JP4302094B2 - Moving picture decoding apparatus and moving picture decoding method - Google Patents

Description

  The present invention relates to a moving picture decoding apparatus and a moving picture decoding method for compressing and encoding a moving picture signal with high efficiency, and more particularly to a moving picture decoding apparatus and a moving picture decoding which are resistant to errors in a transmission path / storage medium. Regarding the method.

  In a system that transmits / stores moving image signals, such as a TV phone, a TV conference system, a portable information terminal, a digital video disk system, and a digital TV broadcast system, the moving image signals are compressed and encoded to the smallest possible amount of information. The code sequence which is the compressed and encoded signal is transmitted / accumulated to the transmission path / storage medium, and the transmitted / accumulated code sequence is decoded to reproduce the original moving image signal.

  As a compression encoding technique of a moving image signal applied to such a system, various systems such as a system such as motion compensation, discrete cosine transform (DCT), subband encoding, pyramid encoding, and a combination thereof are available. Has been developed. In addition, ISO / MPEG1, MPEG2, ITU-T / H. 261, H.M. 262 is defined. All of these encoding methods are based on a combination of motion compensated adaptive prediction and discrete cosine transform. Reference 1: edited by Hiroshi Yasuda, “International Standard for Multimedia Coding”, Maruzen, (1991) (June) etc., the details are described.

  By the way, when a code string obtained by encoding a moving image signal in this way is transmitted / stored via a medium that is prone to error such as a wireless transmission path, an image signal reproduced on the decoding side is transmitted. / May be deteriorated due to errors during storage. As a countermeasure against such an error, for example, under conditions where a code string can be transmitted through a plurality of transmission paths having different error probabilities, the code string is divided into several layers in order to reduce image quality degradation due to errors. Hierarchical coding schemes are known in which a hierarchical code string is transmitted on a transmission path with a lower error probability. As the layer separation, methods such as mode information, motion compensation information, and a low frequency component of an image signal as an upper layer and a high frequency component of an image signal as an upper layer have been proposed.

  As an example of a conventional hierarchical video encoding device, a video encoding device using motion compensated adaptive prediction and DCT will be described with reference to FIG. In FIG. 5, an input moving image signal 11 is first subjected to prediction by a prediction circuit 12. In the prediction circuit 12, a motion vector between the input moving image signal 11 and the reference image signal already obtained by encoding / local decoding stored in the frame memory 13 is detected, and based on this motion vector Motion compensation prediction is performed for each predetermined unit region (referred to as a prediction region), and a motion compensation prediction signal is generated.

  The prediction circuit 12 has a motion compensation prediction mode (interframe prediction mode) and an intraframe prediction mode for encoding the input moving image signal 11 as they are, and selects an optimum mode for encoding from these prediction modes. The prediction signal 14 corresponding to each mode is output. That is, the prediction circuit 12 outputs a motion compensation prediction signal in the motion compensation prediction mode and “0” as the prediction signal 14 in the intraframe prediction mode.

  In the subtracter 15, the prediction residual signal 16 is generated by subtracting the prediction signal 14 from the input moving image signal 11. The prediction residual signal 16 is subjected to discrete cosine transform in a block unit having a certain size in a discrete cosine transform (DCT) circuit 17 to become DCT coefficient information 18. The DCT coefficient information 18 is quantized by the quantization circuit 19. Since the prediction signal 14 is “0” in the intra-frame prediction mode, the input moving image signal 11 is output as it is from the subtracter 15 as the prediction residual signal 16.

  The quantized DCT coefficient information 20 from the quantization circuit 19 is bifurcated, and on the one hand, variable length coding is performed by the first variable length coding circuit 21, and on the other hand, dequantization is performed by the inverse quantization circuit 22. . The output of the inverse quantization circuit 22 is subjected to inverse discrete cosine transform by an inverse discrete cosine transform (inverse DCT) circuit 23. That is, the inverse quantization circuit 22 and the inverse DCT circuit 23 perform processing reverse to that of the quantization circuit 19 and the DCT circuit 17, and a signal approximate to the prediction residual signal 16 is obtained at the output of the inverse DCT circuit 23. The output of the inverse DCT circuit 23 is added to the prediction signal 14 from the prediction circuit 12 in the addition circuit 24 to generate a local decoded signal 25. This local decoded signal 25 is stored in the frame memory 13 as a reference image signal.

