JP4784618B2 - Moving picture encoding apparatus, moving picture decoding apparatus, moving picture encoding program, and moving picture decoding program - Google Patents

Description

  The present invention relates to a moving image encoding device, a moving image decoding device, a moving image encoding program, and a moving image decoding program, and more particularly to encoding a moving image signal by encoding a moving image signal using motion compensated prediction encoding. The present invention relates to a moving image encoding device and a moving image encoding program for obtaining a signal, and a moving image decoding device and a moving image decoding program for decoding an encoded moving image signal.

  In recent years, using information compressed by high-efficiency coding for digitized video and audio signals, devices that record long-term content on recording media, digital broadcast waves from satellites and ground stations, etc. Services for distributing content via a network have been put into practical use. In such a service, in order to broadcast and transmit a large amount of image / sound information having an enormous amount of information, high-efficiency encoding that realizes a large compression rate is required. As the high-efficiency coding of moving pictures, codes such as MPEG2 (Moving Picture Experts Group phase 2), MPEG4 (Moving Picture Experts Group phase 4), ASP (Advanced Simple Profile), MPEG4 AVC (Advanced Video Coding), etc., which are international standards. Is used. In these encoding methods, a method of compressing the amount of information using a correlation between adjacent pixels (spatial direction) of an image signal and a correlation between adjacent frames or fields (time direction) is used. As an example, an MPEG2 image encoding process will be described below.

  In MPEG2, processing is performed by dividing image frames that are continuously input in time into reference frames and predicted frames. By encoding the reference frame using only the correlation in the spatial direction, it is possible to restore only the encoded data of the frame. By encoding the prediction frame using both the correlation in the time direction and the correlation in the spatial direction from the reference frame, the encoding efficiency can be further increased with respect to the reference frame. The encoded data of the prediction frame is restored from the restored reference frame and the encoded data of the prediction frame.

  The image signal of the reference frame is divided into processing units called macroblocks of 16 pixels in the horizontal direction and 16 pixels in the vertical direction as luminance signals. The divided macroblock data is further divided into two-dimensional blocks of 8 pixels in the horizontal direction and 8 pixels in the vertical direction, and a DCT (Discrete Cosine Transform) process, which is a kind of orthogonal transform, is performed.

  Since the signal after DCT conversion shows a value according to the frequency component of the two-dimensional block, the component is concentrated in a low frequency range in a general image. In addition, information degradation of high frequency components has a property that is visually less noticeable than information degradation of low frequency components. Therefore, the amount of information is compressed by finely quantizing the low-frequency components finely and coarsely quantizing the high-frequency components and variable-length coding the continuous length of the coefficient component obtained thereby and coefficient 0 having no component.

  Similar to the reference frame, the image signal of the prediction frame is divided into macroblock units of 16 pixels in the horizontal direction and 16 pixels in the vertical direction by the luminance signal. In the prediction frame, a motion vector between the reference frame and each macroblock is obtained, and a difference block between each pixel of the macroblock and each pixel of the two-dimensional block extracted by the motion vector is generated. . When an accurate motion vector is detected, the information amount of the difference block is significantly smaller than the information amount of the original macroblock, so that a coarser quantization process than the reference frame is possible.

  Motion vector detection is generally obtained by block matching. In this block matching, each pixel of the macroblock and the reference frame of the place where the horizontal / vertical position where the macroblock exists by the motion vector value is moved are blocked into 16 pixels in the horizontal direction and 16 pixels in the vertical direction. The sum of absolute differences (or the sum of squared differences) is obtained, and the value of the motion vector that takes the minimum value is output as the detected motion vector. Depending on the movement of the screen, the image that matches the prediction frame may not exist in the reference frame, so select whether to encode the difference block or non-difference block (intra block) (prediction mode determination) ) DCT / variable length coding processing is performed on the selected block in the same manner as the reference frame to compress the information amount.

  FIG. 12 is a block diagram showing an example of a conventional moving image encoding apparatus. This moving image encoding device is a moving image encoding device based on the above-described MPEG2 encoding method. In the figure, a digital image signal that is a moving image signal to be encoded input from an input terminal 1200 is supplied to and stored in an input image memory 1201 and rearranged in the order of encoding according to an encoding syntax. Delayed to do. From the digital image signal output from the input image memory 1201, a macroblock is cut out by a two-dimensional block converter 1202.

  In the reference frame, the macroblock data output from the two-dimensional block converter 1202 is supplied to the orthogonal transformer 1204 via the subtractor 1203, and is in units of two-dimensional blocks of 8 pixels in the horizontal direction and 8 pixels in the vertical direction. DCT, which is a kind of orthogonal transform, is performed and DCT coefficients are output. The DCT coefficients are further collected by a luminance signal in units of macroblocks of 16 pixels in the horizontal direction and 16 pixels in the vertical direction, and sent to the quantizer 1205. The quantizer 1205 performs quantization processing by dividing the DCT coefficient by a different value for each coefficient by a quantization matrix having a different value for each frequency component. The quantized DCT coefficient is encoded in variable length or fixed length by referring to the address corresponding to the coefficient in the encoding table 1207 in the entropy encoder 1206. The bit stream output from the entropy encoder 1206 is stored in the stream buffer 1208 and then output to the recording medium or transmission path via the output terminal 1209.

  On the other hand, the DCT coefficients quantized by the quantizer 1205 are sequentially subjected to inverse quantization and inverse DCT processing by the inverse quantizer 1210 and the inverse orthogonal transformer 1211 to decode the quantized DCT coefficients. . The decoded DCT coefficient output from the inverse orthogonal transformer 1211 is generated as local decoded data via the adder 1212 and the two-dimensional block inverse transformer 1213, supplied to the reference image memory 1214, and stored. The local decoded data stored in the reference image memory 1214 is used as a reference image during the prediction frame encoding process.

  Subsequently, in the prediction frame, a motion between images is performed between the macroblock data cut out from the input image memory 1201 and the reference image stored in the reference image memory 1214 by a ME (motion estimation) unit 1215. A vector is required. The motion vector output from the ME unit 1215 is supplied to an MC (motion compensation) unit 1216, where a prediction block is cut out from the reference image in the reference image memory 1214. The MC unit 1216 selects an optimal prediction mode from a plurality of prediction blocks cut out, and supplies the selected prediction block to the subtracter 1203 and the adder 1212, respectively.

  The subtractor 1203 sends the difference signal between the input image block and the prediction block to the orthogonal transformer 1204. The orthogonal transformer 1204 performs the same DCT processing as the blocks of the reference frame on the difference signal, and causes the quantizer 1205 to quantize the DCT coefficients obtained thereby. The entropy encoder 1206 encodes and outputs the motion vector and prediction mode output from the MC unit 1216 together with the encoded DCT coefficient output from the quantizer 1205 based on a predetermined syntax structure. Thus, an encoded bit stream is generated. The generated encoded bit stream is output to the output terminal 1209 via the stream buffer 1208.

  On the other hand, the DCT coefficients quantized by the quantizer 1205 are subjected to inverse quantization and inverse DCT processing by the inverse quantizer 1210 and the inverse orthogonal transformer 1211, and the quantized DCT coefficients are decoded and further added. The unit 1212 adds the prediction block to generate local decoded data. In the case of an image that receives a reference, the local decoded data is supplied to the reference image memory 1214 via the two-dimensional block inverse converter 1213 and stored as a reference image. The reference image stored in the reference image memory 1214 is used at the time of the subsequent prediction frame encoding process.

  Regarding the control of the code amount, the code amount controller 1217 compares the code amount of the bit stream output from the entropy encoder 1206 to the stream buffer 1208 with the target code amount, and the target code amount. In order to approach the above, the fineness of quantization (quantization scale) of the quantizer 1205 is controlled. Since the quantization scale varies depending on the degree of information concentration in the encoding process, if the degree of concentration is not sufficient, quantization is performed with a coarse quantization scale. Decreases.

  FIG. 13 shows a block diagram of an example of a conventional video decoding device. This moving picture decoding apparatus is an apparatus that performs decoding processing according to the MPEG2 system. In the figure, a bit stream made up of an encoded signal encoded by the MPEG2 system and input to an input terminal 1300 is stored in a stream buffer 1301 and supplied to an entropy decoder 1302. The entropy decoder 1302 decodes a variable-length or fixed-length code using the encoding table 1303, and restores the quantized DCT coefficient, motion vector, and prediction mode.

  The quantized DCT coefficients are subjected to inverse quantization and inverse DCT processing in an inverse quantizer 1304 and an inverse orthogonal transformer 1305 to generate decoded data. In the case of a reference frame, it is stored in the output image memory 1310 as a decoded image signal via the adder 1306 and the two-dimensional block inverse converter 1307 and rearranged in order of output time, and then output from the output terminal 1311. The decoded image signal is also supplied to and stored in the reference image memory 1308.

  In the prediction frame, a motion vector and a mode signal output from the entropy decoder 1302 are input to the MC (motion compensation) unit 1309, and a reference image input from the reference image memory 1308 is cut out to generate a prediction block. . The generated prediction block is supplied to the adder 1306 and added to the decoded data. The decoded data added by the adder 1306 is stored in the output image memory 1310 as a decoded image signal via the two-dimensional block inverse converter 1307 and rearranged in the order of output time, similarly to the reference frame, and then output from the output terminal 1311. Is output. The decoded image signal is also supplied to and stored in the reference image memory 1308 when the corresponding frame is an image for which reference is made.

  Regardless of the MPEG2 encoding method described above, in many moving image encoding methods such as MPEG4 ASP, MPEG4 AVC, etc., motion compensation prediction is performed on a two-dimensional block image signal, and components in the time direction are applied. At the same time, the orthogonal component such as DCT is applied to the components in the spatial direction to concentrate the components in the low band, and the high frequency component is roughly quantized to compress the amount of information. In such a moving image encoding method, when encoding at a compression rate higher than necessary, the quantization scale goes up and down depending on the degree of convergence of the high frequency of the signal after DCT conversion, and there are many high frequency components. For images and images for which the accuracy of motion compensation prediction is not sufficient, the quantization scale becomes large.

