JP3519441B2 - Video transmission equipment - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture transmission apparatus for transmitting a moving picture signal, and more particularly to an apparatus for performing coding / decoding using motion compensation prediction in block units.

[0002]

2. Description of the Related Art When an image signal is packed in a packet such as a cell in an ATM network and then transmitted, if the cell is discarded, a part of the image signal cannot be decoded until the synchronization of the variable length decoding is restored. Large deterioration occurs. Similarly, if an error occurs in the bit stream on the transmission line, decoding becomes impossible until the synchronization is restored. Therefore, as an error compensating method for reducing image quality deterioration due to cell discard or transmission path error, a method of replacing a block that cannot be decoded on the receiving side with a decoded image signal is being studied. In the following description, the case of cell discard will be described, but the same applies to the case of a transmission path error.

Generally, in the cell discard compensation technique in a moving picture decoding apparatus using motion compensation prediction, an image signal lost due to cell discard is replaced by an image signal stored in a frame memory (or field memory). It is realized by a thing (called concealment).

An example of a conventional cell discard compensation method will be described with reference to FIG. The image signal of the block A that cannot be decoded due to the cell discard in the encoded image is concealed using the motion vector of the decodable block adjacent to the block A (for example, the block B). In particular,
The image signal of the block A ′ in the reference image pointed to by the motion vector of the block B with respect to the block A is motion-compensated to perform concealment. That is, since the correlation of motion vectors between adjacent blocks is generally high,
Degradation of image quality is reduced by performing motion compensation prediction on the block A that cannot be decoded due to cell discard using the motion vector of the block B adjacent thereto.

For example, in the 1992 Image Coding Symposium (PCSJ92), 6-1 "ATM image coding system with cell discard resistance", as shown in FIG. The motion vector candidates (maximum 8 ways) to be used for are set, the weighted majority values of the motion vectors are taken independently in the horizontal and vertical directions, and the concealment is performed by the motion compensation prediction using the motion vector having the largest number. Here, the weight used for the majority decision is set so that the motion vector closer to the block A is more easily selected as shown in FIG. When the correlation between motion vectors is low and sufficient majority voting cannot be performed, the number of motion vector candidates is increased as shown in FIG. In this case, for example, when the correlation between the motion vectors between the blocks 1 to 8 adjacent to the block A is low, but the correlation between the motion vectors between the blocks 9, 11, and 12 is high, the block A is the block 9, Concealment will be performed using either the motion vector 11 or 12. However, since the correlation between the motion vector of the block A and the motion vectors of the blocks 9, 11, and 12 is not necessarily high, if there is no correlation between the motion vectors of the adjacent blocks, It is difficult to select a motion vector to be used for reliable concealment based on the value alone. Furthermore, when there are a plurality of motion compensation prediction modes (for example, MP
EG has forward, backward, and bidirectional prediction. See: Hiroshi Yasuda, "International Standard for Multimedia Coding," Maruzen)
In addition, since the degree of freedom in motion compensation prediction increases, it is difficult to select one candidate from the eight motion compensation prediction candidates by majority voting.

In the 1991 Image Coding Symposium (PCSJ91), 9-2, "Study of ATM Image Coding System", as shown in FIG. As a result, the block A is concealed by detecting the motion vector between the immediately preceding reproduced image and indirectly detecting the motion vector used for the motion compensation prediction of the block A. Will be concealed. However, this method requires a large amount of calculation to detect the motion vector.

Furthermore, when cell discard occurs in the first I picture (image in which all blocks are intra-frame coded) in the sequence, or when cell discard occurs in an image using the image before the scene change after the scene change as the reference image. If it occurs, there is no image data to be used for concealment in the reference image, and thus concealment using motion compensation prediction as described above cannot be performed.

Specifically, when cell discard occurs in the first I picture of the sequence, or when an image before scene change (frame # n-1 in FIG. 22A) is used as a reference image, motion compensation prediction is performed. (Frame #n in FIG. 22A)
When the cell discard occurs in (1), there is no image signal that can be used for concealment (which can be used for motion compensation), so that there is a problem that concealment cannot be performed by motion compensation. Therefore, as shown in FIG. 22B, by using the pixel value in the decodable block adjacent to the block (the block in the shaded area) that cannot be decoded due to the cell discard, the pixels in the undecodable block are used. To conceal by inserting. At this time, the inter-frame correlation between the frame #n and the frame # n + 1 in FIG. 22A is high (when the motion compensation prediction error is almost 0), and the block having a wide band cannot be decoded due to the cell discard. In this case, as shown in FIG.
The block concealed by the pixel interpolation in the above has only a very narrow band component, and the prediction error signal added to this is almost 0, so refer to the above concealed block in the frame # n + 1 on the receiving side. The band of the reproduced image is still only the low frequency signal. Further, when the inter-frame correlation is high also in the subsequent frames, the high frequency component is not restored, and some blocks with a narrow band remain, so that the visual deterioration continues in the subsequent frames. There was a problem. Here, the shaded areas in FIGS. 23A to 23C are signal components.

Also, the first I picture in the sequence,
In the image with the image before the scene change as the reference image, the amount of information generated in the frame is large, so the number of cells of the image data of these frames increases compared to other frames,
There is a problem that is susceptible to cell discard.

On the other hand, in the conventional refresh by the intra slice or the intra column, when performing motion compensation prediction of the block belonging to the refreshed area 2 in FIG. 24, the image signal belonging to the area 1 which is not refreshed as seen from the area 2 Since the search range of the motion vector is limited so as not to refer to, the reliable motion vector cannot be obtained when executing the concealment using the motion vector on the receiving side.

For example, in 1992 Autumn Meeting of the Institute of Electronics, Information and Communication Engineers, D-162, "Study on low-delay interframe predictive coding", as shown in FIG. Lines (shaded areas in FIG. 24) are set periodically (this is called an intra slice).
When the refresh is performed and the rhombic areas surrounded by the intra slices are area 1 and area 2 as shown in FIG. 24 (each includes an intra slice for one cycle on the left side), when the blocks belonging to the area 2 are encoded. In area 1
By limiting the search range of the motion vector so that is not referred to, it is possible to suppress the deterioration of the image quality due to the error that occurred in the area 1 from being propagated to the area 2 and subsequent areas. However, when concealing a block that cannot be decoded due to an error using a motion vector, it cannot be said that the motion vector with a limited search range has a high correlation with the actual motion of the block. However, it cannot be expected that image quality deterioration due to concealment will be reduced. Similarly, instead of forcibly setting the intra-frame coding line in the horizontal direction, forcibly setting the intra-frame coding line in the vertical direction (this is called intra-column) is there. Also, Nikkei Electronics May 10, 1993 issue (no. 580), pp. As shown in 63 to 64, in MPEG2, as a function of error resilience, a motion vector can be added to a macroblock which is intra-frame coded.

