JP6641149B2 - Video encoding device, video decoding device, and program - Google Patents

Description

  The present invention relates to a video encoding device that compresses and encodes a video signal, a video decoding device that decodes a video signal, and a program.

  Conventionally, as a signal encoding method, a method of predicting a signal waveform and encoding a prediction residual has been known. In particular, when encoding a video signal, when encoding a certain frame, the video is divided into blocks, and for each of the encoding target blocks, another encoded frame or an image in the frame is used. Perform prediction processing. Then, transform processing such as discrete cosine transform is performed on the error between the result of the prediction processing and the block to be coded, the transform coefficients that are the result of the transform processing are quantized, and the quantization result is entropy-coded. Thus, the data of the video signal is compression-encoded.

[Conventional video encoding device]
First, a conventional video encoding device will be described. FIG. 6 is a block diagram illustrating a configuration example of a conventional video encoding device. The video encoding apparatus 100 includes a block dividing unit 10, a difference unit 11, an orthogonal transformation unit 13, a quantization unit 14, an entropy encoding unit 15, an inverse quantization unit 16, an inverse orthogonal transformation unit 17, an addition unit 19, a loop The apparatus includes a filter 20, a memory 21, a prediction unit 22, and an optimization unit 23.

  The video encoding device 100 receives a video signal, performs an existing encoding process on the video signal, generates an encoded bit sequence, and outputs the encoded bit sequence to a video decoding device 200 described later.

  The block dividing unit 10 receives a video signal and divides the frame into blocks for each frame of the video signal. The block division unit 10 outputs the divided blocks to the difference unit 11, the prediction unit 22, and the optimization unit 23.

  The block division may be performed by a square of a fixed size, or a rectangle of various sizes as defined in MPEG-4 AVC / H.264, MPEG-H HEVC / H.265, etc. Sometimes it is done in.

  Note that the block shape in the processing of the later-described difference unit 11, orthogonal transformation unit 13, inverse orthogonal transformation unit 17, and addition unit 19 may be different from the block shape in the processing of the prediction unit 22 and the optimization unit 23 described later. obtain. In this case, the block dividing means 10 assigns each divided shape such that the block shape for the difference means 11 and the like and the block shape for the prediction means 22 and the like are different. Then, the block dividing unit 10 divides the frame of the video signal into blocks having respective block shapes, outputs one block to the difference unit 11, and outputs the other block to the prediction unit 22 and the optimization unit 23. .

  The difference unit 11 receives the block from the block division unit 10 and the prediction image from the prediction unit 22, and subtracts the pixel value sequence of the prediction image from the pixel value sequence of the block for the position of the block. Find the resulting residual value sequence. The difference means 11 outputs the residual value sequence of the block to the orthogonal transformation means 13.

  The orthogonal transformation means 13 receives the residual value sequence of the block from the difference means 11 and performs an orthogonal transformation on the residual value sequence using a plurality of predetermined basis functions to generate a transformation coefficient sequence. The orthogonal transform unit 13 outputs a transform coefficient sequence of the block to the quantization unit 14.

  For the basis function, for example, a discrete cosine transform or an integer approximation thereof is used. In MPEG-H HEVC / H.265, discrete sine transform is also used under predetermined conditions. The conversion coefficient sequence is obtained by arranging conversion coefficients obtained by applying different basis functions to each of the horizontal and vertical two-dimensional spatial frequencies.

  The quantization means 14 receives the transform coefficient sequence of the block from the orthogonal transform means 13 and quantizes (discretely converts) the transform coefficient sequence based on a predetermined quantization table to generate a quantization index sequence. The quantization means 14 outputs the quantization index sequence of the block to the entropy coding means 15 and the inverse quantization means 16.