  Deterioration that occurs when the quantization scale becomes large is block distortion caused by the image signal being divided into two-dimensional blocks and subjected to quantization processing, and block subjected to motion compensation prediction in units of two-dimensional blocks. Block distortion due to discontinuity in each predicted image, mosquito distortion due to coarse quantization of the high frequency component of the image signal itself, distance to the camera of the object present in the encoded image signal, The focus etc. changes with time and the object in the screen is deformed, resulting in a mismatch in motion compensation prediction, and there is distortion due to accumulation of error components at various frequency components as a result. Since the reference image signal used for the motion compensation prediction is a decoded image of the encoded signal, the deterioration component is further deteriorated by being mixed in the prediction image.

  As a technique for reducing such distortion components, a filter process for reducing block distortion is considered (for example, see Patent Document 1). In this patent document 1, in order to reduce the deterioration visually by limiting to the block distortion in the deterioration component, the decoding apparatus stores the decoded image stored in the output image memory 1310 in FIG. First, the difference between the pixels at the block boundaries as shown in FIG. 14A is calculated. In FIG. 14A, the target block is C, and adjacent blocks adjacent to the left side, upper side, right side, and lower side of the target block C are L, U, R, and D, respectively. In the decoding device described in Patent Literature 1, the absolute value of the difference between the pixels C00, C10,..., C70 of the target block C and the pixels L07, L17,. , C07 of the target block C and the absolute value of the difference between the pixels U70, U71,..., U77 of the adjacent block U, and the pixels C07, C17 of the target block C ,..., C77 and pixels R00, R10,..., R70 of adjacent blocks R, and absolute values of differences between the pixels, and pixels C70, C71,. Difference absolute values between the pixels D00, D02,..., D07 of the block D are obtained.

  Subsequently, the decoding device described above compares the obtained inter-pixel difference absolute value at the boundary of the block and a predetermined threshold value TH, and when all the difference absolute values are less than the threshold value TH, A low-pass filter process is performed to average the boundary determined that the block distortion is severe, and a filtered decoded image is output from the output terminal 1211. Here, in the low-pass filter processing described above, for example, at the block boundary between the pixel C71 of the target block C and the pixel D01 of the adjacent block D, as shown in FIG. Then, an average value of the values of a total of five pixels of the two pixels C61 and C51 adjacent in the upward direction, the pixel D01 of the adjacent block D and the pixel D11 adjacent in the downward direction is obtained, and as shown in FIG. The average value is set as a value C71 ′ of the pixel C71 of the target block C.

  Furthermore, as a block distortion reduction filter considering discontinuity between blocks in the predicted image by motion compensation prediction, unnecessary filter processing is performed by evaluating the continuity between blocks in motion compensation prediction and changing the strength of the filter processing. There is also disclosed a technique for performing an operation that does not apply (see, for example, Patent Document 2). In the MPEG4 AVC standard, the process as described in Patent Document 2 is defined so that it can be performed within the encoding process.

Japanese Patent Laid-Open No. 10-51775 Japanese Patent No. 3688283

  However, in the decoding device described in Patent Document 1 described above, the distortion of the block boundary that occurs when the DCT component is quantized is performed by smoothing the image signal for pixels near the boundary between the blocks, thereby discontinuous. Although we are trying to alleviate the state, the distortion component evaluation is judged based on the difference between the decoded images, and the discontinuity between blocks may not be alleviated sufficiently due to the strength of smoothing. .

  For example, when the smoothing strength is lower than the required strength, the discontinuity between blocks cannot be improved sufficiently, and the influence still remains as block distortion. In addition, when the smoothing strength is higher than the required strength, discontinuity of the block boundary portion is improved, but the texture information of the block boundary portion is too smoothed. Degradation of the resolution of the decoded image occurs and necessary texture information is lost.

  In the decoding device described in Patent Document 2, the continuity of the predicted image is determined based on the amount of motion used in motion compensation prediction and the identity of the reference image used, and the smoothing strength is changed. ing. For this reason, unnecessary smoothing of the decoded image when no prediction error occurs can be avoided, but it can cope with block distortion caused by quantization of the DCT component of the error when the prediction error occurs. However, the evaluation of the distortion component is judged from the visual detection of the decoded image in the same manner as the decoding device described in Patent Document 1, and the distortion removal that restores the component originally present in the image signal is performed. It has not been.

  From the above, in video encoding and decoding using motion compensation prediction processing, in order to obtain a high-quality decoded image with a high compression rate, distortion component evaluation is based on judgment from visual detection of the decoded image. In addition to realizing distortion removal in a form that restores the components that originally existed in the image signal, eliminating the detrimental effect of resolution degradation due to processing that makes the block boundary distortion components inconspicuous done with the conventional method, There is a need for a method for removing distortion components caused by mosquito distortion and motion compensation mismatch that are not adequately addressed.

  The present invention has been made in view of the above points, and reduces coding distortion caused by a prediction error component in the case of using motion compensation prediction coding without performing smoothing processing on a decoded image signal, and eliminates it. Moving image encoding apparatus, moving image decoding apparatus, moving image encoding program, and moving image decoding program, which can restore a high-resolution component and a motion compensation mismatch component and obtain a decoded image with little deterioration The purpose is to provide.

  In order to achieve the above object, according to a first aspect of the present invention, an input moving image signal to be encoded is divided into rectangular areas each having a predetermined number of pixels, and the rectangular area is processed in the past as a processing unit. The motion compensation prediction process is performed using the decoded image signal of the obtained image as a reference image, and the difference signal between the encoding target image that is the encoding target moving image signal and the motion compensation prediction signal is defined separately from the rectangular area. In a moving image encoding apparatus that performs encoding by sequentially performing orthogonal transform processing and quantization processing on a rectangular region as a processing unit, a motion compensated prediction signal and a result of performing orthogonal transform processing and quantization processing are generated. Orthogonal transform coefficient estimation that receives as input the decoded orthogonal transform coefficient generated by dequantizing the quantized data, and estimates the orthogonal transform coefficient lost as a result of quantization and outputs the estimated orthogonal transform coefficient means , Receiving as input the estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation means, the decoded orthogonal transform coefficient, and the quantized value for each coefficient indicating the fineness of the quantization when the quantization processing is performed, The region information that can be taken by the orthogonal transform coefficient before quantization is calculated from the quantized value, and based on the region information, it is determined whether to replace the decoded orthogonal transform coefficient with the estimated orthogonal transform coefficient. Coefficient correction means for generating, and correction image signal generation for generating a corrected decoded image signal by adding a signal obtained by performing an inverse orthogonal transform process to the corrected orthogonal transform coefficient output from the coefficient correction means and a motion compensated prediction signal And the corrected decoded image signal is used as a reference image of a moving image signal to be encoded subsequently.

  In order to achieve the above object, the second invention divides an input moving image signal to be encoded into rectangular areas each having a predetermined number of pixels, and encodes the past using the rectangular area as a processing unit. The motion compensation prediction process is performed using the decoded image signal of the converted image as a reference image, and the difference signal between the encoding target image that is the encoding target moving image signal and the motion compensation prediction signal is separated from the rectangular area. In a moving picture decoding apparatus that decodes a bit stream of a moving picture signal that has been encoded by sequentially performing orthogonal transform processing and quantization processing using a defined rectangular area as a processing unit, the quantized orthogonality from the bit stream Based on the entropy decoding means for decoding the transform coefficient and the motion vector, and based on the motion vector from the entropy decoding means, the past decoded image signal is cut out as a reference signal and motion compensation prediction is performed. A motion compensation means for generating a signal, a decoded orthogonal transform coefficient generated by dequantizing the quantized orthogonal transform coefficient output from the entropy decoding means, and a motion compensated prediction signal as inputs; The orthogonal transform coefficient estimation means for estimating the orthogonal transform coefficient lost as a result of the conversion and outputting the estimated orthogonal transform coefficient, the estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation means, the decoded orthogonal transform coefficient, and the quantization It receives as input the quantized value for each coefficient indicating the fineness of quantization when processing is performed, calculates area information that can be taken by the orthogonal transform coefficient before quantization from the quantized value, and calculates the area information Based on whether or not to replace the decoded orthogonal transform coefficient with the estimated orthogonal transform coefficient to generate a corrected orthogonal transform coefficient, and inverse orthogonal transform processing to the corrected orthogonal transform coefficient output from the coefficient correction means And a corrected decoded image signal generating means for generating a corrected decoded image signal by adding a motion compensated prediction signal to the signal obtained by performing the correction, and the corrected decoded image signal is subsequently converted into a moving image signal to be encoded. It is used as a reference image.

  In order to achieve the above object, the third invention provides an entropy decoding means for decoding orthogonal transform coefficients and motion vectors quantized from a bitstream in the same video decoding apparatus as the second invention. Based on the motion vector from the entropy decoding means, a motion compensation means for extracting a past decoded image signal as a reference signal to generate a motion compensated prediction signal, and an inverse of the quantized orthogonal transform coefficient output from the entropy decoding means Orthogonal transform coefficient estimation that receives the decoded orthogonal transform coefficient generated by quantization and the motion compensated prediction signal as input, estimates the orthogonal transform coefficient lost as a result of quantization, and outputs the estimated orthogonal transform coefficient , The estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation means, the decoded orthogonal transform coefficient, and the quantization details when the quantization process is performed The region information that can be taken by the orthogonal transform coefficient before quantization is calculated from the quantized value, and the decoded orthogonal transform coefficient is converted into the estimated orthogonal transform coefficient based on the region information. For the correction difference orthogonal transform coefficient output from the coefficient correction difference generation means that outputs the correction difference orthogonal transform coefficient that is the difference information between the replacement coefficient and the orthogonal transform coefficient Inverse orthogonal transform means for transforming into a corrected differential signal by performing inverse orthogonal transform processing, and a decoded image signal decoded using a signal obtained by inverse orthogonal transform of a decoded orthogonal transform coefficient and a motion compensated prediction signal, as a reference image And then adding the corrected difference signal only to a decoded image signal output as an output image and outputting a decoded image signal corrected for the reproduced image. Characterized in that it has and.

  In order to achieve the above object, according to a fourth aspect of the present invention, an input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and encoding is performed in the past using the rectangular region as a processing unit. The motion compensation prediction process is performed using the decoded image signal of the converted image as a reference image, and the difference signal between the encoding target image that is the encoding target moving image signal and the motion compensation prediction signal is separated from the rectangular area. In a moving image encoding program for causing a computer to execute encoding by sequentially performing orthogonal transform processing and quantization processing using a defined rectangular area as a processing unit, the computer is connected to the orthogonal transform according to the configuration of the first invention. It functions as a coefficient estimation unit, a coefficient correction unit, and a corrected image signal generation unit, and uses the corrected decoded image signal as a reference image of a moving image signal to be encoded subsequently. To.