  Prediction mode / motion vector information 26 is output from the prediction circuit 12 as information related to prediction, and variable length coding is performed by the variable length coding circuit 27. The code sequences output from the variable-length encoding circuits 21 and 27 are multiplexed by the multiplexer 30 and divided into an upper layer code sequence 31 and a lower layer code sequence 32, and output to a transmission path / storage medium (not shown). The That is, the upper layer code string 31 is output to a transmission line / storage medium with a relatively low probability of transmission / accumulation error, and the lower layer code string 32 is a transmission line / storage medium with a relatively high probability of transmission / accumulation error. Output to the storage medium.

  Here, the separation of the upper layer code string 31 and the lower layer code string 32 in the multiplexer 30 is performed by, for example, mode information indicating a prediction mode in the prediction circuit 12 from the variable length coding circuit 22 as shown in FIG. The low-frequency DCT coefficient information of the vector information (MV) and the variable-length coded DCT coefficient information from the variable-length code circuit 21 is applied to the upper layer code string 31, and the variable-length coding from the variable-length code circuit 21 is performed. The remaining high-frequency DCT coefficient information in the DCT coefficient information is applied to the lower layer code string 32.

  Such a conventional hierarchical video encoding apparatus has the following problems. First, since there is only one motion vector information for each prediction region that causes image quality to deteriorate greatly if an error occurs, if an error occurs in the motion vector information, no motion information can be decoded for that prediction region. That is, the image quality is greatly deteriorated. In order to reduce such image quality degradation, all the motion vector information (MV) may be applied to the upper layer code string 31 as shown in FIG. However, in general, there is a limit to the ratio that the code amount of the code sequence of each layer can take in the total code amount of the code sequence of all layers, and when all the motion vector information is included in the upper layer code sequence 31 in this way May exceed this limit. In order to prevent this, if the motion vector information is applied to the lower layer code string 32, there arises a problem that the error resistance is greatly lowered.

  In addition, since each code word of the code strings 31 and 32 to be transmitted / stored is composed of a variable length code created by the variable length coding circuits 21 and 27, the variable length code is decoded at the time of decoding due to an error. Out of sync may occur. However, in the conventional moving picture coding apparatus, as shown in FIG. 6, in the code strings 31 and 32 of each layer, if an error occurs, the decoded picture is greatly degraded, such as mode information and motion vector information. Information related to the information and DCT coefficient information of a prediction residual signal that does not cause a great deterioration even if an error occurs are mixed and multiplexed. For this reason, if a loss of synchronization occurs during decoding of a codeword on which unimportant information is placed, an error may be propagated to the codeword on which important information is placed. Cause it to occur. In such a case, since it is impossible to recover the synchronization until the synchronization code appears, all the information of the decoded image up to that time is incorrect, and a large deterioration occurs over a wide range in the screen. is there.

  As described above, in the conventional moving image encoding apparatus, only one piece of information related to prediction such as motion vector information that greatly deteriorates the quality of a decoded image when an error occurs is encoded for each prediction region. Since it is not, error tolerance is low.

  In order to increase error resilience, information related to all predictions must be transmitted / stored via a transmission path / storage medium with a low error probability, but the code amount of each layer of code sequences is the sum of the code sequences of all layers. Since there is a limit on the ratio that can be taken in the code amount, as a result, the transmission / accumulation of the code string using a plurality of transmission paths / accumulation media having different error probabilities reduces the image quality degradation due to errors. The problem of not being able to take advantage of the characteristics of encoding occurs.

  In addition, in the conventional video encoding apparatus, relatively important information such as information related to prediction and other relatively unimportant information are included in the code string in a mixed manner. There is a problem that the error generated in the process spreads to important information and causes a large deterioration in image quality.

  Accordingly, an object of the present invention is to provide a moving picture decoding apparatus and a moving picture decoding method with high error tolerance.

  In order to solve the above problems, the present invention provides a prediction mode selected from a plurality of prediction modes including a motion compensation prediction mode and an intraframe prediction mode for each of a plurality of blocks obtained by dividing an input image constituting an input moving image signal. And the prediction residual signal that is an error signal between the prediction signal generated when the prediction is performed in the motion compensation prediction mode and the input moving image signal, or the prediction is performed in the intra-frame prediction mode. A code sequence including DCT coefficient information quantized with respect to the input video signal, wherein information indicating a prediction mode of the frame is arranged in a picture header of each frame of the input image, and the motion compensation prediction mode is When selected, information indicating the prediction mode related to each block and motion vector information are encoded for a plurality of blocks, and the plurality of blocks are encoded. Means for inputting a code string in which information indicating a prediction mode combined for a plurality of blocks is arranged behind the picture header, with the information indicating the prediction mode combined for each block being the most important information. Means for decoding the code string and reproducing the original moving image signal.