  The width of quantization is determined by a quantization step and a quantization parameter (QP: Quantization Parameter). Further, the quantization width is generally set to be different depending on the spatial frequency, and is set wider (rougher) as the spatial frequency is higher, and narrower (finer) as the spatial frequency is lower. As the value of the quantization parameter increases, the transform coefficient sequence is coarsely quantized and the degree of information compression increases, but the image quality deteriorates greatly.

  The entropy coding unit 15 inputs the block's quantization index sequence from the quantization unit 14 and performs entropy coding on the quantization index sequence based on a model that takes into account the statistical properties of the transform coefficient sequence to encode a code. Generate an allocation and coded bit string. The coded bit string generated by the entropy coding unit 15 is output to a video decoding device 200 described later.

  For the entropy coding, for example, Huffman coding, arithmetic coding, run-length coding, CAVLC (Context-adaptive Variable Length Coding), CABAC (Context-adaptive Binary Arithmetic Coding) and the like are used.

  Here, in addition to the quantization index sequence, the entropy encoding unit 15 adds motion vector information obtained by a prediction unit 22 described later, optimal mode information obtained by an optimization unit 23 described later, Information and other information added by a user or the like are also entropy-encoded and output as encoded bit strings.

  The inverse quantization means 16 receives the quantization index sequence of the block from the quantization means 14 and performs inverse processing of the quantization means 14 to inversely quantize the quantization index sequence and transform the quantization index sequence. Return to coefficient sequence. The inverse quantization means 16 outputs the transform coefficient sequence of the block to the inverse orthogonal transform means 17.

  The quantization means 14 has a predetermined range (width) for the transform coefficient value giving a certain quantization index value, but the inverse quantization means 16 determines the transform coefficient value for the input quantization index value as the range. Is output (for example, a median value).

  The inverse orthogonal transform unit 17 receives the transform coefficient sequence of the block from the inverse quantization unit 16, performs inverse processing of the orthogonal transform unit 13, performs inverse orthogonal transform on the transform coefficient sequence, and obtains a decoded residual value sequence. Generate The inverse orthogonal transform unit 17 outputs the decoded residual value sequence of the block to the adding unit 19. For example, the inverse orthogonal transform unit 17 performs the inverse discrete cosine transform when the orthogonal transform unit 13 performs the discrete cosine transform.

  The adding means 19 inputs the decoded residual value sequence of the block from the inverse orthogonal transform means 17 and a predicted image from the predicting means 22 described later. Then, the adding means 19 adds the pixel value sequence of the predicted image to the decoded residual value sequence for the position of the block, and obtains a decoded pixel value sequence as a result of the addition. The adding means 19 outputs the decoded pixel value sequence to the loop filter 20.

  The loop filter 20 receives the decoded pixel value sequence from the adding unit 19, performs a filtering process on the decoded pixel value sequence, stores the decoded pixel value sequence after the filtering process in the memory 21, and Output. This makes it possible to suppress distortion or the like at the block boundary. As the loop filter 20, for example, a deblocking filter, SAO (Sample Adaptive Offset), ALF (Adaptive Loop Filter), or the like is used.

  The decoded pixel value sequence after the filter processing by the loop filter 20 is stored in the memory 21. Here, the loop filter 20 stores the decoded pixel value sequence in the memory 21 in consideration of the position of the block. Thereby, decoded images are sequentially formed. The term “sequential” means that when sequentially processing the regions of the blocks divided by the block dividing means 10, each time the loop filter 20 completes the processing, the decoded pixel value sequence of the regions of the blocks is embedded, and the decoded image is formed. It is intended to be a process.

  The prediction unit 22 reads out the pixel value sequence of the decoded image formed so far (may be a frame decoded at another time so far) in the memory 21 and, if necessary, the block dividing unit 10. To input a block of a frame of a video signal. Then, the prediction unit 22 predictively synthesizes a pixel value sequence in a block to be processed, and generates a predicted image. The prediction unit 22 outputs the predicted image to the difference unit 11 and the addition unit 19.