  In order to achieve the above object, according to a fifth aspect of the present invention, an input moving image signal to be encoded is divided into rectangular areas each having a predetermined number of pixels, and encoding is performed in the past using the rectangular area as a processing unit. The motion compensation prediction process is performed using the decoded image signal of the converted image as a reference image, and the difference signal between the encoding target image that is the encoding target moving image signal and the motion compensation prediction signal is separated from the rectangular area. In a moving picture decoding program for decoding a bit stream of a moving picture signal encoded by sequentially performing orthogonal transform processing and quantization processing on a defined rectangular area as a processing unit, and causing a computer to execute moving picture decoding The computer includes the entropy decoding means, the motion compensation means, the decoded orthogonal transform coefficient, the orthogonal transform coefficient estimation means, the coefficient correction means, the corrected decoded image signal of the second invention. To function as a generator, the correction decoded image signal, subsequently characterized by using as a reference image of the moving image signal to be encoded.

  In order to achieve the above object, according to a sixth aspect of the present invention, there is provided a moving picture decoding program for causing a computer to execute the same moving picture decoding as in the third aspect of the invention. Means, motion compensation means, orthogonal transform coefficient estimation means, coefficient correction difference generation means, inverse orthogonal transform means, and image signal decoding means.

  According to the present invention, the prediction error signal subjected to the motion compensation prediction process is estimated to be the degradation component generated by the orthogonal transform process and the quantization process, and the truncated component of the quantized prediction error orthogonal transform signal is estimated. The deterioration component can be corrected. As a result, it is not a method that simply smoothes the decoded image signal to make it visually inconspicuous, but also eliminates block distortion caused by discontinuous states between blocks, as well as lost high-resolution components and motion compensation mismatches. It is possible to restore the components and obtain a decoded image with little deterioration.

  Further, according to the present invention, when estimating the decoding orthogonal transform coefficient, it is expressed by a predetermined source model such as a quadratic function that defines the behavior of the signal in the block, and the signal in the block is used as the boundary condition of the block. By using the technology to generate the estimated signal based on the signal characteristics that the entire block holds with the distortion component between the blocks removed, the significant estimated signal that is selected and replaced by the region information can be Can be generated and the correction process can function more appropriately. As a result, according to the present invention, it is possible to provide a moving picture coding apparatus and a moving picture coding program that are less deteriorated even when the compression rate is increased than before.

  Furthermore, according to the present invention, it is possible to provide a moving picture decoding apparatus and a moving picture decoding program capable of reproducing transmitted information with a higher compression rate than before with less deterioration.

  Next, the best mode for carrying out the invention will be described with reference to the drawings.

(First embodiment)
First, a first embodiment of the video encoding device and video decoding device of the present invention will be described. FIG. 1 shows a block diagram of a first embodiment of a video encoding apparatus according to the present invention.

  As shown in FIG. 1, the encoding apparatus according to the present embodiment includes an input terminal 100, an input image memory 101, a two-dimensional block transformer 102, a subtractor 103, an orthogonal transformer 104, a quantizer 105, and entropy coding. 106, coding table 107, stream buffer 108, output terminal 109, inverse quantizer 110, inverse orthogonal transformer 111, adder 112, two-dimensional block inverse transformer 113, reference image memory 114, ME (motion estimation) 115, MC (motion compensation) unit 116, code amount controller 117, orthogonal transform coefficient estimator 118, coefficient corrector 119, and image buffer 120. Regarding the code amount controller 117 from the input terminal 100, the same configuration as the code amount controller 1217 from the input terminal 1200 of FIG. The present embodiment is characterized in that an orthogonal transform coefficient estimator 118 and a coefficient corrector 119 are additionally provided.

  Next, the operation of the moving picture coding apparatus according to the present embodiment will be described. In FIG. 1, a digital image signal that is a moving image signal to be encoded input from an input terminal 100 is supplied to and stored in an input image memory 101, and macroblock data is cut out by a two-dimensional block converter 102. The macroblock data is supplied to the orthogonal transformer 104 via the subtractor 103, and DCT, which is a kind of orthogonal transformation, is performed on a two-dimensional block unit of 8 pixels in the horizontal direction and 8 pixels in the vertical direction to output DCT coefficients. . The DCT coefficients are further collected as a luminance signal in units of macroblocks of 16 pixels in the horizontal direction and 16 pixels in the vertical direction, and sent to the quantizer 105.

  The quantizer 105 performs a quantization process on the DCT coefficient. The entropy encoder 106 refers to the encoding table 107 for the quantized DCT coefficient output from the quantizer 105 and performs variable length or fixed length encoding to generate a bitstream. . The bit stream output from the entropy encoder 106 is output from the output terminal 109 via the stream buffer 108.

  On the other hand, the DCT coefficient quantized in the quantizer 105 is also subjected to inverse quantization processing in the inverse quantizer 110, and the DCT coefficient is decoded. The decoded DCT coefficient is input to the orthogonal transform coefficient estimator 118 and the coefficient corrector 119. The orthogonal transform coefficient estimator 118 is a coefficient component in which the quantized DCT coefficient is lost from the decoded DCT coefficient. And the estimated DCT coefficients are output. The coefficient corrector 119 uses the estimated DCT coefficient output from the orthogonal transform coefficient estimator 118 and the decoded DCT coefficient output from the inverse quantizer 110 to determine the DCT coefficient possessed by the input image signal. The component lost by quantization is judged and the coefficient is corrected. Detailed operation will be described later.

  The corrected DCT coefficient output from the coefficient corrector 119 is subjected to inverse DCT processing by the inverse orthogonal transformer 111, and the decoded image signal is added to the reference image memory via the adder 112 and the two-dimensional block inverse transformer 113. 114 and stored as a reference image. In the reference frame, since the prediction block input to the adder 112 is not generated by the MC unit 116, the image buffer 120 does not function. The decoded image signal stored in the reference image memory 114 is used as a reference image during the prediction frame encoding process.

  Subsequently, in the prediction frame, the ME (motion estimator) 115 obtains a motion vector between the macroblock data cut out from the input image memory 101 and the image stored in the reference image memory 114. The MC (motion compensation) unit 116 extracts a prediction block from the reference image in the reference image memory 114 based on the motion vector obtained by the ME unit 115. Further, the MC unit 116 selects an optimal prediction mode from a plurality of prediction blocks cut out, and supplies the selected prediction block to the subtractor 103, the image buffer 120, and the orthogonal transform coefficient estimator 118, respectively.

  The subtracter 103 sends the difference signal between the input image block and the prediction block to the orthogonal transformer 104. The orthogonal transformer 104 performs DCT processing similar to that of each block of the reference frame on the difference signal, and causes the quantizer 105 to quantize the prediction difference DCT coefficient obtained thereby. The entropy encoder 106 encodes and outputs the motion vector and prediction mode output from the MC unit 116 together with the encoded DCT coefficient output from the quantizer 105 based on a predetermined syntax structure. Thus, an encoded bit stream is generated. The generated encoded bit stream is output to the output terminal 109 via the stream buffer 108.

  On the other hand, the prediction difference DCT coefficient quantized by the quantizer 105 is subjected to inverse quantization processing by the inverse quantizer 110, and the prediction difference DCT coefficient is decoded. The decoded prediction difference DCT coefficient is input to the orthogonal transform coefficient estimator 118 and the coefficient corrector 119. The orthogonal transform coefficient estimator 118 is supplied together with the decoded prediction difference DCT coefficient supplied from the inverse quantizer 110. The prediction block is received from the MC unit 116, the coefficient component in which the quantized differential DCT coefficient is lost is estimated, and the estimated differential DCT coefficient is output. The coefficient corrector 119 quantizes the DCT coefficient held by the motion compensated prediction difference signal from the difference DCT coefficient estimated by the orthogonal transform coefficient estimator 118 and the difference DCT coefficient decoded by the inverse quantizer 110. The lost component is judged and the coefficient is corrected. Details of this processing will also be described later.

  The corrected differential DCT coefficient output from the coefficient corrector 119 is subjected to inverse DCT processing by the inverse orthogonal transformer 111, and becomes a decoded predicted difference signal. The image buffer 120 is a buffer for absorbing the delay required for the coefficient estimation processing in the orthogonal transform coefficient estimator 118. The image buffer 120 delays the prediction block signal output from the MC unit 116 for a fixed time, and sends it to the adder 112. Output. The size of the image buffer 120 depends on the delay required by the orthogonal transform coefficient estimator 118, and the image buffer 120 can be deleted when performing an estimation process that does not cause a delay. The adder 112 adds the decoded prediction difference signal obtained by the inverse orthogonal transformer 111 and the prediction block delayed by the image buffer 120 to generate and output a decoded image signal.

  In the case of an image that receives a reference, the decoded image signal output from the adder 112 is supplied to the reference image memory 114 via the two-dimensional block inverse transformer 113 and stored as a reference image. The reference image stored in the reference image memory 114 is used at the time of encoding processing of the subsequent prediction frame. In the case of a non-referenced prediction frame, the decoded image signal generation process after the inverse quantizer 110 may not function.

  Next, the orthogonal transform coefficient estimator 118 will be described in more detail. FIG. 2 shows a block diagram of an embodiment of orthogonal transform coefficient estimator 118. In the figure, an orthogonal transform coefficient estimator 118 receives a motion compensated prediction signal from the MC unit 116 and a prediction difference DCT coefficient decoded from the inverse quantizer 110. The motion compensated prediction signal is subjected to orthogonal transform (DCT which is a kind of orthogonal transform in this case) in units of two-dimensional blocks of 8 pixels in the horizontal direction and 8 pixels in the vertical direction in the orthogonal transformer 200 and the predicted DCT coefficients. Is output. The adder 201 generates a decoded DCT coefficient by adding the predicted DCT coefficient output from the orthogonal transformer 200 and the decoded predicted difference DCT coefficient output from the inverse quantizer 110 in coefficient units. . The decoded DCT coefficient has a component equivalent to the result obtained by performing orthogonal transformation on the decoded image signal. The decoded DCT coefficients are delayed in the coefficient buffer 202 by a macroblock period corresponding to the horizontal direction. In this embodiment, by accumulating macroblock coefficients for the horizontal direction, decoded DCT coefficients for the DCT blocks on the top, bottom, left, and right of the target DCT block are obtained.