  Further, the present invention performs prediction by a prediction mode selected from a plurality of prediction modes including a motion compensation prediction mode and an intra-frame prediction mode for each of a plurality of blocks obtained by dividing an input image forming an input moving image signal, A prediction residual signal which is an error signal between a prediction signal generated when prediction is performed in the motion compensation prediction mode and the input moving image signal, or the input moving image when prediction is performed in the intra-frame prediction mode. A code string including DCT coefficient information quantized for a signal, information indicating a prediction mode of the frame is arranged in a picture header of each frame of the input image, and the motion compensation prediction mode is selected Predicting the prediction mode related to each block and motion vector information by encoding them for multiple blocks. The information indicating the mode is the most important information, and means for inputting a code string in which information indicating the prediction mode combined for the plurality of blocks is arranged behind the picture header and multiplexed, and the input code string is And a means for decoding and reproducing the original moving image signal.

  According to the present invention, it is possible to provide a moving picture decoding apparatus and a moving picture decoding method that have high error tolerance, that is, little deterioration in the image quality of a decoded picture due to errors that occur during transmission / accumulation of a coded sequence.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of an embodiment of a moving picture coding apparatus according to the present invention, showing an example in which the present invention is applied to a moving picture coding apparatus that combines motion compensated adaptive prediction and DCT.

  In FIG. 1, an input moving image signal 101 is first subjected to prediction by a prediction circuit 102. That is, the prediction circuit 102 detects a motion vector between the input moving image signal 101 and the reference image signal already obtained by encoding / local decoding stored in the frame memory 103, and based on this motion vector. Thus, a motion compensated prediction signal is generated. The prediction circuit 102 has a motion compensation prediction mode (interframe prediction mode) and an intraframe prediction mode in which the input moving image signal 101 is encoded as it is, and selects a prediction mode optimal for encoding and selects each mode. The prediction signal 104 corresponding to is output. That is, the prediction circuit 102 outputs a motion compensation prediction signal in the motion compensation prediction mode and “0” as the prediction signal 104 in the intraframe prediction mode.

  In the subtractor 105, the prediction residual signal 106 is generated by subtracting the prediction signal 104 from the input moving image signal 101. The prediction residual signal 106 is subjected to discrete cosine transform in units of blocks having a certain size in a discrete cosine transform (DCT) circuit 107, and becomes DCT coefficient information 108. The DCT coefficient information 108 is quantized by the quantization circuit 109. In the intraframe prediction mode, since the prediction signal 104 is “0”, the subtractor 105 outputs the input moving image signal 101 as it is as the prediction residual signal 106.

  The quantized DCT coefficient information 110 from the quantization circuit 109 is bifurcated, and on the one hand, variable length coding is performed by the first variable length coding circuit 111, and on the other hand, dequantization is performed by the inverse quantization circuit 112. . The output of the inverse quantization circuit 112 is subjected to inverse discrete cosine transform by an inverse discrete cosine transform (inverse DCT) circuit 113. That is, the inverse quantization circuit 112 and the inverse DCT circuit 113 perform processing reverse to that of the quantization circuit 109 and the DCT circuit 107, and a signal approximate to the prediction residual signal 106 is obtained at the output of the inverse DCT circuit 113. The output of the inverse DCT circuit 113 is added to the prediction signal 104 from the prediction circuit 102 in the addition circuit 114, and a local decoded signal 115 is generated. This local decoded signal 115 is stored in the frame memory 103 as a reference image signal.

  As will be described later, the prediction circuit 102 outputs large area prediction mode / motion vector information 116 and small area prediction mode / motion vector information 117 as information related to prediction, and variable length encoding circuits 118 and 119 respectively have variable lengths. Encoded. The code sequences output from the variable-length encoding circuits 111, 118, and 119 are multiplexed by the multiplexer 120 and are divided into an upper layer code sequence 121 and a lower layer code sequence 122 to a transmission path / storage medium (not shown). Is output.