[0012]

In the conventional method for compensating for cell discard and transmission path error, the motion compensation prediction suitable for the undecodable block is selected from the motion compensation predictions used in the adjacent blocks. When selecting the candidate, the motion vector used for concealment is selected by comparing only the motion vector values of the surrounding blocks by utilizing the correlation of motion vectors between adjacent blocks. Therefore,
When the correlation of motion vectors between adjacent blocks is small, it is difficult to select an appropriate motion vector, and the performance of cell loss compensation deteriorates.

Further, when cell discard occurs in the first I picture of the sequence (image in which all blocks are intra-frame coded), or in the image using the image before the scene change after the scene change as the reference image, the cell discard occurs. If it occurs, there is no image data to be used for concealment in the reference image, and thus concealment using motion compensation prediction as described above cannot be performed. In addition, since the amount of information generated in a frame is large in the first I picture in the sequence or in an image in which the image before the scene change is used as a reference image, the number of cells of image data in these frames is increased compared to other frames. Then, there is a problem that the cell discard is easily affected.

Further, in the conventional refresh by the intra slice or the intra column, when the motion compensation prediction of the block belonging to the refresh area is performed, the image signal belonging to the area other than the refresh area is not referred to.
Since the search range of the motion vector was limited, a highly reliable motion vector could not be obtained when executing concealment using the motion vector on the receiving side.

The present invention has been made to solve the above problems, and a first object of the present invention is to make a block that cannot be decoded due to a cell discard or a transmission path error of another block adjacent to the block. A moving picture transmission apparatus for motion-compensated prediction using motion-compensated prediction, in which an appropriate motion is achieved even when the motion vector correlation between the undecodable block and the adjacent block is low. It is an object of the present invention to provide a moving image transmission apparatus that enables selection of compensation prediction candidates and makes motion compensation prediction with higher accuracy.

The present invention has been made to solve the above problems, and a second object of the present invention is to provide a first I picture in a sequence or a scene after a scene change using a picture before the scene change as a reference picture. The image is made less susceptible to cell discard and transmission path error, and even when cell discard or transmission path error occurs in the I picture and the image after the scene change, the influence of image deterioration is propagated to the subsequent images. An object is to provide a moving image transmission device that performs difficult encoding / decoding.

The present invention has been made to solve the above problems, and a third object thereof is to perform refresh by intra slice or intra column without limiting the search range of motion vectors. An object of the present invention is to provide a moving image transmission device capable of performing the above.

[0018]

In order to achieve the first object, according to the present invention (claim 1), a moving image for transmitting data obtained by encoding a moving image signal in block units using motion compensation prediction. In a moving image transmission device including a transmitting device and a moving image receiving device that receives the encoded data and decodes the encoded data in block units, the moving image receiving device detects an error in the received encoded data. Error detection means for outputting an error detection signal, identification means for identifying a block that is undecodable due to the error based on the error detection signal, and a plurality of blocks in the vicinity of the block identified as undecodable by the identification means Selecting means for selecting a motion vector and motion compensation method to be applied to the block identified as undecodable from among the motion vectors and motion compensation of Motion compensation prediction means for performing motion compensation prediction on the block identified as undecodable using the selected motion vector and motion compensation method to generate a prediction signal; and for the block identified as undecodable Interpolation prediction means for performing interpolation prediction of pixel values in the block identified as undecodable using a plurality of neighboring pixel values to generate an interpolation prediction signal; Processing means for setting a weighted average value with the interpolated prediction signal as an image signal of the block identified as undecodable.

[0019]

[0020]

In order to achieve the first object, in the present invention (claim 2), a moving image transmitting apparatus for transmitting data obtained by encoding a moving image signal in block units using motion compensation prediction, and the code. In a moving image transmission apparatus including a moving image receiving apparatus for receiving encoded data and decoding the encoded data in block units, the moving image receiving apparatus detects an error in the received encoded data and outputs an error detection signal. Error detection means, identification means for identifying a block that is undecodable due to the error based on the error detection signal, and a plurality of decodable blocks in the vicinity of the block identified as undecodable by the identification means Selecting means for selecting a motion vector and motion compensation method to be applied to the block identified as undecodable from among motion vectors and motion compensation; A prediction means for predicting a low frequency component in the block identified as undecodable by using pixel values in the vicinity of the identified block, and a motion vector and motion compensation method selected by the selection means are used. And a signal for separating the predicted low frequency component and the separated high frequency component from each other. The processing means for converting the image signal of the entire band into the image signal of the block identified as undecodable.

[0022]

[0023]

[0024]

[0025]

According to the present invention (Claim 1), in the concealment method using motion compensation prediction of adjacent blocks, packet discarding or transmission line is performed by using the selected appropriate motion vector and motion compensation method. A prediction signal generated by motion-compensated prediction of a block that cannot be decoded due to an error, and an interpolative prediction generated by interpolating pixel values in the block using a plurality of pixel values in the vicinity of the block The weighted average value with the signal is used as the image signal of the block. As a result, it is possible to reduce block-like distortion that may occur due to discontinuity of low frequency components between adjacent blocks.

[0026]

[0027]

Further, according to the present invention (claim 2), in the concealment method using the motion compensation prediction of adjacent blocks, packet discarding or transmission path is performed by using the selected appropriate motion vector and motion compensation method. A pixel value in the block is interpolated and predicted using a high frequency component of a prediction signal generated by motion compensation prediction of a block that cannot be decoded due to an error and a plurality of pixel values in the vicinity of the block. The image signal of the entire band, which is formed by combining the low-frequency component of the interpolated prediction signal generated as described above, is used as the image signal of the block. As a result, it is possible to reduce block-like distortion that may occur due to discontinuity of low frequency components between adjacent blocks.

[0029]

[0030]

[0031]

[0032]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) First, a moving image transmission apparatus according to the first embodiment of the present invention will be described. FIG. 1 is a block diagram showing a moving image receiving apparatus according to this embodiment.

The moving image transmitting apparatus according to the present embodiment is ITU (formerly CCITT) standard H.264 standard. Any coding device may be used as long as it is a coding device using motion compensation for moving images such as 26X in block units.