  From the coefficient buffer 202, the target DCT block and the decoded DCT coefficients of the upper, lower, left, and right DCT blocks are input to the slope-cancellation orthogonal transform coefficient generator 203. The slope cancellation orthogonal transform coefficient generator 203 generates a DCT coefficient of a block boundary slope cancellation function for smoothly connecting the block boundaries between the image signal of the target block and the image signal of the adjacent block.

  As a technique for generating a DCT coefficient of a block boundary slope cancellation function, in this embodiment, a prediction signal in a block is estimated by applying the slope of an image signal at a block boundary to a Poisson equation as a boundary condition, and the DCT Apply the estimation technique used in the multiple harmonic local cosine transform method to generate the coefficients. Next, the multiple harmonic local cosine transform method will be described.

  FIG. 3 is a conceptual diagram for illustrating the basic concept of the multiple harmonic local cosine transform method. The multiple harmonic local cosine transform method improves the coding efficiency of DCT used in JPEG (Joint Photographic Experts Group), etc. as described in well-known literature (Improvement of DCT-based Compression Algorithms Using Poisson's Equation) It is a technique to do. FIG. 3 shows a particular block in an original image frame in which attention is paid to a particular line in the block, and the state of the original signal on the noticed line and the boundary of the original signal on the block boundary is shown in FIG. Shows how it changes.

  With respect to the original signal 301 shown in FIG. 3A, the inclinations 302 and 303 of the image signal at both ends of the original signal 301 corresponding to the block boundaries 304 and 305 are obtained. The slopes 302 and 303 of the image signals at both ends are used as boundary conditions. The multiharmonic local cosine transform method defines a predetermined source model such as a quadratic function that defines the behavior of a signal in a block, expresses a signal in the block by the predetermined source model, and expresses it by a predetermined source model. In this method, generation of an estimated signal based on the boundary condition of the block can be derived from DCT (discrete cosine transform) coefficients.

  Here, the estimation signal 306 is generated using the gradients of the image signals at both ends of the original signal 301 as shown in FIG. As the predetermined source model, it is desirable to adopt a model that can estimate a signal closest to the original signal under boundary conditions. In general, in the case of a one-dimensional signal as shown in FIG. 3, it can be easily obtained by applying a quadratic function as a predetermined source model, but here it is not particularly limited to this model. It may be a low-order function such as a linear function or a higher-order function such as a cubic function or a quartic function. As described above, in the present embodiment, by introducing a predetermined source model, an estimated signal in a block is generated analytically without spending a large amount of calculation and solving the Poisson equation numerically.

  In the multiharmonic local cosine transform method, after generating an estimated signal 306 based on the boundary condition as shown in FIG. 3B, a difference is obtained between the original signal 301 and the estimated signal 306, so that FIG. A residual signal 307 as shown in c) is generated. The residual signal 307 thus obtained is subjected to normal orthogonal transform DCT, quantization, and entropy encoding, thereby reducing the amount of code generated compared to the case of performing normal DCT encoding. It is possible to suppress.

  In the present embodiment, estimation processing within a block in the multiple harmonic local cosine transform method is performed on the decoded image signal to estimate a signal close to the input original signal, and motion compensation prediction is performed from the estimated signal. By subtracting the signal, the present invention is applied to a process of restoring the originally predicted difference DCT coefficient that has been truncated by the quantization process.

  Next, specific processing in the slope cancellation orthogonal transform coefficient generator 203 will be described. First, an image signal of a moving image frame is captured as each block shown in FIG. 4 expressed by Equation 1 with respect to an input moving image frame (in this embodiment, a decoded image signal is indicated).

図４のΩ^(0,0)が現在処理対象となっているブロックであるものとする。また、各ブロック内の各画素は、数２で表現される。

^Assume that Ω ^{(0,0) in} FIG. 4 is the block currently being processed. In addition, each pixel in each block is expressed by Equation 2.

続いて、図４に示すブロックの境界部分Γ⁽¹⁾、Γ^（2）、Γ^(3）、Γ⁽⁴⁾の境界条件である画像信号の傾きを周波数領域で議論することができるように、ＤＣＴ級数で表現する。ここでは数３に示した境界条件のＤＣＴ級数展開の式を利用して画像信号の傾きをｇとした場合に数３に示した式中のＧ_ｋのようなＤＣＴ係数で表現する。

Subsequently, the inclination of the image signal, which is the boundary condition of the block boundary portions Γ ⁽¹⁾ , Γ ⁽²⁾ , Γ ⁽³⁾ , Γ ⁽⁴⁾ shown in FIG. 4, can be discussed in the frequency domain. , Expressed as a DCT series. Here, when the gradient of the image signal is set to g using the DCT series expansion formula of the boundary condition shown in Formula 3, it is expressed by a DCT coefficient such as G _{k in} the formula shown in Formula 3.

また、ポアソン方程式の概念を導入してブロック内の原信号に対する推定信号を求める。ここで、ポアソン方程式は、処理対象ブロックＱｊにおいて、ブロック内の推定信号ｕのラプラシアンであるΔｕ_ｊがソース項Ｋ_ｊとの間において、次式が成立する方程式である。

In addition, the concept of the Poisson equation is introduced to obtain an estimated signal for the original signal in the block. Here, the Poisson equation is an equation in the processing target block Qj in which Δu _j, which is a Laplacian of the estimation signal u in the block, and the source term K _j hold the following equation.

Δu _j = K _j
The estimated signal u in the block can be expressed as shown in Equation 5 using Neumann boundary conditions and DCT series notation. This can be obtained by adding the DCT series expansion components of the estimated signal from each boundary as shown in Equation 4.

このような推定信号ｕに対してＤＣＴを行うことで、数７を伴う数６で示されるような推定信号Ｕを得る。

By performing DCT on such an estimated signal u, an estimated signal U as shown in Expression 6 accompanied by Expression 7 is obtained.

数７に示した式より、ηおよびη^*はブロック内の位置情報のみに依存し、ブロック内の画像信号には依存しないことから、予め一意に計算して求めておくことができる。従って、推定信号Ｕを求めるためには、傾きのＤＣＴ係数情報Ｇ_kを求めることが重要となる。

From the equation shown in Equation 7, since η and η ^* depend only on the position information in the block and do not depend on the image signal in the block, they can be uniquely calculated in advance. Therefore, in order to obtain the estimated signal U, it is important to obtain the slope DCT coefficient information G _k .

Here, in order to make it possible to obtain G _k by using DCT coefficient information after performing DCT on the original signal in the block, when obtaining the inclination of the image signal at the block boundary, which is a boundary condition. The slope between the blocks is calculated using the average value of the signal components in each direction in the block as represented by the equation (8) as a representative value, and g is obtained by approximating this slope as the slope at the block boundary.

続いて、この近似されたｇに対してＤＣＴを行うことでＧを求める。ここで、傾きのＤＣＴ係数情報Ｇを求める際に数８に示した式を考慮して数９に示した式変形を行うことで、Ｇを数９に示した式中のＦ、つまりブロック内のＤＣＴ係数情報を利用して求めることが可能となる。

Subsequently, G is obtained by performing DCT on the approximated g. Here, when obtaining the DCT coefficient information G of the slope, the equation shown in Equation 9 is taken into consideration when the equation shown in Equation 8 is taken into consideration, so that G in the equation shown in Equation 9, that is, in the block It is possible to obtain using the DCT coefficient information.

傾き相殺直交変換係数生成器２０３においては、係数バッファ２０２より入力されたＤＣＴ係数情報をＦとして数９に示した式から傾きのＤＣＴ係数情報Ｇを求めることで、推定信号ｕのＤＣＴ係数Ｕを生成する処理が施される。

In the inclination canceling orthogonal transform coefficient generator 203, the DCT coefficient information G inputted from the coefficient buffer 202 is set as F, and the DCT coefficient information G of the estimated signal u is obtained from the equation shown in Equation 9 to obtain the DCT coefficient U of the estimated signal u. Processing to generate is performed.

  Specifically, as shown in FIG. 5, the target decoding DCT block 501 as a target and the decoding DCT blocks 502, 503, 504 adjacent to the left, top, right, and bottom of the target decoding DCT block 501, respectively. A process of calculating G with the coefficient values of the DC term in the two-dimensional DCT and the DC term in the one-dimensional DCT and calculating the estimated signal DCT coefficient U excluding the DC term of the target decoding DCT block 501 to be the target 505. Applied.

  Here, in FIG. 5, c00 is the DC term of the two-dimensional DCT of the target decoding DCT block 501, and c01 to c07 arranged in the horizontal direction are DC terms of the vertical one-dimensional DCT of the decoding DCT block 501, and c10 to c70 are arranged in the vertical direction. Indicates the DC term of the horizontal one-dimensional DCT of the decoded DCT block 501. Further, l00, u00, r00, and d00 indicate DC terms of the two-dimensional DCT of adjacent decoding DCT blocks 502, 503, 504, and 505. Also, l01 to l07, u01 to u07, r01 to r07, and d01 to d07 arranged in the horizontal direction are respectively the DC terms of the vertical one-dimensional DCT of adjacent decoding DCT blocks 502, 503, 504, and 505, and c10 arranged in the vertical direction. c70, u10 to u70, r10 to r70, and d10 to d70 are DC terms of the horizontal one-dimensional DCT of adjacent decoding DCT blocks 502, 503, 504, and 505, respectively.

  Returning to FIG. The estimated signal DCT coefficient output from the slope cancellation orthogonal transform coefficient generator 203 is input to the subtractor 205 together with the predicted DCT coefficient output from the coefficient buffer 204, and the predicted DCT coefficient is subtracted from the estimated signal DCT coefficient in coefficient units. And output to the coefficient corrector 119 as the estimated prediction difference DCT coefficient.

  In the case of the image signal of the reference frame, since the decoded DCT coefficient is input from the inverse quantizer 110 and the motion compensated prediction signal is not input from the MC unit 116, the slope cancellation orthogonal transform coefficient generator 203 receives the adder. 201, the motion compensated prediction DCT coefficient is not input to the subtracter 205, and the estimated DCT coefficient is output to the coefficient corrector 119.