  Here, under the condition that the code string can be transmitted / stored via a plurality of transmission paths / storage media having different error probabilities, the transmission path / storage medium having a lower error probability is provided for the upper layer code string 121. The lower layer code string 122 is transmitted / stored via a transmission path / storage medium having a higher error probability, so that the upper layer code string 121 is less likely to cause errors. Like that. In addition, when error correction coding is performed on the coded sequences 121 and 122, stronger error correction is performed so that the higher layer coded sequence 121 has a lower error rate than the lower layer coded sequence 122. Encoding is performed.

  Next, the configuration and operation of the prediction circuit 102 will be described in detail with reference to FIG. In the prediction circuit 102, the input moving image signal 101 is divided into a larger number of regions sequentially from the upper level to the lower level, and motion compensation prediction is performed on the input moving image signal for each of the divided regions. Generate a prediction signal. In the example of FIG. 2, the region division and prediction in the prediction circuit 102 are performed in two stages. That is, in the first stage, the prediction circuit 102 divides the input moving image signal 101 into large areas as indicated by solid line areas (referred to as large areas) in FIG. 2, performs motion compensation prediction with coarse pixel accuracy for these large areas, Next, in the second stage, if necessary, the large area is further divided as indicated by the broken line area (referred to as a small area) in the figure, and motion compensation prediction is performed with fine pixel accuracy for these small areas.

  Then, the variable length encoding circuits 118 and 119 encode not only the information related to the prediction for the large area output from the prediction circuit 102 but also the information related to the prediction for the large area for the small area. In this way, even if the information about the prediction for the small area is lost due to an error, if the information about the prediction for the large area is decoded without error in the decoding device, the prediction can be performed with a rough accuracy. This can be performed, and the deterioration of the image quality of the decoded image can be prevented.

  The information related to prediction output from the prediction circuit 102 includes information indicating a prediction mode and information indicating a motion vector. The information indicating the large region prediction mode and the information indicating the motion vector (indicated by the solid line arrow in FIG. 2), that is, the large region prediction mode / motion vector information 116 are variable length encoded by the variable length encoding circuit 118. . At this time, the motion vector information may be variable-length encoded with respect to the difference from the adjacent already-encoded large-area motion vector information, or may be fixed-length encoded as it is without taking the difference. Alternatively, the motion vector information may be fixed-length encoded for each region at a certain interval, and the motion vector information of other regions may be variable-length encoded.

  On the other hand, the information indicating the small region prediction mode and the information indicating the motion vector (indicated by the broken arrow in FIG. 2), that is, the small region prediction mode / motion vector information 117 are variable length encoded by the variable length encoding circuit 119. Is done. In this case, the difference between the motion vector information and the large region motion vector information may be taken for each small region and variable length coding may be performed, or block coding, vector quantization, etc. may be used for each large region collectively. Encoding may be performed. When variable length coding is performed on the difference value, if a large area motion vector can be expressed by a reversible function (for example, an average value) based on the small area motion vector, one of the small area motion vectors is large. Since it can be obtained by calculation based on the region motion vector and other small region motion vectors, there is no need for encoding.

  FIG. 3 shows a configuration example of the upper layer code sequence 121 and the lower layer code sequence 122. In the upper layer code string 121 shown in FIG. 3A, a synchronous code that can be uniquely decoded is placed at the head for each encoded frame or for each region unit. PSC indicates a frame-by-frame synchronization code. A picture header indicating the encoding information of the frame is attached after the synchronization code PSC. The picture header includes a frame number indicating the temporal position of the frame, information indicating the prediction mode of the frame (mode information), and information (code amount) indicating the lengths of the code sequences of the upper and lower layers of the frame. Consists of.

  Furthermore, as shown in FIG. 3A, if information (large region MC accuracy, small region MC accuracy) indicating the sizes of large regions and small regions and pixel accuracy of motion compensation is added to the picture header, the accuracy of motion compensation Can be varied in units of frames to control the amount of code of motion vector information. Thereby, even when the ratio of the code amount of the upper layer code sequence 121 and the lower layer code sequence 122 is defined by the restrictions of the transmission path / storage medium, the code amount corresponding to the code amount can be allocated. In addition, since it is possible to control the total code amount of the motion vector information of each frame, it is possible to select the optimal motion compensation accuracy in view of the relationship between the motion compensation accuracy and the code amount of the motion vector information. Efficiency can be improved.