  The predicting unit 22 includes, for example, a pixel value sequence of a decoded image at another time that has already been formed in the memory 21 and a current frame formed by a block input from the block dividing unit 10 in an area of a block to be processed. Are compared with each other, and a spatial relative positional relationship where the pixel value sequence is most matched (the error is reduced) is obtained as a motion vector (a motion vector is obtained by motion search). Then, the prediction unit 22 generates a predicted image by spatially moving (motion-compensating) the pixel value sequence of the decoded image at another time based on the motion vector.

  For example, the prediction unit 22 spatially adjoins the pixel value sequence spatially adjacent to the area of the block to be processed and based on the pixel value sequence already formed in the memory 21 in a fixed direction. A predicted image is generated by extrapolating or filling with a result obtained by performing an operation between the pixel value strings.

  The optimizing unit 23 includes a block dividing unit 10, an orthogonal transform unit 13 (corresponding inverse orthogonal transform unit 17), a quantizing unit 14 (corresponding inverse quantizing unit 16), an entropy coding unit 15, a loop The experiment is performed by changing various parameters used in the filter 20 and the prediction unit 22. Then, the optimization unit 23 determines a parameter (the value representing the rate distortion characteristic is a predetermined value from the maximum value) that maximizes or approximates the maximum in the rate distortion characteristic (to be better with a smaller code amount and a smaller error). Parameters that exist between the lower values).

  The optimizing unit 23 uses the specified parameters as the parameters of the optimal mode, and uses the parameters of the optimal mode as the block dividing unit 10, the orthogonal transform unit 13 (the inverse orthogonal transform unit 17 corresponding thereto), and the quantization unit 14 (corresponding to this). To the inverse quantization means 16), the entropy coding means 15, the loop filter 20 and the prediction means 22, and operate according to the parameters.

  Thereby, the block dividing means 10, the orthogonal transform means 13 (the corresponding inverse orthogonal transform means 17), the quantizing means 14 (the corresponding inverse quantizing means 16), the entropy encoding means 15, the loop filter 20, The prediction unit 22 performs a process according to the parameters input from the optimization unit 23.

  Further, the optimization unit 23 outputs the parameters of the optimal mode to the entropy encoding unit 15 as optimal mode information.

  As a result, the optimal mode information is entropy-encoded by the entropy encoding means 15 and output to the video decoding device 200 described later as an encoded bit sequence.

  As described above, by the existing encoding processing of the video encoding device 100, the quantization index sequence, the optimal mode information, and the like are converted into encoded bit sequences, and output to the video decoding device 200 described later.

[Conventional video decoding device]
Next, a conventional video decoding device will be described. FIG. 7 is a block diagram illustrating a configuration example of a conventional video decoding device. The video decoding device 200 includes an entropy decoding unit 31, an inverse quantization unit 32, an inverse orthogonal transformation unit 33, an addition unit 35, a loop filter 36, a memory 37, and a prediction unit 38.

  The video decoding device 200 receives a quantization index sequence and an encoded bit sequence such as optimal mode information from the video encoding device 100, performs an existing decoding process according to the optimal mode parameter indicated by the optimal mode information, and performs quantization index sequence. From the original video signal. Then, the video decoding device 200 outputs the decoded video signal.

  The entropy decoding unit 31 receives the coded bit sequence from the video coding device 100, performs the reverse process of the entropy coding unit 15 shown in FIG. 6, and performs entropy decoding on the coded bit sequence to obtain the original information. Reversibly decode. The original information includes, in addition to the quantization index sequence, the motion vector information obtained by the prediction unit 22 shown in FIG. 1, the optimal mode information obtained by the optimization unit 23, the information such as the size of the video, and the like. , And information added by the user or the like.

  The entropy decoding unit 31 outputs the parameters of the optimal mode, which are the entropy-decoded optimal mode information, to the below-described inverse quantization unit 32, inverse orthogonal transform unit 33, loop filter 36, and prediction unit 38, and operates according to the parameters. .