  Next, the operation of the coefficient corrector 119 will be described. In the case of the reference frame, the coefficient corrector 119 receives the decoded DCT coefficient from the inverse quantizer 110 and the estimated DCT coefficient from the orthogonal transform coefficient estimator 118. When motion compensated prediction is used, the decoded DCT coefficient is a decoded prediction difference DCT coefficient, and the estimated DCT coefficient is an estimated prediction difference DCT coefficient.

  The coefficient corrector 119 performs a delay necessary for the orthogonal transform coefficient estimator 118 to calculate the estimated prediction difference DCT coefficient for the decoded prediction difference DCT coefficient, matches the processing timing of the target DCT block, and decodes the prediction difference The DCT coefficient and the estimated prediction difference DCT coefficient are compared for each coefficient, and which DCT coefficient is to be applied is selected.

  An algorithm for selecting a coefficient in units of DCT blocks in the coefficient corrector 119 will be described with reference to the flowchart of FIG. First, the coefficient corrector 119 obtains a decoded DCT coefficient DecC and an estimated DCT coefficient EstC for a target DCT block (step S600).

  Subsequently, a quantization step Q and a quantization matrix M when the target DCT block is quantized are obtained (step S601). Next, the coefficient corrector 119 calculates a possible range for each coefficient term of the DCT block before the decoded DCT coefficient is quantized. In MPEG2 coding, for DCT blocks using motion compensated prediction, DCT coefficients are rounded off when quantized, motion compensated prediction is not used, and the image signal itself is DCT processed. Processing is performed. Therefore, it is determined whether the target DCT block is a prediction difference block using motion compensated prediction (step S602).

  When the target DCT block is the prediction difference DCT, the coefficient corrector 119 converts the horizontal x and vertical y coefficients in the decoded prediction difference DCT into a quantization matrix of terms of DecC (x, y), horizontal x, and vertical y. If the value is M (x, y), the range of absolute values of coefficients that can be taken before quantization is calculated as follows (step S603).

Minimum value EstCmin = abs (DecC (x, y))
Maximum value EstCmax = abs (DecC (x, y)) + Q * M (x, y)
On the other hand, when the target DCT block is not the prediction difference DCT, the coefficient corrector 119 performs the following calculation (step S604).

Minimum value EstCmin = 0: When DecC (x, y) = 0
abs (DecC (x, y))-Q * M (x, y) / 2
: When DecC (x, y) ≠ 0, the maximum value EstCmax = abs (DecC (x, y)) + Q * M (x, y) / 2
Subsequently, the coefficient corrector 119 satisfies a condition satisfying that the maximum value, the minimum value, and the estimated DCT coefficient calculated in step S603 or S604 are within the range of coefficients that can be taken before the following quantization. (Step S605), and if the condition is satisfied, the estimated DCT coefficient EstC (x, y) is selected as the corrected DCT coefficient CorC (x, y) (step S606). If not satisfied, the decoded DCT coefficient DecC (x, y) is selected as the corrected DCT coefficient CorC (x, y) (step S607).

  Here, the condition that satisfies the above-described range of the coefficients that can be taken is that all of the following (1), (2), and (3) are simultaneously satisfied.

(1) The polarity of the estimated DCT coefficient EstC (x, y) and the decoded DCT coefficient DecC (x, y) are the same, or DecC (x, y) = 0
(2) EstCmin ≦ abs (EstC (x, y)) <EstCmax
(3) Not DC term (x = 0, y = 0) Finally, the coefficient corrector 119 selects the decoded DCT as the DC term of the DCT block, and for all other coefficient terms, in step S606 or S607 The obtained coefficients of DecC (x, y) and EstC (x, y) are stored as corrected DCT coefficients CorC (x, y) (step S608). The coefficient corrector 119 calculates corrected DCT coefficients for all the DCT blocks in the macroblock according to the above algorithm, and outputs the corrected DCT coefficients to the inverse orthogonal transformer 111 in FIG.

  In the present embodiment, there is provided a mechanism for estimating and correcting a prediction residual component truncated by quantization in the DCT region of the prediction residual component. Since the range of the component cut in the DCT region can be estimated from the quantization scale, whether or not the estimated DCT coefficient is in an appropriate correction amount range is determined individually for each component of the decoded DCT coefficient. With regard to the DCT coefficient estimation process, an assumed signal from which the distortion component of the decoded image is removed is generated, and a DCT coefficient to be output for decoding the signal and a prediction difference DCT coefficient are calculated. Realize.

  In the present embodiment, a method for specifying an estimation signal based on block boundary continuity between an image signal of a target block and an image signal of an adjacent block is used, but other decoded signal estimation techniques can also be used. It is. In this embodiment, the validity of the estimated signal is realized by confirming that it exists in the quantization error range of the DCT coefficient and the prediction difference DCT coefficient subjected to the quantization process, Therefore, the deteriorated component can be restored without changing unnecessary components.

  Next, the moving picture decoding apparatus according to the present invention will be described. FIG. 7 shows a block diagram of the first embodiment of the moving picture decoding apparatus according to the present invention. As shown in the figure, the decoding apparatus according to the present embodiment includes an input terminal 700, a stream buffer 701, an entropy decoder 702, a coding table 703, an inverse quantizer 704, an inverse orthogonal transformer 705, and an adder 706. It comprises a two-dimensional block inverse transformer 707, a reference image memory 708, an MC (motion compensation) unit 709, an output image memory 710, an output terminal 711, an orthogonal transform coefficient estimator 712, a coefficient corrector 713, and an image buffer 714. The

  With respect to each part from the input terminal 700 to the output terminal 711, the same configuration as that from the input terminal 1300 to the output terminal 1311 of the conventional video decoding device in FIG. 13 is used. The present embodiment is characterized in that an orthogonal transform coefficient estimator 712, a coefficient corrector 713, and an image buffer 714 are provided in addition to the configuration of the conventional moving picture decoding apparatus.

  The bit stream generated by the moving picture coding apparatus in FIG. 1 or the moving picture coding program described later input to the input terminal 700 is stored in the stream buffer 701 and supplied to the entropy decoder 702. The entropy decoder 702 decodes a variable-length or fixed-length code using the encoding table 703 and restores the quantized DCT coefficient, motion vector, and prediction mode. The quantized DCT coefficient is subjected to an inverse quantization process by an inverse quantizer 704, and a decoded DCT coefficient is calculated.

  The decoded DCT coefficient is input to an orthogonal transform coefficient estimator 712 and a coefficient corrector 713. The orthogonal transform coefficient estimator 712 estimates a coefficient component in which the DCT coefficient is lost due to quantization during encoding, and outputs the estimated DCT coefficient. The coefficient corrector 713 selects a component in which the DCT coefficient of the encoded image signal is lost from the estimated DCT coefficient and the decoded DCT coefficient, and corrects it to the decoded DCT coefficient. To calculate a corrected DCT coefficient.

  The corrected DCT coefficient output from the coefficient corrector 713 is subjected to inverse DCT processing by an inverse orthogonal transformer 705, and the decoded image signal is added to a reference image memory via an adder 706 and a two-dimensional block inverse transformer 707. 708 to be stored. In the reference frame, since the prediction block input to the adder 706 via the image buffer 714 is not generated by the MC unit 709, the image buffer 714 does not function.

  The decoded image signal stored in the reference image memory 708 is used when the prediction frame is encoded. Also, the decoded image signal is stored in the output image memory 710, rearranged in the order of output time, and output from the output terminal 711, whereby the encoded moving image signal can be reproduced.

  In the prediction frame, the prediction block is generated by inputting the motion vector and the mode signal output from the entropy decoder 702 to the MC unit 709 and cutting out the reference image from the reference image memory 708. The generated prediction block is supplied to the image buffer 714 and the orthogonal transform coefficient estimator 712. In the prediction frame, the decoded differential DCT coefficient is output from the inverse quantizer 704 and sent to the orthogonal transform coefficient estimator 712 and the coefficient corrector 713.

  The orthogonal transform coefficient estimator 712 receives the prediction block from the MC unit 709 together with the decoded differential DCT coefficient, estimates the coefficient component in which the quantized differential DCT coefficient is lost, and outputs the estimated differential DCT coefficient To do. The coefficient corrector 713 selects, from the estimated differential DCT coefficient and the decoded differential DCT coefficient, a component in which the DCT coefficient possessed by the motion compensated prediction differential signal has been lost due to quantization, and determines the decoded differential DCT coefficient. Correction is applied and a correction difference DCT coefficient is calculated.

  The corrected differential DCT coefficient output from the coefficient corrector 713 is subjected to inverse DCT processing by the inverse orthogonal transformer 705, and is added to the prediction block from the image buffer 714 in the adder 706 as a decoded prediction difference signal. It is output as a decoded image signal. The image buffer 714 is a buffer for absorbing the delay required for the coefficient estimation process in the orthogonal transform coefficient estimator 712, delays the prediction block output from the MC unit 709 for a fixed time, and outputs it to the adder 706. To do. The capacity of the image buffer 714 depends on the delay required by the orthogonal transform coefficient estimator 712, and the image buffer 714 can be deleted when performing an estimation process that does not cause a delay.

  The decoded image signals output from the adder 706 are stored in the output image memory 710, rearranged in order of output time, and output from the output terminal 711. Also, the decoded image signal output from the adder 706 is supplied to and stored in the reference image memory 708 via the two-dimensional block inverse converter 707 when the corresponding frame is an image to be referenced. The image stored in the reference image memory 708 is used at the time of the subsequent prediction frame encoding process.

  Regarding the detailed operations of the orthogonal transform coefficient estimator 712 and the coefficient corrector 713 in the video decoding apparatus of the first embodiment, the orthogonal operations in the video encoding apparatus of the first embodiment shown in FIG. By performing the same operation as that of the transform coefficient estimator 118 and the coefficient corrector 119, the restored image at the time of encoding and that at the time of decoding are made coincident with each other so that the mismatch of the reference image in the motion compensated prediction encoding process does not occur.

  In the first embodiment, each coefficient value generated by applying DCT to a decoded image is obtained, and the DCT coefficient that the decoded image of the target block has the coefficient component held at the time of input of the target DCT block In addition, the estimation is performed on the basis of the DCT coefficients of the decoded images of the surrounding blocks, and the estimation is performed by subtracting the prediction difference DCT component at the time of motion compensation prediction encoding from the DCT component of the prediction block from the estimated DCT component of the input image.