In this embodiment, first, a decodable block (A in FIG. 2A) adjacent to a block (X in FIG. 2A) that cannot be decoded due to packet discard or transmission path error.
The motion-compensated prediction (prediction mode, prediction mode such as backward prediction, and motion vector) used in H) is set as a candidate for motion-compensated prediction used for concealment. next,
Each motion compensation prediction candidate is applied to the pixel value in the block adjacent to the block X, the motion compensation prediction error value is calculated for each candidate, and the motion compensation prediction having the smallest error evaluation value is selected from the candidates. Select and conceal block X using this motion compensated prediction.

First, an example of the error detection circuit 100 shown in FIG. 1 will be described with reference to FIG. As shown in FIG. 3A, when the packet is supplied via the line 10, the packet decomposing circuit 101 decomposes the packet, and the bit stream is demultiplexer / variable length decoding circuit 10
2, the packet discard identification information (sequence number in the ATM cell) is detected, and the packet discard information is supplied to the undecodable identification information flag circuit 103. The bit stream supplied to the demultiplexer / variable length decoding circuit 102 is separated into discrete cosine transform (DCT) coefficients and side information, and then variable length decoding is performed, and a signal sent as difference data (for example, a motion vector). ) Is returned to the original shape and then supplied to the decoder 110. At this time, when an error in the syntax of the bitstream such that a code that does not appear in the variable length code table appears, that information is supplied to the undecodable identification information flag circuit 103. In the undecodable identification information flag circuit 103, the packet decomposing circuit 101 and the demultiplexer / variable length decoding circuit 1
An address and a non-decipherable identification information flag are set in a block that cannot be decoded due to the error detected in 02, and the block is output via the line 20.

On the other hand, as shown in FIG. 3B, line 1
When the bitstream is supplied via 0, the demultiplexer / variable length decoding circuit 1 outputs the bitstream in which the correctable error is corrected in the error correction circuit 104.
Supply to 02. Demultiplexer / Variable length decoding circuit 1
The bit stream supplied to 02 is separated into Discrete Cosine Transform (DCT) coefficients and side information, and then variable length decoding is performed, and the signal (for example, motion vector) sent as difference data is returned to the original form. After returning, it is supplied to the decoder 110. At this time, if an error in the syntax of the bit stream such that a code that does not appear in the variable length code table appears due to an error that cannot be completely corrected or is erroneously corrected, An undecipherable identification information flag is set to the address of the block that cannot be decrypted, and is output via the line 20.

In the decoder 110 of FIG. 1, the DCT coefficient supplied from the error detection circuit 100 is the inverse quantization circuit 111.
The side information including the motion vector and the motion compensation prediction mode is supplied to the motion compensation prediction circuit 113 and the motion compensation information memory 120 via the line 30.

In the inverse quantization circuit 111, the inversely quantized DC
The T coefficient is supplied to the inverse DCT circuit 112. On the other hand, the motion compensation prediction circuit 113 reads the reference image signal from the frame memory 130 via the line 40 in accordance with the motion compensation prediction mode for each block, creates a prediction value, and supplies the prediction value to the adder 114. However, the prediction value input to the adder 114 is 0 in the block of the intra-frame coding mode. In the adder 114, the prediction error supplied from the inverse DCT circuit 112 and the prediction value supplied from the motion compensation prediction circuit 113 are added, and the frame memory 1 is added via the line 50.
Supply to 30.

The decoder of this embodiment may have another configuration as long as it includes the motion compensation prediction circuit 113. In the concealment circuit 140, when the block X shown in FIG. 2 is concealed using the information of the blocks A to H, in the motion compensation information memory 120,
For the hatched block in FIG. 2 (a), the line 3
The value of the motion compensation prediction mode and the motion vector supplied via 0 and the non-decodable identification information flag supplied from the error detection circuit 100 via the line 20 are stored. In addition, the frame memory 130 has a structure shown in FIG.
The image signals of the blocks hatched in (a) are stored. That is, the signal of the block X is delayed and then concealed and output until the decoding of the block H is completed and the signal of the block H is stored in the motion compensation information memory 120 and the frame memory 130.
However, it is not necessary to rewrite the motion compensation information and the image data of the block which is supplied with the undecodable identification information flag via the line 20 and is recognized as undecodable.

In the concealment circuit 140, first, in the first motion compensation circuit 141, motion compensation of a block which is decodable in blocks A to H stored in the motion compensation information memory 120 and which is not intraframe coding is performed. A prediction value for a pixel value near the block X is created from the reference image read from the frame memory 130 using the information, and is supplied to the error evaluation value calculation circuit 142.

The motion compensation error evaluation circuit (hereinafter referred to as MC error evaluation circuit) 142 reads the reproduced image of the pixels in the vicinity of the block X from the frame memory 130, and the block X supplied from the first motion compensation prediction circuit 141. After calculating the error evaluation value (sum of absolute values of error signals or sum of squares) of prediction values of neighboring pixels, the error evaluation value is selected by the selection circuit 14
Supply to 3. However, among the pixels near the block X, the motion compensation prediction error of the pixel included in the undecodable block is not evaluated.

In the selection circuit 143, the MC error evaluation circuit 1
The motion-compensated prediction having the smallest error evaluation value supplied from the second motion compensation circuit 14 is selected, the corresponding motion compensation information is read from the motion compensation information memory 120, and the second motion compensation circuit 14 is selected.
Supply to 4.

The second motion compensation circuit 144 reads the motion compensation prediction value (concealment image) of the block X from the frame memory 130 using the most appropriate motion compensation prediction selected by the selection circuit 143. It is created from the signal and supplied to the selector 150. Note that the block adjacent to the block X cannot be decoded,
When the motion compensation information cannot be obtained because the block is not in the motion compensation mode (for example, the intraframe coding mode), for example, the same position as that of the block X in the reference image stored in the frame memory 130 is set. It suffices to read the image data of the block in the block and supply it to the selector 150.

In the selector 150, if the block (block X) can be decoded by the undecodable identification information flag supplied from the motion compensation information memory 120, the block X stored in the frame memory 130.
Is output via the line 60, and if it is a block that cannot be decoded, the concealment image supplied from the second motion compensation prediction circuit 144 is output via the line 60. Here, the coding of the block H is completed, and the information is stored in the frame memory 130 and the motion compensation information memory 1.
After being stored in 20, the image of the block X is delayed and output from the selector 150 via the line 60.

