

As described above, since the moving image coding device and the moving image decoding device of the embodiment employ the sequential refresh method, the peak of the information amount of the coded data for each frame is suppressed. Further, since an image in an optimum reference region can be used when performing inter-frame coding, deterioration of image quality can be suppressed. Furthermore, if the reference area contains pixels in the ineffective area, a complementary image generated from the pixel data in the effective area is used instead of the image in the ineffective area, so that an error in the ineffective area occurs. Propagation can be suppressed.

The sequential refresh may be executed for each block, for each one or a plurality of lines, or for other units.<u style="single">Further, in the moving image coding apparatus of the embodiment, the complement unit may generate a complement image by a complement method selected from a plurality of complement methods prepared in advance. In this case, the moving image coding device of the embodiment may include a transmission unit that transmits the complement method information representing the selected complement method to the decoding device. At this time, the transmitting unit may transmit the complement method information to the decoding device only when a part of the reference area extends outside the effective area. Further, the complementary unit included in the moving image decoding device of the embodiment may generate a complementary image by the complementary method used in the moving image coding device.</u>