  In the first embodiment, a prediction error signal subjected to motion compensation prediction processing is estimated to be a degradation component caused by orthogonal transform processing and quantization processing, and a truncated component of a quantized prediction error orthogonal transform signal is estimated. By doing so, the deterioration component is corrected. Since the range of the difference component rounded down by quantization can be limited by the quantization scale and the quantization matrix, the validity of the estimated component can be determined for each coefficient. As a result, according to the present embodiment, the prediction error signal is caused by a mismatch between a component that corrects block distortion, a component that corrects mosquito distortion, and a motion-compensated prediction that is a degradation component caused by orthogonal transform processing and quantization processing. Thus, it is possible to appropriately select a component obtained by truncating the difference signal, a component lost for each of the difference signals, and restore the signal based on the quantization processing performed at the time of encoding.

  Since the restored image signal is also used as a reference image for motion compensation prediction, the time direction correlation between the reference image with improved quality and the input image can be calculated, the accuracy of the motion vector can be improved, and the degradation component can be reduced. Deterioration due to the prediction frame remaining in the decoded image can also be prevented.

  Furthermore, according to the present embodiment, when estimating the decoding orthogonal transform coefficient, it is expressed by a predetermined source model such as a quadratic function that defines the behavior of the signal in the block, and the signal in the block is expressed as a block boundary. By using a technique that generates an estimated signal based on conditions, it is possible to estimate the signal characteristics held by the entire block in a form in which distortion components between blocks are removed. A signal can be generated, and correction processing can function more appropriately.

(Second Embodiment)
Next, a second embodiment of the video encoding device and video decoding device of the present invention will be described. This second embodiment provides a moving picture decoding apparatus that improves the quality of a decoded picture that is decoded when standard coding processing is performed on the moving picture coding apparatus.

  FIG. 8 shows a block diagram of a second embodiment of the moving picture decoding apparatus according to the present invention. As shown in the figure, the moving picture decoding apparatus according to the present embodiment includes an input terminal 800, a stream buffer 801, an entropy decoder 802, a coding table 803, an inverse quantizer 804, an inverse orthogonal transformer 805, an addition 806, two-dimensional block inverse transformer 807, reference image memory 808, MC (motion compensation) unit 809, output image memory 810, output terminal 811, orthogonal transform coefficient estimator 812, coefficient correction difference generator 813, inverse orthogonal transform And a correction image difference memory 815 and an adder 816. The circuit unit from the input terminal 800 to the output terminal 811 has the same configuration as the circuit unit from the input terminal 1200 to the output terminal 1211 of the conventional video decoding device in FIG.

  Also, the orthogonal transform coefficient estimator 812 uses the same configuration as the orthogonal transform coefficient estimator 712 in the video decoding device according to the first embodiment. Compared with the moving picture decoding apparatus according to the first embodiment, this embodiment includes a coefficient correction difference generator 813, an inverse orthogonal transformer 814, and a corrected image difference memory 815, and an image corresponding to the image buffer 714. It is characterized in that the buffer is deleted.

  In FIG. 8, a bit stream obtained by performing standard encoding processing by the moving image encoding apparatus of FIG. 12 or the like is input to the input terminal 800, and further, a stream buffer 801, an entropy decoder 802, an encoding table 803, Decoded DCT coefficients are calculated by a circuit unit including the inverse quantizer 804. This operation is the same as that of a conventional video decoding device.

  The decoded DCT coefficient is input to the inverse orthogonal transformer 805, the orthogonal transform coefficient estimator 812, and the coefficient correction difference generator 813. The orthogonal transform coefficient estimator 812 lost the DCT coefficient due to quantization during encoding. The coefficient component is estimated, and the estimated DCT coefficient is output. The coefficient correction difference generator 813 selects, from the estimated DCT coefficients and the decoded DCT coefficients, a component in which the DCT coefficient that the encoded image signal has lost due to quantization, and the selected component and Difference information with the decoded DCT coefficient is calculated. The difference from the configuration of the moving picture decoding apparatus according to the first embodiment is that the local decoding process is performed on the conventional DCT coefficient without performing the process of directly adding the correction coefficient to the decoded DCT coefficient. is there.

  Here, an algorithm for selecting a coefficient in units of DCT blocks in the coefficient correction difference generator 813 will be described with reference to the flowchart of FIG. First, the coefficient correction difference generator 813 obtains a decoded DCT coefficient DecC and an estimated DCT coefficient EstC for the target DCT block (step S900).

  Subsequently, a quantization step Q and a quantization matrix M when the target DCT block is quantized are obtained (step S901). Next, the coefficient correction difference generator 813 calculates, for each coefficient term of the DCT block, a possible range before the decoded DCT coefficient is quantized. In MPEG2 coding, for DCT blocks using motion compensated prediction, DCT coefficients are rounded off when quantized, motion compensated prediction is not used, and the image signal itself is DCT processed. Processing is performed. Therefore, it is determined whether the target DCT block is a prediction difference block using motion compensated prediction (step S902).

  When the target DCT block is the prediction difference DCT, the coefficient correction difference generator 813 quantizes the horizontal x and vertical y coefficients in the decoded DCT coefficient into terms of DecC (x, y), horizontal x, and vertical y. Assuming that the matrix value is M (x, y), the range of absolute values of coefficients that can be taken before quantization is calculated as follows (step S903).

Minimum value EstCmin = abs (DecC (x, y))
Maximum value EstCmax = abs (DecC (x, y)) + Q * M (x, y)
On the other hand, when the target DCT block is not the prediction difference DCT, the coefficient correction difference generator 813 performs the following calculation (step S904).

Minimum value EstCmin = 0: When DecC (x, y) = 0
abs (DecC (x, y))-Q * M (x, y) / 2
: When DecC (x, y) ≠ 0, the maximum value EstCmax = abs (DecC (x, y)) + Q * M (x, y) / 2
Subsequently, the coefficient correction difference generator 813 determines that the maximum value and minimum value calculated in step S903 or S904 and the obtained estimated DCT coefficient are within the range of coefficients that can be taken before the following quantization. It is compared whether or not the condition to be satisfied is satisfied (step S905). If the condition is satisfied, the estimated DCT coefficient EstC (x, y) is selected as the corrected DCT coefficient CorC (x, y) (step S906). If the condition is not satisfied, the decoded DCT coefficient DecC (x, y) is selected as the corrected DCT coefficient CorC (x, y) (step S907). Here, the condition that satisfies the above-described range of possible coefficients is the same as the condition in step S605 in the flowchart of FIG. 6 described above.

Finally, the coefficient correction difference generator 813 sets the correction DCT coefficient difference information DifC (x, y) to 0 for the DC term of the DCT block, and the following expression DifC for all the remaining decoded DCT coefficients. (X, y) = CorC (x, y) −DecC (x, y)
Based on the above, the corrected DCT coefficient difference information DifC is calculated and output (step S908). The coefficient correction difference generator 813 obtains corrected DCT coefficient difference information DifC for all DCT blocks in the macro block according to the above algorithm.

  The corrected DCT coefficient difference information DifC output from the coefficient correction difference generator 813 differs from the processing performed by the conventional decoding device, and is subjected to inverse DCT processing by the inverse orthogonal transformer 814 in FIG. The corrected image difference signal for correcting the signal is stored in the corrected image difference memory 815. In the corrected image difference memory 815 and the output image memory 810, the corrected image difference signal and the output image are output so that the signals are rearranged in the order of the output time according to the encoding processing structure and the signals at the same time and the same position are output. Match timing.

  On the other hand, the inverse orthogonal transformer 805 receives the decoded DCT coefficient from the inverse quantizer 804 and performs inverse DCT processing to generate decoded data. In the case of a reference frame, the decoded data is stored in the output image memory 810 as a decoded image signal via the adder 806 and the two-dimensional block inverse transformer 807, and is also supplied to the reference image memory 808 for reference. Stored as an image.

  In the prediction frame, the prediction block is generated by inputting the motion vector and the mode signal output from the entropy decoder 802 to the MC unit 809 and cutting out the reference image from the reference image memory 808. The generated prediction block is supplied to the adder 806 and added with the decoded data from the inverse orthogonal transformer 805. The added decoded data is stored in the output image memory 810 as a decoded image signal via the two-dimensional block inverse transformer 807 as in the case of the base frame, and when the corresponding frame is a reference image, the reference image memory Also supplied to 808 and stored as a reference image. As described above, in the image stored internally as the reference image, the processing is configured so that the information generated by the conventional moving image encoding apparatus can be stored.

  The adder 816 generates a decoded image signal corrected by adding the decoded image signal output from the output image memory 810 and the corrected image difference signal output from the corrected image difference memory 815, and outputs from the output terminal 811. Output as a corrected decoded image signal.

  In the second embodiment, similarly to the first embodiment, the DCT coefficient held at the time of input by the decoded image signal of the target block is estimated, and the prediction difference at the time of motion compensation prediction encoding is estimated. The DCT component is estimated by subtracting from the estimated DCT component of the input image from the DCT component of the prediction block. The validity of the estimated component is the same for the part determined by the coefficient based on the quantization scale and the quantization matrix, but the estimated component is not reflected in the reference image used for motion compensation prediction of the subsequent moving image. As a result, it becomes possible to decode the bitstream generated by the existing video encoding device without accumulating errors.

In the present embodiment, a difference component with the decoded DCT component is calculated from the estimated DCT component, and an inverse orthogonal transform (IDCT) is performed on the difference component, whereby a correction signal for the decoded image signal is obtained. Generate. Then, by adding the generated correction signal when the decoded image signal is output, correction of the component in which the difference signal generated by the mismatch of the block distortion correction, the mosquito distortion correction, and the motion compensation prediction is rounded down by the quantization is performed. The signal can be reproduced. Therefore, according to the present embodiment, it is possible to construct a moving picture decoding apparatus capable of correcting and reproducing a deterioration component of information encoded and transmitted / recorded by an existing encoding process.
(Third embodiment)
Next, a third embodiment of the present invention will be described. In the third embodiment, the functions of the moving picture coding apparatus and the moving picture decoding apparatus according to the first and second embodiments are realized in a central processing control apparatus which is a computer shown in FIGS. A moving picture encoding program and a moving picture decoding program are described below.