As described above, the error-compensated image signal output via the line 60 is stored in the frame memory 130 as a reference image for the next frame. As the pixels near the block X in the above embodiment, the block A
Without evaluating the motion-compensated prediction error values of all the pixels included in ~ H, a group of pixels (pixel group) close to the block X (having a strong inter-pixel correlation) is used as shown in FIG. The hardware scale can be reduced. However, since the influence of the error propagates in the horizontal direction, when the block X cannot be decoded due to the error, the blocks D and E are also often undecodable. Therefore, the error is evaluated only by the pixels in these blocks. You shouldn't do it.

The following may be considered as another embodiment of the concealment circuit 140. That is,
As shown in FIG. 4, a pixel value interpolation circuit 145 and a mode determination circuit 146 are provided, and an output obtained by motion-compensating and predicting the block X by the second motion compensation circuit 144 and the pixel value interpolation circuit as in the above embodiment. An output obtained by interpolating and predicting the inside of the block X by using pixel values near the block X in 145 is supplied to each selector 147, and an output determined to be appropriate for concealing the block X in the mode determination circuit 146. Is output via the selector 147. Specifically, in the pixel value interpolation circuit 145, for example, the frame memory 13
Pixel values adjacent to the block X of blocks A to H stored in 0 (n pixels in each block when the block size is n × n pixels, n: integer, pixel indicated by diagonal lines in FIG. 2C) ) Is read from the frame memory 130, and a pixel value in the block X is weighted and interpolated by linear combination of pixel values to create an interpolated image. However, the pixel values in the undecodable block are not used when creating the interpolated image of the block X. As the interpolation image creating method for the block X, another interpolation method such as using the average value of surrounding blocks may be used. Further, in the mode determination circuit 146, in a predetermined case, for example, when the motion compensation modes of the decodable blocks stored in the motion compensation information memory 120 are all intraframe coding, the selector 147 selects pixels. A signal for selecting the output of the value interpolation circuit 145 is supplied.

As another embodiment for setting the pixel for evaluating the motion compensation error value, an attribute determination circuit 148 is provided in the concealment circuit 140 as shown in FIG. 5, and the motion compensation error value is evaluated here. The first motion compensation prediction circuit 141 is detected by using a pixel group in a region (for example, edge portion) having a large brightness gradient.
The MC error evaluation circuit 142 may obtain the motion compensation prediction error evaluation value. As a specific example, block A to
Edges in H are detected and motion compensated prediction errors are evaluated using pixels in blocks containing edges. In addition, as in the embodiment of FIG. 4 described above, the second motion compensation prediction circuit 144
May be switched to the output of the pixel value interpolation circuit 145.

(Second Embodiment) Next, a moving image transmission apparatus according to a second embodiment of the present invention will be described. Figure 6
It is a block diagram which shows the moving image receiving apparatus concerning a present Example.

In FIG. 6, in the error detection circuit 100,
As in the first embodiment, the coded data supplied via the line 10 is separated into discrete cosine transform (DCT) coefficients and side information, and then variable length decoding is performed, and the DCT coefficients are inverse quantization circuits. The motion compensation information such as a motion vector (hereinafter referred to as MV) 111 is stored in the first motion compensation prediction circuit 22.
0 and the motion compensation information memory 240. In addition, an address and an undecodable identification information flag are set in the undecodable block and output through the line 20.

In the inverse quantization circuit 111, the inversely quantized DC
The T coefficient is supplied to the inverse DCT circuit 112. The prediction error signal output from the inverse DCT circuit 112 is supplied to the first activity calculation circuit 200, and the value of the activity of the prediction error signal (sum of squares or sum of absolute values, which is an index of the amount of information included in the signal. ) Is calculated. In the second activity calculation circuit 210, the first motion compensation prediction circuit 220 causes the first activity calculation circuit 20 to
A prediction signal of the same block as the block whose activity is calculated at 0 is supplied, and the activity of this signal (sum of squares or sum of absolute values of signals separated from the average value) is calculated.

The reliability judgment circuit 230 is supplied with the activity of the prediction error signal and the activity of the prediction signal calculated by the activity calculation circuits 200 and 210, respectively, and judges the reliability of the MV. For example, even if the power of the prediction error signal is small, if the brightness gradient of the prediction signal is small, there is a possibility that an MV detection error may occur due to noise or the like during MV detection on the transmission side. On the other hand, when the brightness gradient of the prediction signal is large, the reliability of the MV is high unless the power of the prediction error signal is extremely large. Therefore,
Ratio E / of the activity E of the prediction error signal (output of 200) and the activity P of the prediction signal (output of 210)
This value is supplied to the motion compensation information memory 240, assuming that the smaller P is, the higher the reliability of MV is.

When estimating the motion compensation information of the block X shown in FIG. 2 using the motion compensation information of blocks A to H, the motion compensation information memory 240 stores the motion compensation information of FIG.
For the hatched block in (a), the motion compensation prediction mode and motion vector values supplied via the line 30 and the undecodable identification information supplied from the error detection circuit 100 via the line 20. The flag and the reliability of the MV supplied from the reliability determination circuit 230 are stored. If the block (block X) is a decodable block, the motion compensation information of the block is supplied to the second motion compensation prediction circuit 113. If the block is undecodable, the block A The motion compensation information of the block having the most reliable MV among H is supplied to the second motion compensation prediction circuit 113. In addition, in the blocks A to H, the average value of the most reliable MV and the second most reliable MV is calculated as the second motion compensation prediction circuit 113.
May be supplied to

In the second motion compensation prediction circuit 113, the reference image signal is read from the frame memory 130 in accordance with the motion compensation prediction mode supplied from the motion compensation information memory 240 for each block, a prediction value is created, and an adder is created. 11
Supply to 4. However, the prediction value input to the adder 114 is 0 in the block of the intra-frame coding mode.

In the adder 114, when the value of the block X is supplied from the second motion compensation prediction circuit 113, the inverse DCT circuit in the delay circuit 250 is stored until the values of the block H are stored in the motion compensation information memory 240. After delaying the output of 112, the values are added and the frame memory 130
Supply to. However, when the block cannot be decoded, the value of the prediction error signal is reset in the delay circuit 250.

(Third Embodiment) Next, a moving image transmission apparatus according to a third embodiment of the present invention will be described. Figure 7
It is a block diagram which shows the moving image receiving apparatus concerning a present Example.

In FIG. 7, in the error detection circuit 100,
As in the first embodiment, the coded data supplied via the line 10 is separated into discrete cosine transform (DCT) coefficients and side information, and then variable length decoding is performed, and the DCT coefficients are inverse quantization circuits. At 111, motion compensation information such as MV is supplied to the motion compensation information memory 300. Further, an address and an undecodable identification information flag are set in the undecodable block and output via the line 20.