  Characteristically, the encoded information of each region is arranged in order from the most important information behind the picture header of the upper layer code sequence 121 shown in FIG. Here, information having a high degree of importance is information that causes a large deterioration in the decoded image if an error occurs. That is, after the picture header of the upper layer code sequence 121, first, information (mode information) indicating a prediction mode having the highest importance, that is, the large region prediction mode / motion vector information 116 output from the prediction circuit 102 and the small information. The code sequence of the prediction mode information among the code sequences obtained by encoding the region prediction mode information 117 by the variable length encoding circuit 118 is inserted.

  Next, the DC component (intra DC) of the DCT coefficient information in the region where the intra-frame prediction mode is selected, that is, the DCT coefficient information obtained by passing the prediction residual signal 106 through the DCT circuit 107 and the quantization circuit 109 is variable-length encoded. The DC component code string of the code string obtained by encoding by the circuit 111 is inserted. Further, in the region in which the motion compensation prediction mode is selected, the large region motion vector information (large region MV) indicating rough motion information, that is, the large region prediction mode / motion vector information 116 output from the prediction circuit 102 is variable length. A code string of motion vector information in the code string obtained by encoding by the encoding circuit 118 is input.

  On the other hand, in the lower layer code sequence 122 shown in FIG. 3B, the small area motion vector information (small area MV), that is, the small area prediction mode / motion vector information 117 output from the prediction circuit 102 is variable length code. The code sequence of the motion vector information in the code sequence obtained by the encoding circuit 119 is inserted, and the high-frequency component of the DCT coefficient information, that is, the prediction residual signal 106 is further added to the DCT circuit 107 and the quantization circuit 109. The code sequence of the high-frequency component in the code sequence obtained by encoding the DCT coefficient information obtained through the variable length encoding circuit 111 is inserted.

  In this way, motion compensation prediction is performed in a hierarchical manner, and mode information indicating a prediction mode and large region motion vector information indicating rough prediction information are stored in the upper layer code sequence 121, and fine motion vector information is stored in the lower layer code sequence 122. Assigned to each. Therefore, even when the small region motion vector information included in the lower layer code sequence 122 is lost due to an error in the transmission path / storage medium, the moving image decoding apparatus uses the large region motion vector information included in the upper layer code sequence 121. Therefore, it is possible to perform motion compensation prediction with rough accuracy, so that the probability of significant image quality degradation in the decoded image can be reduced.

  Further, in the present embodiment, since the code sequences are arranged in the order of important information in the upper and lower hierarchical code sequences 121 and 122, an error caused by unimportant information does not spread to important information, Significant image quality degradation can be prevented.

  Next, an embodiment of the moving picture decoding apparatus according to the present invention will be described. FIG. 4 is a block diagram showing a configuration of a moving picture decoding apparatus corresponding to the moving picture encoding apparatus of FIG.

  In FIG. 4, the upper layer code sequence 201 and the lower layer code sequence that are input via the transmission path / storage medium with the upper layer code sequence 121 and the lower layer code sequence 122 output from the moving picture encoding apparatus of FIG. 202 is separated by the demultiplexer 203 into a variable length code 204 of quantized DCT coefficient information, a variable length code 205 of large area prediction mode / motion vector information, and a variable length code 206 of small area prediction mode / motion vector information. Are input to variable length decoding circuits 207, 208 and 209, respectively.

  The variable length decoding circuit 207 outputs quantized DCT coefficient information 210 by variable length decoding the variable length code 204. The quantized DCT coefficient information 210 is inversely quantized by an inverse quantization circuit 213 and further subjected to inverse discrete cosine transform by an inverse DCT circuit 214 to generate a prediction residual signal 215. The prediction residual signal 215 is added to the prediction signal 219 from the prediction circuit 217 by the addition circuit 216, and a reproduced image signal 221 is generated. The reproduced image signal 221 is output to the outside of the moving image decoding apparatus and is stored in the frame memory 218 as a reference image signal.

  On the other hand, the variable length decoding circuits 208 and 209 perform variable length decoding of the large region prediction mode / motion vector information 211 and the small region prediction mode / motion vector information 212 from the variable length codes 205 and 206. These pieces of information 211 and 212 are input to the prediction circuit 217. The prediction circuit 217 predicts a moving image signal from the reference image signal stored in the frame memory 218, the large region prediction mode / motion vector information 211, and the small region prediction mode / motion vector information 212, and generates a prediction signal 219. To do.