  FIG. 10 shows a configuration diagram of an example of an information processing apparatus that operates according to the moving picture encoding program of the present invention. In the figure, an information processing device 1000 includes an input device 1001 for inputting various types of information, an output device 1002 for outputting various types of information, and a central operation that is operated by the moving image encoding program of the present embodiment. A processing control device 1003, an external storage device 1004, a temporary storage device 1005 used as a work area for arithmetic processing by the central processing control device 1003, and a communication device 1006 for communicating with the outside include a bidirectional bus. 1007 is connected.

  The central processing control apparatus 1003 receives the moving picture coding program of the present embodiment from a recording medium or via a communication network and is taken in by the communication apparatus 1006, and is stored in the two-dimensional block converter 102 shown in FIG. Corresponding to a two-dimensional block transform unit 1008 corresponding to the subtractor 1009 corresponding to the subtractor 103, an orthogonal transform unit 1010 corresponding to the orthogonal transformer 104, a quantizing unit 1011 corresponding to the quantizer 105, and an entropy encoder 106. Entropy encoding means 1012, inverse quantization means 1013 corresponding to inverse quantizer 110, inverse orthogonal transform means 1014 corresponding to inverse orthogonal transformer 111, addition means 1015 equivalent to adder 112, two-dimensional block inverse transformer 2 dimensional block inverse transform means 1016 corresponding to 113 and ME device 115 E means 1017, MC means 1018 corresponding to MC unit 116, coding control means 1019 corresponding to code amount controller 117, orthogonal transform coefficient estimation means 1020 equivalent to orthogonal transform coefficient estimator 118, and coefficient corrector 119 And at least the functions of the coefficient correction unit 1021 for performing the same operation as the moving image encoding apparatus shown in FIG. 1 by software processing.

  FIG. 11 shows a configuration diagram of an example of an information processing apparatus that operates according to an embodiment of a moving picture decoding program according to the present invention. In this figure, the basic configuration of the apparatus included in the information processing apparatus 1100 is the same as that of the information processing apparatus 1000 in FIG. The central processing control device 1103 has the entropy corresponding to the entropy decoder 702 in FIG. 7 as the moving picture decoding program according to the present embodiment is distributed from a recording medium or distributed via a communication network and taken in by the communication device 1106. Decoding means 1108, inverse quantization means 1109 corresponding to inverse quantizer 704, inverse orthogonal transform means 1110 corresponding to inverse orthogonal transformer 705, addition means 1112 corresponding to adder 706, and two-dimensional block inverse transformer 707 Two-dimensional block inverse transform unit 1112 corresponding to, MC unit 1113 corresponding to MC unit 709, and orthogonal transform coefficient estimation unit 1114 corresponding to orthogonal transform coefficient estimator 712, and corresponding to coefficient corrector 713 By having the function of the coefficient correction means 1115 to perform the decoding of the first embodiment shown in FIG. The apparatus similar operations performed by software processing.

  In addition, instead of the coefficient correction means 1115, the coefficient correction difference generation means 1116 corresponding to the coefficient correction difference generator 813 shown in FIG. 8 is provided, so that the moving picture decoding according to the second embodiment shown in FIG. The same operation as that of the computer is executed by software.

  10 and 11, the memory for storing images and coefficients is configured in the temporary storage devices 1002 and 1102, and the processing of the inverse orthogonal transformer 814 in FIG. 8 can be shared by the inverse orthogonal transform means 1014 and 1110 as software. As such, it does not have any additional means.

  In each embodiment of the present invention, the MPEG2 encoding method is used as a moving image encoding configuration. However, other methods using motion compensated prediction and orthogonal transform such as MPEG1, MPEG4 ASP, MPEG4 AVC, and VC1 are used. It is expected that the same effect can be exhibited even if the moving image coding standard is used, and it can be effectively utilized.

  The present invention can be applied to an encoding device and an encoding program that compress and transmit a moving image signal. Further, the present invention can be applied to a decoding device, a decoding program, and the like that restore information transmitted by compressing a moving image signal.

It is a block diagram of 1st Embodiment of the moving image encoder of this invention. FIG. 2 is a block diagram of an embodiment of an orthogonal transform coefficient estimator in FIG. 1. It is a conceptual diagram showing the basic concept of multiple harmonic local orthogonal transformation. It is a conceptual diagram for showing the definition regarding the boundary between blocks in a decoded image block. It is a conceptual diagram which shows the coefficient data processed in the orthogonal transformation coefficient estimator of the 1st Embodiment of this invention. It is a flowchart explaining operation | movement of the coefficient corrector in FIG. It is a block diagram of the moving picture decoding apparatus of the 1st Embodiment of this invention. It is a block diagram of the moving image decoding apparatus of the 2nd Embodiment of this invention. It is a flowchart explaining operation | movement of the coefficient correction | amendment difference generator in FIG. It is a functional block diagram showing the basic composition of the information processor which realizes the function with which the moving picture coding device of the present invention is provided by running the moving picture coding program of the present invention. It is a functional block diagram showing the basic composition of an information processor which realizes the function with which the video decoding device of the present invention is provided, by running the video decoding program of the present invention. It is a block diagram which shows an example of the conventional MPEG2 encoding apparatus. It is a block diagram which shows an example of the conventional MPEG2 decoding apparatus. It is a conceptual diagram which shows the operation | movement concept of the conventional block distortion removal filter.

Explanation of symbols

100 Digital Image Signal Input Terminal 101 Input Image Memory 102 Two-dimensional Block Transformer 103, 205 Subtractor 104, 200 Orthogonal Transformer 105 Quantizer 106 Entropy Encoder 107, 703, 803 Coding Table 108, 701, 801 Stream Buffer 109, 711, 811 Output terminal 110, 704, 804 Inverse quantizer 111, 705, 805 Inverse orthogonal transformer 112, 201, 706, 806, 816 Adder 113, 707, 807 Two-dimensional block inverse transformer 114, 708, 808 Reference image memory 115 ME (motion estimator) 116, 709, 809 MC (motion compensation) unit 117 Code amount controller 118 Orthogonal transform coefficient estimator 119 Coefficient corrector 120 Image buffer 202, 204 Coefficient buffer 203 Inclination cancellation Interchange coefficient generator 700, 800 Bit stream input terminal 702, 802 Entropy decoder 710 Output image memory 711 Output terminal 712, 812 Orthogonal transform coefficient estimator 713 Coefficient corrector 714 Image buffer 813 Coefficient correction difference generator 815 Corrected image difference Memory 1000, 1100 Information processing device 1001, 1101 Input device 1002, 1102 Output device 1003, 1103 Central processing control device 1004, 1104 External storage device 1005, 1105 Temporary storage device 1006, 1106 Communication device 1007, 1107 Bidirectional bus 1008 Two-dimensional Block transformation means 1009 Subtraction means 1010 Orthogonal transformation means 1011 Quantization means 1012 Entropy coding means 1013, 1109 Inverse quantization means 1014, 1110 Inverse orthogonality Switch means 1015,1111 addition means 1016,1112 2-dimensional block inverse transform unit 1017 ME unit 1018,1113 MC 1019 code amount control unit 1020,1114 orthogonal transform coefficient estimating means 1021,1115 coefficient correction unit 1116 coefficient correction difference generation means

Claims (10)