When the motion compensation information of the block X shown in FIG. 2 is estimated by using the motion compensation information of the blocks A to H, the motion compensation information memory 300 stores the motion compensation information of FIG.
For the hatched block in (a), the motion compensation prediction mode and motion vector values supplied via the line 30 and the undecodable identification information supplied from the error detection circuit 100 via the line 20. Flags will be stored. Here, in the motion compensation estimation circuit 310, if the block (block X) is a decodable block, the motion compensation information of the block is stored, and if the block is a non-decodable block, it is stored in the motion compensation information memory 300. Appropriate motion compensation is selected from the motion compensation information of the blocks A to H that are provided, and the motion compensation information is supplied to the motion compensation prediction circuit 113.

In the inverse quantization circuit 111, the inversely quantized DC
The T coefficient is supplied to the inverse DCT circuit 112. The prediction error signal output from the inverse DCT circuit 112 is delayed by the delay circuit 350 as in the second embodiment, and then added by the adder circuit 11.
4 and is added to the prediction signal supplied from the motion compensation prediction circuit 113, and then added to the frame memory 130. Here, the prediction error signal of the block that cannot be decoded in the delay circuit 320 is reset to zero.

In the pixel value interpolation circuit 330, pixel values (blocks adjacent to the blocks X of the blocks A to H stored in the frame memory 130, for example, as in the circuit of FIG. 4 described in the first embodiment. When the size is n × n pixels, n pixels for each block, n: an integer) is read from the frame memory 130, and the pixel values in the block X are weighted and interpolated according to the inter-pixel distance by linear combination of the pixel values. Thus, an interpolated image is created and supplied to the weighted average circuit 350.

In the weighted average circuit 350, the signal supplied from the frame memory 130 and the pixel value interpolation circuit 33 are used.
A weighted average value with the interpolated signal supplied from 0 is calculated based on the signal supplied from the mode determination circuit 340. For example, when the mode determination circuit 340 determines that the block X stored in the motion compensation information memory 300 can be decoded, the frame memory 130.
When it is determined that the block X is undecodable and the motion compensation modes of the blocks decodable among the blocks A to H adjacent to the block X are all intraframe coding, After calculating the average value of the input from the pixel value interpolation circuit 330, and the average value of the input from the frame memory 130 and the input from the pixel value interpolation circuit in the weighted average circuit 350, the line 40 is calculated. It is also output through the frame memory 130.

As described above, when the undecodable block is concealed by the motion compensation, the correlation of the low frequency component with the adjacent block is small, and the block-shaped distortion is generated, the pixel value interpolation is performed. The block-shaped distortion is reduced by taking the average value of the adjacent blocks created in the circuit 330 and the signal in which the block boundaries are smoothly connected.

Further, the concealment circuit 140 shown in FIG. 4 described above is modified as shown in FIG.
6, when it is determined that the motion compensation modes of the decodable blocks among the blocks A to H adjacent to the block X are all intraframe coding, the input from the pixel value interpolation circuit 330 is input. In other cases, the same effect can be obtained by calculating the average value of the input from the frame memory 130 and the input from the pixel value interpolation circuit in the weighted average circuit 350 and then outputting it to the selector 150.

(Fourth Embodiment) Next, a moving image transmission apparatus according to a fourth embodiment of the present invention will be described. Figure 9
It is a block diagram which shows the moving image receiving apparatus concerning a present Example.

In FIG. 9, in the error detection circuit 100,
As in the first embodiment, the coded data supplied via the line 10 is separated into discrete cosine transform (DCT) coefficients and side information, and then variable length decoding is performed, and the DCT coefficients are inverse quantization circuits. At 111, motion compensation information such as MV is supplied to the motion compensation information memory 400. Further, an address and an undecodable identification information flag are set in the undecodable block and output via the line 20.

When estimating the motion compensation information of the block X shown in FIG. 2 using the motion compensation information of the blocks A to H, the motion compensation information memory 300 stores the motion compensation information shown in FIG.
For the hatched block in (a), the motion compensation prediction mode and motion vector values supplied via the line 30 and the undecodable identification information supplied from the error detection circuit 100 via the line 20. Flags will be stored. Here, in the motion compensation estimation circuit 410, the motion compensation information of the block (block X) is stored in the motion compensation information memory 400 if the block (block X) is decodable and in the motion compensation information memory 400 if the block is undecodable. Appropriate motion compensation is selected from the motion compensation information of the blocks A to H that are provided, and the motion compensation information is supplied to the motion compensation prediction circuit 113.

In the inverse quantization circuit 111, the inversely quantized DC
The T coefficient is supplied to the inverse DCT circuit 112. The prediction error signal output from the inverse DCT circuit 112 is delayed by the delay circuit 420 as in the second embodiment, and then added by the adder circuit 11.
4 and is added to the prediction signal supplied from the motion compensation prediction circuit 113, and then added to the frame memory 130. Here, the prediction error signal of the block that cannot be decoded in the delay circuit 420 is reset to zero.

In the low-order coefficient predicting circuit 430, reference is made to H.Su.
n, K.Challapali, J.Zdepski, ERROR CONCEALMENT IN DIGI
TAL SIMULCAST AD-HDTV DECODER ", IEEE Trans.Consumer
Electoronics, Vol.38, No.3, Aug.1992, (see Fig. 10)
As shown in FIG.
Using DC9), the low-order coefficient (AC01, AC10, AC02, AC11, AC20) of the DCT block corresponding to DC5
Is predicted on the condition that the amplitude change on the screen is approximated by a quadric surface. DC5, DC6, DC8, DC
If the macroblock containing 9 cannot be decoded, the value of DC in the adjacent macroblock (DC2, DC3, etc.)
Is used for bilinear interpolation prediction. The dotted line in FIG. 10A indicates the boundary of the DCT block, and the solid line indicates the boundary of the macro block.

On the other hand, in the coefficient separation / synthesis circuit 440, in the mode determination circuit 450, the motion compensation information memory 40
Block A adjacent to block X stored in 0
If it is determined that the motion compensation modes of the decodable blocks in H are all intra-frame coding, only the input from the low-order coefficient prediction circuit 430 is output, otherwise DCT The low-order coefficient predicting circuit 43 supplies the low-order coefficient of the signal, which is supplied from the circuit 460 and is read from the frame memory 130 and converted into the DCT coefficient, to the DCT coefficient.
The signal replaced with the count supplied from 0 is output. The output of the coefficient separation / synthesis circuit 440 is the inverse DCT circuit 470.
After being inversely converted in, the signal is supplied to the selector 480. If the block X is a decodable block, the selector 480 outputs the signal of the block X supplied from the frame memory via the line 40. If the block X cannot be decoded, the signal is supplied from the inverse DCT circuit. The output signal on line 40 is supplied to the frame memory 130 while the output signal on line 40 is output.