  The error determination circuit 220 determines whether or not there is an error in the upper layer code sequence 201 and the lower layer code sequence 202 based on the states of the demultiplexer 203 and the variable length decoding circuits 207, 208, and 209, and the determination result is a prediction circuit. 217. When the error determination circuit 220 detects no error in any of the upper and lower layer code sequences 201 and 202, the prediction circuit 217 predicts the prediction in FIG. 1 based on the reference image signal stored in the frame memory 218. The same prediction signal 219 as the signal 104 is output.

  The above processing is processing for reproducing an image signal corresponding to the moving image encoding apparatus in FIG. 1. The processing performed by the inverse quantization circuit 213, the inverse DCT circuit 214, the addition circuit 216, and the frame memory 218 is respectively The processing of the inverse quantization circuit 112, the inverse DCT circuit 113, the addition circuit 114, and the frame memory 103 in FIG. In addition, the variable length decoding circuits 207, 208, and 209 and the multiplexer 203 perform processing reverse to the processing of the variable length encoding circuits 111, 118, and 119 and the multiplexer 120 in FIG.

  On the other hand, when an error is detected in at least one of the upper and lower layer code sequences 201 and 202 by the error determination circuit 220, for example, reproduction is performed using information having higher importance than information in which an error is detected as follows. An image is created.

  That is, (1) when an error is detected in the DCT coefficient information of a block subjected to motion compensation prediction included in the lower layer code sequence 122, the prediction residual signal of that block is set to 0 and decoded correctly A reproduced image signal 221 is obtained using a motion compensated prediction signal obtained using the mode information, the large area motion vector information, and the small area motion vector information as a prediction signal 219.

  Also, (2) when an error occurs in the small area motion vector information, the reproduced image signal 221 is obtained using the motion compensated prediction signal obtained using the large area motion vector information for the small area as the prediction signal 219. .

  Further, (3) when an error occurs in the large area motion vector information, when the large area motion compensation prediction can be performed for the large area from the surrounding area or the motion vector information of the already decoded frame, If the estimated value is used and there is an error in the surrounding motion vector information, the image signal of the already decoded frame is used as it is as the reproduced image signal 221.

  (4) When an error occurs in the AC component of the DCT coefficient information of the block for which the intra-frame prediction mode is selected, the image of that block is determined from the DC component of the DCT coefficient information and the image signal correctly decoded from the surrounding blocks. The signal is predicted to be the reproduced image signal 221 or the image signal of the block is predicted from the image signal of the already decoded frame using the prediction signal of the image signal correctly decoded in the surrounding blocks, and the reproduced image Signal 221 is assumed.

  By the way, when a variable length code is used for encoding various types of information, loss of synchronization may occur due to an error, and the error may spread to subsequent codes until synchronization recovery is achieved by synchronization code detection or the like. In such a case, subsequent codes are not used for decoding. For example, if an error occurs in the small area motion vector information in the lower layer code sequence 122, the error may spread to the motion vector information after the small area and the DCT coefficient information that follows. Does not use the information spreading the error for decoding. Even when such out-of-synchronization occurs, codewords are arranged in the code sequence in descending order of importance, so that an error caused by information of low importance may spread to information of high importance. Therefore, it is possible to prevent the reproduction image from being greatly deteriorated.

  As a specific method for detecting that an error has occurred in the code strings 201 and 202 by the error determination unit 220, the following method can be cited.

  The first is a method using an error detection code such as a parity code or a CRC code. In this case, error detection coding is performed on the variable-length code in the multiplexer 120 of the video encoding device shown in FIG. 1, and error detection processing is performed in the demultiplexer 203 of the video decoding device shown in FIG. The detection result is given to the error determination unit 220.

  Second, if a variable-length decoding circuit 207, 208, or 209 detects a codeword that does not exist in the codeword table that represents the correspondence between the input codeword and the output value, this is regarded as an error. It is a method of determination. In particular, when variable-length codes are used, errors may be spread not only in the part where the error is detected but also in the code string before and after that, so error detection is performed for all the codewords. Process.

  Third, whether motion vector information, prediction signal, DCT coefficient information, prediction residual signal, reproduction image signal, and the like reproduced based on the decoded codeword cannot be generated in moving picture encoding. This is a method of performing error determination by determining whether or not. Since this method is characteristic in the present invention, it will be described in more detail.

  For example, if the motion vector indicated by the motion vector information exceeds the search range defined in advance or protrudes outside the screen, it is determined as an error.

  It is also possible to detect an error by determining the DCT coefficient information inversely quantized by the inverse quantization circuit 213. Assuming that the pixel value range of the input image signal 101 can be 0 to D-1 and the DCT block size is N × N, the DCT coefficient takes a value within the following range.