  Accordingly, the inverse quantization means 32, the inverse orthogonal transform means 33, the loop filter 36, and the prediction means 38, which will be described later, perform processing according to the parameters of the optimal mode input from the entropy decoding means 31.

  Further, the entropy decoding unit 31 outputs the quantization index sequence of the entropy-decoded block to the inverse quantization unit 32.

  The inverse quantization means 32 receives the block quantization index sequence from the entropy decoding means 31 and performs the same processing as the inverse quantization means 16 shown in FIG. 6 to convert the quantization index sequence into a transform coefficient sequence. Convert. The inverse quantization means 32 outputs the transform coefficient sequence of the block to the inverse orthogonal transform means 33.

  The inverse orthogonal transform unit 33 receives the transform coefficient sequence of the block from the inverse quantization unit 32 and performs the same processing as the inverse orthogonal transform unit 17 shown in FIG. 6 to perform the inverse orthogonal transform on the transform coefficient sequence. , And generates a decoded residual value sequence. The inverse orthogonal transform unit 33 outputs the decoded residual value sequence of the block to the adding unit 35.

  The adding unit 35 receives the decoded residual value sequence of the block from the inverse orthogonal transforming unit 33 and also receives a predicted image from a predicting unit 38 described later. Then, the adding unit 35 performs the same processing as the adding unit 19 shown in FIG. 6 to add the pixel value sequence of the predicted image to the decoded residual value sequence for the position of the block, and obtains the result of the addition. A certain decoded pixel value sequence is obtained. The adder 35 outputs the decoded pixel value sequence to the loop filter 36.

  The loop filter 36 receives the decoded pixel value sequence from the adding unit 35 and performs the same processing as the loop filter 20 shown in FIG. 6 to perform a filtering process on the decoded pixel value sequence. The decoded pixel value sequence is stored in the memory 37.

  In the memory 37, similarly to the memory 21 illustrated in FIG. 6, the decoded pixel value sequence that has been subjected to the filtering process by the loop filter 36 is sequentially stored at the corresponding block position. Thus, decoded images are sequentially formed in the memory 37.

  After the decoded image is completely formed in the memory 37, the video decoding device 200 reads the decoded image from the memory 37 and outputs the decoded image to the outside as a decoded video signal.

  The prediction unit 38 reads out the pixel value sequence of the decoded image formed so far (may be a frame decoded at another time so far) in the memory 37, and reads the pixel value sequence from the prediction unit 22 shown in FIG. Similarly, a pixel value sequence in a block to be processed is synthesized predictively to generate a predicted image. The prediction unit 38 outputs the predicted image to the addition unit 35.

  For example, when using the motion vector information, the prediction unit 38 inputs the motion vector information as a parameter from the entropy decoding unit 31 and spatially converts the pixel value sequence of the decoded image at another time based on the motion vector information. (A motion compensation) to generate a predicted image. In this case, no motion search is performed.

  As described above, by the existing decoding processing of the video decoding device 200, the coded bit sequence is converted into the quantization index sequence and the optimal mode information and the like, and the frame of the original video signal is restored.

  The coding scheme described in FIGS. 6 and 7 is called hybrid coding or MC-DCT (Motion Compensation / Discrete Cosine Transform) coding. For example, MPEG-1, MPEG-2, MPEG-4, MPEG-4 AVC / H. H.264, and MPEG-H HEVC / H. Each of the H.265 video coding schemes is a scheme belonging to MC-DCT coding.