As described above, when the undecodable block is concealed by motion compensation, the correlation of the low frequency component with the adjacent block is small, and even if block-like distortion occurs, the low-order coefficient of the DCT is Since the low-order coefficient predicting circuit 430 replaces the adjacent block with a coefficient predicted so that the block boundary portion is smoothly connected, block-shaped distortion is reduced.

Further, the concealment circuit 140 shown in FIG. 4 described above is modified as shown in FIG. 11, and in the coefficient separation / synthesis circuit 440, in the mode determination circuit 146, among the blocks A to H adjacent to the block X. When it is determined that all the motion compensation modes of the decodable block are intra-frame coding, the input from the low-order coefficient prediction circuit 430 is output as it is. In other cases, the DCT circuit 4 is used.
60, the low-order coefficient of the signal obtained by converting the signal of the block X read from the frame memory 130 and converted into the DCT coefficient is replaced with the coefficient supplied from the low-order coefficient prediction circuit 430, and the inverse signal is output. The same effect can be obtained by performing inverse conversion in the DCT circuit 470 and then outputting to the selector 150.

The low-order coefficient predicting circuit 430, the DCT circuit 460, and the inverse DCT circuit 470 may be replaced with a circuit suitable for an orthogonal transform other than DCT or a band division filter.

(Fifth Embodiment) Next, a moving image transmission apparatus according to a fifth embodiment of the present invention will be described. 12
FIG. 3 is a block diagram showing a moving image transmitting apparatus according to this embodiment.

In FIG. 12, the input image is stored in a plurality of frame memories 500. Motion vector detection circuit 5
In FIG. 10, after detecting a motion vector between frames stored in the frame memory 500, a sum of absolute values of error signals (obtained when a motion vector is detected) for each block and a motion vector when the frame compensates for another frame. The scene change detection circuit 52 outputs a signal for identifying whether or not it becomes a reference image.
Supply to 0.

The scene change detection circuit 520 calculates the activity of the input image supplied from the frame memory 500 (for example, the sum of absolute values of AC components), and calculates the sum of absolute values of error signals supplied from the motion vector detection circuit 510. Compare. Here, for example, if the number of blocks in which the latter value is large is more than half of the total number of blocks in a frame, it is determined that a scene change has occurred, and if the frame is a reference image of another frame, The scene change identification signal is converted into a resolution conversion circuit 530 and an encoder 540.
Supply to each.

The embodiment of scene change detection is not limited to the above method, and other methods may be used. In the resolution conversion circuit 530, when the scene change identification signal supplied from the scene change detection circuit 520 is enabled, the image signal whose resolution has been reduced by applying a 0th-order phase low-pass filter to the frame is sent to the encoder 540. Supply.

In the encoder 540, the resolution conversion circuit 5
The image signal supplied from 30 is divided into blocks, the motion vector supplied from the motion vector detection circuit 510 is used to perform motion compensation predictive coding for each block, and then the coded image signal is output.

As another embodiment for reducing the spatial resolution of the frame immediately after the scene change, as shown in FIG. 15, the spatial resolution is not lowered by the resolution conversion circuit 530, but
When the scene change identification signal supplied from the scene change detection circuit 520 is enabled, the encoder 5
In 60, all the blocks of the frame may be intra-frame coded, high-order coefficients may be discarded, and only low-order coefficients as shown in FIG. 10B may be coded to reduce the spatial resolution.

A scene change identification flag (1 bit) may be added for each frame. Similarly, in FIG. 13, in the dynamic range conversion circuit 535 instead of the resolution conversion circuit 530, when the scene change identification signal supplied from the scene change detection circuit 520 is enabled, the dynamic range of the frame is calculated by the following equation. The image signal from which the signal has been dropped is supplied to the encoder 540.

X ′ = α · (x− 128) +128 0 <α <1 Here, x is a pixel value of the frame, and x ′ is a pixel value with a reduced dynamic range.

As another embodiment for reducing the dynamic range, the average value of x may be used instead of 128 in the above equation. This average value may be the average value of the entire frame or the average value of a small area including x.

Further, instead of multiplying the amplitude value by α, the dynamic range of the amplitude value may be limited by a limiter. The region in which the band and the dynamic range are reduced as described above may be the entire frame, or may be only the region in which information is largely lost by spatial interpolation. In the latter case, it is necessary to detect the edge portion or the area where the power of the high frequency component is large, and therefore it is preferable to add the edge portion detection circuit 550 as shown in FIG. In this case, the edge part detection circuit 550 applies an edge detection filter to the signal of the frame supplied from the frame memory 500 to detect the edge part, and outputs a signal for identifying a region including the edge part to the dynamic range conversion circuit 535. Supply. The dynamic range conversion circuit 535 reduces the dynamic range only in the area determined to be the edge part by the edge part detection circuit 550.

The moving image receiving apparatus can be realized if it has the same configuration as that of the first to fourth embodiments. For example, in FIG.
The mode determination circuit 146 determines that a scene change has occurred if all the decodable blocks in the blocks A to H are intra-frame encoded, and the selector 147.
The signal from the pixel value interpolation circuit 145 may be output via the. Further, as shown in FIG. 16, when the scene change identification flag is sent from the transmitting side,
The mode determination circuit 146 may be configured to supply the selector 147 with a signal that outputs the input from the pixel value interpolation circuit 145 without fail.

(Sixth Embodiment) Next, a moving image transmission apparatus according to the sixth embodiment of the present invention will be described. FIG. 17
FIG. 3 is a block diagram showing a moving image transmitting apparatus according to this embodiment.

In FIG. 17, the input image is stored in a plurality of frame memories 500. Motion vector detection circuit 5
In FIG. 10, after detecting a motion vector between frames stored in the frame memory 500, a sum of absolute values of error signals (obtained when a motion vector is detected) for each block and a motion vector when the frame compensates for another frame. The scene change detection circuit 52 outputs a signal for identifying whether or not it becomes a reference image.
Supply to 0.