  <Intraframe prediction mode> DC component: 0 to N × D AC component: − (N / 2 × D) to (N / 2 × D) <Interframe prediction mode> − (N / 2 × D) to (N Therefore, when the value of the decoded DCT coefficient takes a value outside this range, it is determined as an error. In this case, all or a part of DCT coefficients of a block in which an error is detected may be set to 0, or a decoded value may be estimated from the decoded values of surrounding blocks.

  Further, it is possible to determine an error based on the pixel value of the reproduced image signal 221. Assuming that the pixel value range of the input image signal 101 is 0 to D-1, the DCT block size is N × N, and the quantization width is Q (in the case of linear quantization), the pixel value of the reproduced image signal 221 is captured. The range obtained is −N × Q to D + N × Q. Therefore, when the pixel value of the decoded reproduced image signal 221 exceeds this range, it is determined as an error. In this case, for example, in the inter-frame prediction mode (motion compensation prediction mode), the prediction residual signal 215 is set to “0”, and in the intra-frame prediction mode, a part of the DCT coefficients input to the inverse DCT circuit 214 is set to “0”. Inverse DCT may be performed to obtain the reproduced image signal 221, or estimation may be performed from pixel values of blocks around the reproduced image signal 221.

  As described above, according to the present invention, it is determined whether or not the reproduced information or signal is information or signal that cannot be generated in the transmission path / storage medium in the error determination in the error determination unit 220 in the video decoding device. In addition, more accurate error determination can be performed. For this reason, it is possible to suppress quality degradation of a reproduced image that occurs when information or a signal in which an error has occurred is directly used for reproducing a moving image signal without error processing.

  The present invention can be implemented with various modifications. For example, in the above-described embodiment, an example in which the code string output from the video encoding device is divided into two layers has been described, but encoding may be performed by dividing into three or more layers. For example, the frame synchronization code (PSC), the picture header, and the mode information are set to the highest first layer, and the DC component (intra DC) of the DCT coefficient information and the large area motion vector information in the intra-frame prediction mode are set to the second layer, DCT coefficient information other than the DCT coefficient information assigned to the third layer and the first layer may be assigned to the fourth layer. The DCT coefficient may be further divided into several layers such as a low-frequency component and a high-frequency component, and encoding may be performed.

  Also, in the coding of large area motion vector information, the motion vector information for variable length coding the difference between the motion vector information to be fixed length coding and the motion vector information to be fixed length coding as described above. In the case of encoding in two ways, fixed-length encoded motion vector information that does not cause an error to propagate to the subsequent code string due to loss of synchronization is put together into a frame or a certain area unit first, It is effective to insert variable length encoded motion vector information behind. In this way, even if an error occurs in the variable length coding part and loss of synchronization occurs, the error is not propagated to the variable length coded motion vector information. Since it is possible to estimate a motion vector in which an error has occurred based on the motion vector information and create a prediction signal with rough accuracy, it is possible to reduce degradation in the quality of a reproduced image due to the error.

  Note that a method for determining whether or not the information reproduced in the moving picture decoding apparatus in FIG. 4 is information that cannot be generated in moving picture coding is not limited to a hierarchically encoded code string, and is a general moving picture. The present invention can also be applied to a moving image decoding apparatus that decodes an original image signal from a code string obtained by an image encoding apparatus.

The block diagram which shows one Embodiment of the moving image encoder by this invention The figure which shows the motion compensation area | region and motion vector corresponding to it in the moving image encoder of FIG. The figure which shows the example of the high-order hierarchy and the low-order hierarchy code sequence output from the moving image encoder of FIG. The block diagram which shows one Embodiment of the moving image decoding apparatus corresponding to the moving image encoding apparatus of FIG. Block diagram showing an example of a conventional video encoding device The figure which shows the example of the code sequence output from the moving image encoder of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... Input image signal 102 ... Prediction circuit 103 ... Frame memory 104 ... Prediction signal 105 ... Subtractor 106 ... Prediction residual signal 107 ... Discrete cosine transform circuit 108 ... DCT coefficient information 109 ... Quantization circuit 110 ... Quantization DCT coefficient information DESCRIPTION OF SYMBOLS 111 ... Variable length encoding circuit 112 ... Inverse quantization circuit 113 ... Inverse discrete cosine transform circuit 114 ... Adder 115 ... Local decoded signal 116 ... Large area prediction mode / motion vector information 117 ... Small area prediction mode / motion vector information 118 , 119 ... variable length coding circuit 120 ... multiplexer 121 ... upper layer code string 122 ... lower layer code string 201 ... upper layer code string 202 ... lower layer code string 203 ... demultiplexer 204 to 206 ... variable length code 207 to 209 ... Variable length coding circuit 210... Quantized DCT coefficient 21 ... large area prediction mode / motion vector information 212 ... small area prediction mode / motion vector information 213 ... inverse quantization circuit 214 ... inverse discrete cosine transform circuit 215 ... prediction residual signal 216 ... adder 217 ... prediction circuit 218 ... frame memory 219 ... Prediction signal 220 ... Error determination circuit 221 ... Reproduced image signal