  Also, a technique has been disclosed in which pre-processing and post-processing are added to existing video coding methods including MC-DCT coding to improve coding efficiency and subjective image quality (for example, see Patent Document 1). ). In the method of Patent Literature 1, the compression ratio in the existing video encoding method is reduced by using the resolution reduction processing in the pre-processing, while the resolution is reduced by applying the super-resolution technique to the post-processing. The restoration has been realized. At this time, it is possible to adjust the resolution restoring process, obtain an optimal adjustment value that can reproduce the input video best on the transmission side in advance, and transmit the optimal adjustment value to the decoding side as auxiliary information, thereby achieving high efficiency. Enables compressed transmission

Japanese Patent No. 5419795

  However, the method of Patent Document 1 is a form in which pre-processing and post-processing are added outside the existing encoding processing. For this reason, optimization performed in existing encoding processing (for example, Rate-Distortion Optimization (RDO)) and optimization performed in pre-processing and post-processing (performed in resolution restoration processing and the like) Optimization is a different process and is independent of each other. Even if optimization is performed in existing encoding processing and optimization is performed in pre-processing and post-processing, optimization is not necessarily performed as a whole. Therefore, the technique of Patent Document 1 is not always suitable for optimization from the overall viewpoint of the existing encoding processing, pre-processing, and post-processing.

  For example, pre-processing and post-processing are performed independently of block division in existing encoding processing. For this reason, the image shape or the sampling structure after the preprocessing needs to be in a form that can be input to the existing encoding processing, and the conversion processing applicable as the preprocessing is limited. In addition, in existing encoding processing, resolution conversion of a different magnification is applied to each location in a block, the sampling grid shape is enlarged or reduced in directions other than the vertical and horizontal directions, and a resampling is performed by a non-linear sampling structure. Is generally not allowed.

  As described above, by adding the pre-processing and the post-processing to the existing encoding processing, the encoding efficiency can be increased by the technique of Patent Document 1, but as described above, the existing encoding processing and the pre-processing are performed. In addition, there is a problem that the optimization is insufficient as a whole because it is not possible to sufficiently cooperate with the post-processing. For this reason, further improvement has been desired in order to realize optimization from a global perspective.

  Then, this invention is made in order to solve the said subject, The objective is to provide the video encoding device which can improve the encoding efficiency of a video signal, a video decoding device, and a program.

In order to solve the problem, the video encoding device according to claim 1 divides a frame of a video signal into blocks, and calculates a pixel value of each sample point of the block, or a pixel value of each of the sample values and a pixel value of a prediction image. In a video encoding device that performs orthogonal transformation and quantization on a residual value that is a difference between the two, and outputs an encoded bit sequence that is entropy-encoded, the number, arrangement, and Changing at least one of the pixel value or the residual value to resample the sample points, a parameter used in the resampler , and the block. Process, changing a plurality of parameters including respective parameters used in the orthogonal transformation process and the quantization process, respectively, to obtain evaluation values representing coding efficiency, Optimizing means for specifying the plurality of parameters at a fixed range of values as optimal plurality of parameters, and quantization performed by the orthogonal transformation and the quantization, which are resampled by the resampling means. An index sequence, and entropy-encoding means for entropy-encoding the plurality of parameters specified by the optimizing means to generate the encoded bit string, wherein the re-sampling means is configured by the optimizing means The sampling point is resampled according to a resampling rule corresponding to the specified parameter.

According to such a configuration, after the frame of the video signal is divided into blocks, resampling is performed in units of the blocks. As a result, the resampling process and the orthogonal transform and quantization processes performed on a block-by-block basis can be linked, so that pre-processing and post-processing irrespective of the block unit are performed as compared with the related art. As a whole, effective optimization can be realized. In addition, since resampling corresponding to the optimal parameter is performed in units of blocks, optimization can be reliably realized. As a result, it is possible to improve coding efficiency.

In the video encoding apparatus according to a second aspect of the present invention, in the video encoding apparatus according to the first aspect , the sample points input by the resampling unit are set as input sample points, and resampled by the resampling unit. When the sampled points are output sample points, the re-sampling rule, in units of the block, to reduce the number of the input sample points, or to change the density of the input sample points, Set the arrangement of the output sample points, based on the pixel value or the residual value of a predetermined number of the input sample points present in a predetermined range based on the output sample points, the pixel of the output sample points Value or a rule for calculating the residual value.