The scene change detection circuit 520 calculates the activity of the input image supplied from the frame memory 500 (for example, the sum of absolute values of AC components), and calculates the sum of absolute values of error signals supplied from the motion vector detection circuit 510. Compare. Here, for example, if the number of blocks in which the latter value is large is more than half of the total number of blocks in a frame, it is determined that a scene change has occurred, and if the frame is a reference image of another frame, The scene change identification signal is supplied to the encoder 600. The edge part detection circuit 550 detects the edge part by applying an edge detection filter to the signal of the frame supplied from the frame memory 500, and identifies for each block whether the block belongs to a region including the edge part. Encoder 600
Supply to.

In the encoder 600, the frame memory 5
The image signal supplied from No. 00 is divided into blocks, and the motion vector supplied from the motion vector detection circuit 510 is used to perform motion compensation predictive coding for each block, and then the coded image signal is output. Here, when the scene change identification signal supplied from the scene change detection circuit 520 is enabled (for example, frame #n in FIG. 22A), all the blocks of the frame are intra-frame encoded. In addition, a frame that uses this frame as a reference image for motion compensation prediction (for example, FIG. 22A)
In frame # n + 1), the block for which the edge detection circuit 550 determines that the frame belongs to the edge part is forcibly intraframe-encoded.

The moving image receiving apparatus can be realized if it has the same configuration as that of the first to fourth embodiments. For example, in FIG.
If all the decodable blocks among the blocks A to H are intra-frame coded in the mode determination circuit 146 of 1, the scene change is determined to have occurred, and the signal from the pixel value interpolation circuit 145 is output via the selector 147. It should output it. Further, as shown in FIG. 16, when the scene change identification flag is sent from the transmission side, the mode determination circuit 146 always outputs the input from the pixel value interpolation circuit 145 to the selector 147. It is sufficient to supply such a signal.

In this way, for example, even if cell discard occurs in frame #n of FIG. 22A, the spatial resolution drops due to the concealment by spatial interpolation, and the edge portion is blurred,
In the frame # n + 1, the edge portion is forcibly intra-coded to eliminate the edge blur.

In the first to seventh embodiments described above, if the input image is an interlaced signal,
The type of concealment may be switched for each field.

In all of the above embodiments (first to sixth), when these embodiments are applied to MPEG, the motion compensation circuit (14) relating to concealment is used.
1, 144, 220), the pixel average value calculation circuit required in the 1/2 pixel motion compensation, the interpolation prediction mode of the B picture, etc. is omitted (for example, the motion vector is rounded in 1 pixel units). When the interpolative prediction mode is selected, the circuit scale can be reduced by performing concealment using only forward prediction.

(Seventh Embodiment) Next, a moving image transmission apparatus according to a seventh embodiment of the present invention will be described. FIG.
FIG. 3 is a block diagram showing a moving image transmitting apparatus according to the present embodiment.

In FIG. 18A, first, the video signal supplied for each frame is stored in the first frame memory 700. Next, the motion vector detection circuit 710 detects the MV from the first frame memory 700 and the second frame memory 720 that stores the locally decoded image, and supplies it to the motion compensation prediction circuit 730. The motion compensation prediction circuit 730 creates a predicted image signal from the second frame memory 720 using the MV supplied from the motion vector detection circuit 710 and supplies it to the encoding control circuit 740 and the intra-frame / inter-frame switching circuit 750. To do.

Intra-frame / inter-frame switching circuit 75
At 0, when the interframe coding is selected, the optimum prediction value created by the motion compensation prediction circuit 730 is set, and when the intraframe coding is selected, the prediction value is set to 0 and the subtractor 760 is set. To the adder 770.

The subtractor 760 creates an error signal between the input image and the prediction signal and supplies it to the DCT circuit 780. The prediction error signal converted into the transform coefficient in the DCT circuit 780 is supplied to the quantizer 790. The transform coefficient quantized in the quantizer 790 is the inverse quantizer 8
The transform coefficient dequantized in 00 is supplied to the inverse DCT circuit 810. The inverse DCT circuit 810 inversely transforms the transform coefficient into a prediction error signal and supplies it to the adder 770. In the adder 770, the prediction error signal and the prediction value are added to create a locally decoded image, and the second frame memory 720 is generated.
Supply to. Further, the transform coefficient quantized by the quantizer 790 is also supplied to the variable length coding circuit 820.

In the variable length coding circuit 820, the quantizer 7
The transform coefficient quantized in 90 is subjected to variable length coding together with additional information including MV, and then supplied to the multiplexing circuit 830.

The multiplexing circuit 830 multiplexes the variable length code supplied from the variable length coding circuit 820 and outputs the coded bit stream via the output buffer 840. Also, in the output buffer 840, the amount of stay of the encoded bitstream in the buffer is determined by the encoding control circuit 740.
Supply to.

The encoding control circuit 740 uses the predicted image signal supplied from the motion compensation prediction circuit 730 and the input image signal supplied from the first frame memory 700 to perform an optimum prediction method (for example, forward prediction). (Prediction, backward prediction, etc.) and intra-frame coding and inter-frame coding are determined for each block to determine the prediction mode. Here, in the motion compensation prediction of the block belonging to the region 2 in FIG. 24, if the image signal belonging to the region 1 is referred to, the intra-frame prediction mode is forcibly set.
Here, if error resilience is required, the mode is switched to a mode in which a motion vector is added in intraframe prediction coding, and the motion vector information is also supplied to the variable length coding circuit 820.

By the above, the determined in-frame / inter-frame determination information is supplied to the in-frame / inter-frame switching circuit 750. Further, the quantization step size is determined so as to be adapted to the set code amount from the staying amount of the coded bit stream supplied from the output buffer 840, and is supplied to the quantizer 790 and the dequantizer 800.

[0099]

According to the present invention (claims 1 and 2), in a concealment method using motion compensation prediction of adjacent blocks, it can occur due to discontinuity of low frequency components with adjacent blocks. It is possible to reduce the block-shaped distortion having the characteristic by using the signal of the low frequency component predicted from the signals in the adjacent blocks.

[0100]

[0101]

[0102]

[0103]

[Brief description of drawings]

FIG. 1 is a block diagram showing a main configuration of a moving image receiving apparatus according to a first embodiment of the present invention.

FIG. 2 is a diagram illustrating data used for concealment.

3 is a block diagram showing a configuration example of an error detection circuit in FIG.

FIG. 4 is a block diagram showing another configuration example of the concealment circuit of FIG.

FIG. 5 is a block diagram showing still another configuration example of the concealment circuit of FIG.