Claims (4)

Prediction is performed in a prediction mode selected from a plurality of prediction modes including a motion compensation prediction mode and an intraframe prediction mode for each of a plurality of blocks obtained by dividing an input image forming an input moving image signal. The prediction residual signal, which is an error signal between the prediction signal generated when prediction is performed and the input video signal, or the input video signal is quantized when prediction is performed in the intra-frame prediction mode. A code string including DCT coefficient information,
Information indicating the prediction mode of the frame is arranged in the picture header of each frame of the input image, and when the motion compensation prediction mode is selected, a plurality of pieces of information indicating the prediction mode and motion vector information related to each block are provided. A code string in which the information indicating the prediction mode combined for the plurality of blocks is information having the highest importance, and the information indicating the prediction mode combined for the plurality of blocks is arranged behind the picture header. Means to input;
A moving picture decoding apparatus comprising: means for decoding the inputted code string and reproducing the original moving picture signal. Prediction is performed in a prediction mode selected from a plurality of prediction modes including a motion compensation prediction mode and an intraframe prediction mode for each of a plurality of blocks obtained by dividing an input image forming an input moving image signal. The prediction residual signal, which is an error signal between the prediction signal generated when prediction is performed and the input video signal, or the input video signal is quantized when prediction is performed in the intra-frame prediction mode. A code string including DCT coefficient information,
Information indicating the prediction mode of the frame is arranged in the picture header of each frame of the input image, and when the motion compensation prediction mode is selected, a plurality of pieces of information indicating the prediction mode and motion vector information related to each block are provided. A code obtained by coding blocks and encoding the information indicating the prediction mode combined for the plurality of blocks as the most important information and multiplexing the information indicating the prediction mode combined for the blocks after the picture header. A means of entering a column;
A moving picture decoding apparatus comprising: means for decoding an input code string and reproducing an original moving picture signal. Prediction is performed in a prediction mode selected from a plurality of prediction modes including a motion compensation prediction mode and an intraframe prediction mode for each of a plurality of blocks obtained by dividing an input image forming an input moving image signal. The prediction residual signal, which is an error signal between the prediction signal generated when prediction is performed and the input video signal, or the input video signal is quantized when prediction is performed in the intra-frame prediction mode. A code string including DCT coefficient information,
Information indicating the prediction mode of the frame is arranged in the picture header of each frame of the input image, and when the motion compensation prediction mode is selected, a plurality of pieces of information indicating the prediction mode and motion vector information related to each block are provided. A code string in which the information indicating the prediction mode combined for the plurality of blocks is information having the highest importance, and the information indicating the prediction mode combined for the plurality of blocks is arranged behind the picture header. A moving picture decoding method, comprising: inputting, decoding the inputted code string, and reproducing an original moving picture signal. Prediction is performed in a prediction mode selected from a plurality of prediction modes including a motion compensation prediction mode and an intraframe prediction mode for each of a plurality of blocks obtained by dividing an input image forming an input moving image signal. The prediction residual signal, which is an error signal between the prediction signal generated when prediction is performed and the input video signal, or the input video signal is quantized when prediction is performed in the intra-frame prediction mode. A code string including DCT coefficient information,
Information indicating the prediction mode of the frame is arranged in the picture header of each frame of the input image, and when the motion compensation prediction mode is selected, a plurality of pieces of information indicating the prediction mode and motion vector information related to each block are provided. A code obtained by coding blocks and encoding the information indicating the prediction mode combined for the plurality of blocks as the most important information and multiplexing the information indicating the prediction mode combined for the blocks after the picture header. A video decoding method, comprising: inputting a sequence, decoding the input code sequence, and reproducing an original video signal.

Images