The input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and motion compensation prediction is performed using the rectangular region as a processing unit and a decoded image signal of a previously encoded image as a reference image. An orthogonal transform process using a rectangular area defined separately from the rectangular area as a processing unit for the difference signal between the encoding target image that is the moving image signal to be encoded and the motion compensated prediction signal; In a video encoding device that performs encoding by sequentially performing quantization processing,
The motion compensated prediction signal and a decoded orthogonal transform coefficient generated by dequantizing the quantized data generated as a result of the orthogonal transform process and the quantization process are input and quantized. An orthogonal transform coefficient estimating means for estimating an orthogonal transform coefficient lost as a result and outputting an estimated orthogonal transform coefficient;
Using the estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation means, the decoded orthogonal transform coefficient, and a quantized value for each coefficient indicating the fineness of quantization when the quantization process is performed Then, region information that can be taken by the orthogonal transform coefficient before quantization is calculated from the quantized value, and whether or not the decoded orthogonal transform coefficient is replaced with the estimated orthogonal transform coefficient is determined based on the region information. Coefficient correction means for generating a corrected orthogonal transform coefficient;
Correction image signal generation means for generating a corrected decoded image signal by adding a signal obtained by performing inverse orthogonal transform processing to the corrected orthogonal transform coefficient output from the coefficient correction means and the motion compensation prediction signal. ,
A moving image encoding apparatus, wherein the corrected decoded image signal is used as the reference image of a moving image signal to be subsequently encoded. The orthogonal transform coefficient estimation means includes
Orthogonal transform means for generating orthogonal transform coefficients for the motion compensated prediction signal;
Means for generating an orthogonal transform coefficient of a decoded image signal from the motion compensated prediction signal and the decoded orthogonal transform coefficient;
Based on the boundary condition of the corresponding boundary portion of the generated orthogonal transform coefficient of the decoded image signal and the corresponding rectangular region decoded image signal in the rectangular region unit to be subjected to orthogonal transformation processing, the Poisson equation An inclination-cancellation orthogonal transform coefficient generating means for generating an estimated orthogonal transform coefficient of the decoded image signal by generating an estimated decoded image signal in the rectangular region that satisfies
Subtracting means for generating an estimated prediction difference orthogonal transform coefficient by subtracting an orthogonal transform coefficient for the motion compensated prediction signal output from the orthogonal transform means from the estimated orthogonal transform coefficient. Item 4. A moving image encoding apparatus according to Item 1.   The inclination canceling orthogonal transform coefficient generating means inputs a rectangular area to be subjected to the orthogonal transform process and an orthogonal transform coefficient for a decoded image signal of the rectangular area existing around the rectangular area to be subjected to the orthogonal transform process. The moving picture coding apparatus according to claim 2, wherein the estimated orthogonal transform coefficient of the decoded image signal is generated by adding and multiplying the input orthogonal transform coefficients. The input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and motion compensation prediction is performed using the rectangular region as a processing unit and a decoded image signal of a previously encoded image as a reference image. An orthogonal transform process using a rectangular area defined separately from the rectangular area as a processing unit for the difference signal between the encoding target image that is the moving image signal to be encoded and the motion compensated prediction signal; In a moving picture decoding apparatus that sequentially decodes a bit stream of a moving picture signal that has been encoded by performing quantization processing,
Entropy decoding means for decoding orthogonal transform coefficients and motion vectors quantized from the bitstream;
Based on the motion vector from the entropy decoding unit, a motion compensation unit that cuts out a past decoded image signal as a reference signal and generates a motion compensated prediction signal;
The decoded orthogonal transform coefficient generated by dequantizing the quantized orthogonal transform coefficient output from the entropy decoding means and the motion compensated prediction signal are received as inputs, and the quantized result is lost. Orthogonal transform coefficient estimating means for estimating the orthogonal transform coefficient and outputting the estimated orthogonal transform coefficient;
The estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation unit, the decoded orthogonal transform coefficient, and a quantized value for each coefficient indicating the fineness of quantization when the quantization process is performed are received as inputs. Calculating the region information that can be taken by the orthogonal transform coefficient before quantization from the quantized value, and determining whether or not to replace the decoded orthogonal transform coefficient with the estimated orthogonal transform coefficient based on the region information. Coefficient correction means for generating
Corrected decoded image signal generating means for generating a corrected decoded image signal by adding the motion compensated prediction signal to a signal obtained by performing inverse orthogonal transform processing on the corrected orthogonal transform coefficient output from the coefficient correcting means; Have
A moving picture decoding apparatus, wherein the corrected decoded picture signal is used as the reference picture of a moving picture signal to be encoded subsequently. The input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and motion compensation prediction is performed using the rectangular region as a processing unit and a decoded image signal of a previously encoded image as a reference image. An orthogonal transform process using a rectangular area defined separately from the rectangular area as a processing unit for the difference signal between the encoding target image that is the moving image signal to be encoded and the motion compensated prediction signal; In a moving picture decoding apparatus that sequentially decodes a bit stream of a moving picture signal that has been encoded by performing quantization processing,
Entropy decoding means for decoding orthogonal transform coefficients and motion vectors quantized from the bitstream;
Based on the motion vector from the entropy decoding unit, a motion compensation unit that cuts out a past decoded image signal as a reference signal and generates a motion compensated prediction signal;
The decoded orthogonal transform coefficient generated by dequantizing the quantized orthogonal transform coefficient output from the entropy decoding means and the motion compensated prediction signal are received as inputs, and the quantized result is lost. Orthogonal transform coefficient estimating means for estimating the orthogonal transform coefficient and outputting the estimated orthogonal transform coefficient;
The estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation unit, the decoded orthogonal transform coefficient, and a quantized value for each coefficient indicating the fineness of quantization when the quantization process is performed are received as inputs. , Calculating region information that can be taken by the orthogonal transform coefficient before quantization from the quantized value, and determining whether to replace the decoded orthogonal transform coefficient with the estimated orthogonal transform coefficient based on the region information, Coefficient correction difference generation means for outputting a correction difference orthogonal transformation coefficient that is difference information from the orthogonal transformation coefficient;
An inverse orthogonal transform unit that converts the correction difference orthogonal transform coefficient output from the coefficient correction difference generation unit into a correction difference signal by performing an inverse orthogonal transform process;
After the decoded image signal decoded using the signal obtained by inverse orthogonal transformation of the decoded orthogonal transform coefficient and the motion compensated prediction signal is stored as the reference image, the decoded image signal output as an output image is stored. A video decoding apparatus comprising: an image signal decoding unit that adds the corrected difference signal only and outputs a decoded image signal corrected for a reproduced image. The orthogonal transform coefficient estimation means includes
Orthogonal transform means for generating orthogonal transform coefficients for the motion compensated prediction signal;
Means for generating an orthogonal transform coefficient of a decoded image signal from the motion compensated prediction signal and the decoded orthogonal transform coefficient;
Based on the boundary condition of the corresponding boundary portion of the generated orthogonal transform coefficient of the decoded image signal and the corresponding rectangular region decoded image signal in the rectangular region unit to be subjected to orthogonal transformation processing, the Poisson equation An inclination-cancellation orthogonal transform coefficient generating means for generating an estimated orthogonal transform coefficient of the decoded image signal by generating an estimated decoded image signal in the rectangular region that satisfies
Subtracting means for generating an estimated prediction difference orthogonal transform coefficient by subtracting an orthogonal transform coefficient for the motion compensated prediction signal output from the orthogonal transform means from the estimated orthogonal transform coefficient. Item 6. The moving picture decoding device according to Item 4 or 5.   The inclination canceling orthogonal transform coefficient generating means inputs a rectangular area to be subjected to the orthogonal transform process and an orthogonal transform coefficient for a decoded image signal of the rectangular area existing around the rectangular area to be subjected to the orthogonal transform process. 7. The moving picture decoding apparatus according to claim 6, wherein the corrected orthogonal transform coefficient is generated by adding and multiplying the input orthogonal transform coefficients. The input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and motion compensation prediction is performed using the rectangular region as a processing unit and a decoded image signal of a previously encoded image as a reference image. An orthogonal transform process using a rectangular area defined separately from the rectangular area as a processing unit for the difference signal between the encoding target image that is the moving image signal to be encoded and the motion compensated prediction signal; In a moving image encoding program that causes a computer to execute encoding by sequentially performing quantization processing,
The computer,
The motion compensated prediction signal and a decoded orthogonal transform coefficient generated by dequantizing the quantized data generated as a result of the orthogonal transform process and the quantization process are input and quantized. An orthogonal transform coefficient estimating means for estimating an orthogonal transform coefficient lost as a result and outputting an estimated orthogonal transform coefficient;
Using the estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation means, the decoded orthogonal transform coefficient, and a quantized value for each coefficient indicating the fineness of quantization when the quantization process is performed Then, region information that can be taken by the orthogonal transform coefficient before quantization is calculated from the quantized value, and whether or not the decoded orthogonal transform coefficient is replaced with the estimated orthogonal transform coefficient is determined based on the region information. Coefficient correction means for generating a corrected orthogonal transform coefficient;
A corrected image signal generating unit that generates a corrected decoded image signal by adding a signal obtained by performing an inverse orthogonal transform process to the corrected orthogonal transform coefficient output from the coefficient correcting unit, and the motion compensated prediction signal;
And the corrected decoded image signal is used as the reference image of the moving image signal to be encoded subsequently. The input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and motion compensation prediction is performed using the rectangular region as a processing unit and a decoded image signal of a previously encoded image as a reference image. An orthogonal transform process using a rectangular area defined separately from the rectangular area as a processing unit for the difference signal between the encoding target image that is the moving image signal to be encoded and the motion compensated prediction signal; In a moving picture decoding program for causing a computer to execute an operation of decoding a bit stream of a moving picture signal encoded by sequentially performing quantization processing,
The computer,
Entropy decoding means for decoding orthogonal transform coefficients and motion vectors quantized from the bitstream;
Based on the motion vector from the entropy decoding unit, a motion compensation unit that cuts out a past decoded image signal as a reference signal and generates a motion compensated prediction signal;
The decoded orthogonal transform coefficient generated by dequantizing the quantized orthogonal transform coefficient output from the entropy decoding means and the motion compensated prediction signal are received as inputs, and the quantized result is lost. Orthogonal transform coefficient estimating means for estimating the orthogonal transform coefficient and outputting the estimated orthogonal transform coefficient;
The estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation unit, the decoded orthogonal transform coefficient, and a quantized value for each coefficient indicating the fineness of quantization when the quantization process is performed are received as inputs. Calculating the region information that can be taken by the orthogonal transform coefficient before quantization from the quantized value, and determining whether or not to replace the decoded orthogonal transform coefficient with the estimated orthogonal transform coefficient based on the region information. Coefficient correction means for generating
A corrected decoded image signal generating unit that generates a corrected decoded image signal by adding the motion compensation prediction signal to a signal obtained by performing an inverse orthogonal transform process on the corrected orthogonal transform coefficient output from the coefficient correction unit. And the corrected decoded image signal is used as the reference image of the moving image signal to be encoded subsequently. The input moving image signal to be encoded is divided into rectangular regions each having a predetermined number of pixels, and motion compensation prediction is performed using the rectangular region as a processing unit and a decoded image signal of a previously encoded image as a reference image. An orthogonal transform process using a rectangular area defined separately from the rectangular area as a processing unit for the difference signal between the encoding target image that is the moving image signal to be encoded and the motion compensated prediction signal; In a moving picture decoding program for causing a computer to execute an operation of decoding a bit stream of a moving picture signal encoded by sequentially performing quantization processing,
The computer,
Entropy decoding means for decoding orthogonal transform coefficients and motion vectors quantized from the bitstream;
Based on the motion vector from the entropy decoding unit, a motion compensation unit that cuts out a past decoded image signal as a reference signal and generates a motion compensated prediction signal;
The decoded orthogonal transform coefficient generated by dequantizing the quantized orthogonal transform coefficient output from the entropy decoding means and the motion compensated prediction signal are received as inputs, and the quantized result is lost. Orthogonal transform coefficient estimating means for estimating the orthogonal transform coefficient and outputting the estimated orthogonal transform coefficient;
The estimated orthogonal transform coefficient output from the orthogonal transform coefficient estimation unit, the decoded orthogonal transform coefficient, and a quantized value for each coefficient indicating the fineness of quantization when the quantization process is performed are received as inputs. , Calculating region information that can be taken by the orthogonal transform coefficient before quantization from the quantized value, and determining whether to replace the decoded orthogonal transform coefficient with the estimated orthogonal transform coefficient based on the region information, Coefficient correction difference generation means for outputting a correction difference orthogonal transformation coefficient that is difference information from the orthogonal transformation coefficient;
An inverse orthogonal transform unit that converts the correction difference orthogonal transform coefficient output from the coefficient correction difference generation unit into a correction difference signal by performing an inverse orthogonal transform process;
After the decoded image signal decoded using the signal obtained by inverse orthogonal transformation of the decoded orthogonal transform coefficient and the motion compensated prediction signal is stored as the reference image, the decoded image signal output as an output image is stored. A moving picture decoding program that functions as an image signal decoding unit that adds the corrected difference signal only and outputs a decoded image signal corrected for a reproduced image.

Images