FIG. 6 is a block diagram showing a main configuration of a moving image receiving apparatus according to a second embodiment of the present invention.

FIG. 7 is a block diagram showing a main configuration of a moving image receiving apparatus according to a third embodiment of the present invention.

FIG. 8 is a block diagram showing another configuration example of the moving image receiving apparatus according to the third embodiment of the present invention.

FIG. 9 is a block diagram showing a main configuration of a moving image receiving apparatus according to a fourth embodiment of the present invention.

FIG. 10 is a diagram illustrating a DCT coefficient prediction method.

FIG. 11 is a block diagram showing another configuration example of the moving image receiving apparatus according to the fourth embodiment of the present invention.

FIG. 12 is a block diagram showing a main configuration of a moving image transmitting apparatus according to a fifth embodiment of the present invention.

FIG. 13 is a block diagram showing another configuration example of the moving image transmitting apparatus according to the fifth embodiment of the present invention.

FIG. 14 is a block diagram showing still another configuration example of the moving image transmitting apparatus according to the fifth embodiment of the present invention.

FIG. 15 is a block diagram showing still another configuration example of the moving image transmitting apparatus according to the fifth embodiment of the present invention.

FIG. 16 is a block diagram showing still another configuration example of the moving image transmitting apparatus according to the fifth embodiment of the present invention.

FIG. 17 is a block diagram showing a main configuration of a moving image transmitting apparatus according to a sixth embodiment of the present invention.

FIG. 18 is a block diagram showing the main configuration of a moving image transmitting apparatus according to a seventh embodiment of the present invention.

FIG. 19 is a diagram for explaining the principle of conventional packet discard compensation.

FIG. 20 is a diagram illustrating a conventional motion vector selection method.

FIG. 21 is a diagram illustrating an example of conventional packet discard compensation.

FIG. 22 is a diagram for explaining concealment by space interpolation in the frame.

23 is a diagram for explaining the effect of the band of frame #n on the transmitting side in FIG. 22 on the band of frame # n + 1 on the receiving side.

FIG. 24 is a diagram illustrating a method of setting an intra slice in a frame to limit the search range of motion vectors.

[Explanation of symbols]

100 ... Error detection circuit, 101 ... Packet decomposition circuit, 102 ... Demultiplexer / variable length decoding circuit, 103 ... Undecodable identification flag circuit, 104 ... Error correction circuit, 110 ... Decoder, 111 ... Dequantization circuit 112, 470 Inverse DCT circuit, 113, 141, 144, 220 ... Motion compensation circuit, 114 ... Adder, 120, 240, 300, 400 ... Motion compensation information memory, 130, 500 ... Frame memory, 140 ... Concealment circuit, 142 ... Motion compensation error evaluation circuit, 143 ... Selection circuit, 145, 330 ... Pixel value interpolation circuit, 146, 340, 450 ... Mode determination circuit, 148 ... Attribute determination circuit, 150, 147, 480 ... Selector, 200, 210 ... Activity Calculation circuit, 230 ... Reliability determination circuit, 250, 320, 420 Delay circuit, 310, 410 ... Motion compensation estimation circuit, 350 ... Weighted averaging circuit, 430 ... Low-order coefficient prediction circuit, 440 ... Coefficient separation / synthesis circuit, 460 ... DCT circuit, 510 ... Motion vector detection circuit, 520 ... Scene change Detection circuit, 530 ... Resolution conversion circuit, 535 ... Dynamic range conversion circuit, 540, 560, 600 ... Encoder, 550 ... Edge detection circuit, 700, 720 ... Frame memory, 710 ... Motion vector detection circuit, 730 ... Motion compensation prediction Circuit, 740 ... Encoding control circuit, 750 ... In-frame / inter-frame switching circuit, 760 ... Subtractor, 770 ... Adder, 780 ... Discrete cosine transform circuit, 790 ... Quantizer, 800 ... Inverse quantizer, 810 ... Inverse Discrete Cosine Transform Circuit, 820 ... Variable Length Coding Circuit, 830 ... Multiplexing circuit, 840 ... Output buffer

Continuation of front page (56) Reference JP-A-3-46481 (JP, A) JP-A-4-72985 (JP, A) JP-A-2-188079 (JP, A) JP-A-2-172388 (JP , A) JP-A-2-219387 (JP, A) JP-A-2-172384 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (2)

(57) [Claims] 1. A moving image transmitting apparatus that transmits data obtained by encoding a moving image signal in units of blocks using motion compensation prediction, and a moving image receiving apparatus that receives the encoded data and decodes in units of blocks. In the moving picture transmission apparatus, the moving picture receiving apparatus detects an error in the received coded data and outputs an error detection signal, and an error detection unit that does not decode due to the error based on the error detection signal. Identification means for identifying a possible block, motion vectors of a plurality of decodable blocks in the vicinity of the block identified as undecodable by the identification means, and a method of motion compensation are identified as undecodable. Selecting means for selecting a motion vector and a motion compensation method to be applied to the selected block, and using the selected motion vector and the motion compensation method, Motion-compensated prediction means for motion-compensated predicting a block identified as undecodable to generate a prediction signal; and using a plurality of pixel values in the vicinity of the block identified as undecodable Interpolation prediction means for interpolatively predicting pixel values in the identified block to generate an interpolated prediction signal, and a weighted average value of the motion compensation predicted signal and the interpolated predicted signal is identified as the undecodable And a processing means for converting the image signal of the selected block into a moving image transmission device. 2. A moving image transmitting apparatus that transmits data obtained by encoding a moving image signal in block units using motion compensation prediction, and a moving image receiving apparatus that receives the encoded data and decodes in block units. In the moving picture transmission apparatus, the moving picture receiving apparatus detects an error in the received coded data and outputs an error detection signal, and an error detection unit that does not decode due to the error based on the error detection signal. Identification means for identifying a possible block, motion vectors of a plurality of decodable blocks in the vicinity of the block identified as undecodable by the identification means, and a method of motion compensation are identified as undecodable. Selecting means for selecting a motion vector and a motion compensation method to be applied to the selected block, and using pixel values in the vicinity of the block identified as undecodable, Prediction means for predicting a low frequency component in a block identified as undecodable, and motion compensation prediction for the block identified as undecodable using the motion vector and motion compensation method selected by the selecting means Separation means for separating a high frequency component from the predicted signal, and a full band image signal configured by combining the predicted low frequency component and the separated high frequency component signal are identified as undecodable And a processing means for converting the image signal of the block into a moving image transmission apparatus.