  According to such a configuration, for example, for a video signal in which pixel values or residual values have similar values in a predetermined wide range, when a wide range of blocks is divided corresponding to an optimal parameter, , The number of sample points may be reduced corresponding to the optimal parameters. As a result, the number of sample points is reduced, and the support range of the equivalent orthogonal transformation process can be expanded while suppressing the computational load of the subsequent orthogonal transformation process. Further, the resampling means can change the density of the sampling grid in accordance with the texture in the block, the degree or complexity of the residual, etc., in accordance with the optimal parameters. Thereby, many sample points can be arranged at necessary places in the block, and the approximation accuracy of the video signal can be improved. As a result, it is possible to improve coding efficiency.

Furthermore, the image decoding apparatus according to claim 3, enter the encoded bit string output by the video encoding apparatus according to claim 1, wherein the encoded bit stream to the entropy decoding the original and the inverse quantization and inverse orthogonal transform In a video decoding device for restoring a frame of a video signal, entropy decoding of the coded bit sequence, entropy decoding means for generating a quantization index sequence and a plurality of parameters, the block of the frame of the video signal as a unit In a process reverse to the process in which the video encoding device resamples the sample points, the number, arrangement, and the number of the sample points according to a reverse resampling rule corresponding to the parameter generated by the entropy decoding means. And the quantization index sequence is at least one of an inverse orthogonal transform and a decoded pixel value or a decoded residual value subjected to the inverse orthogonal transform. By changing the sampling point, reverse resampling means for reverse resampling the sample points, based on the decoded pixel value or the decoded residual value inversely resampled by the reverse resampling means. And restoring the frame of the original video signal.

According to such a configuration, inverse resampling is performed for each block. This makes it possible to link the inverse quantization and inverse orthogonal transform processing performed in units of blocks with the inverse sampling processing, so that the prior art in which preprocessing and postprocessing independent of the block units are performed. As a whole, effective optimization can be realized. In addition , since inverse resampling corresponding to the optimal parameters is performed in units of blocks, optimization can be reliably achieved. As a result, it is possible to improve coding efficiency.

According to a fourth aspect of the present invention, in the video decoding apparatus according to the third aspect , the sample points input by the inverse resampling unit are set as input sample points, and the inverse resampling unit performs inverse resampling. When the sampled points are set as output sample points, the inverse re-sampling rule increases the number of the sample points in units of the block or changes the density of the input sample points. Setting the arrangement of the output sample points, based on the decoded pixel values or the decoded residual values of a predetermined number of the input sample points present in a predetermined range based on the output sample points, Is a rule for calculating the decoded pixel value or the decoded residual value.

  According to such a configuration, for example, for a video signal whose pixel value or residual value is a similar value in a predetermined wide range, a wide range of blocks is divided according to the optimal parameter in the video encoding device, When the number of sample points corresponding to the optimal parameters is reduced by the resampling unit of the encoding device, the number of sample points corresponding to the optimal parameters is reduced by the inverse resampling unit of the video decoding device. Can be increased back to the original sample point. Further, when the sampling grid is changed in density according to the texture in the block, the degree or complexity of the residual, etc., in accordance with the optimal parameter by the resampling means of the video encoding device, By the inverse resampling means of the video decoding device, it is possible to return to the original sample points corresponding to the optimal parameters. As a result, it is possible to improve coding efficiency.

  As described above, according to the present invention, it is possible to improve the coding efficiency of a video signal.

1 is a block diagram illustrating a configuration example of a video encoding device according to an embodiment of the present invention. FIG. 9 is a diagram illustrating an example of a resampling process. FIG. 9 is a diagram illustrating Formula (1) for performing bilinear interpolation based on residual values (pixel values) of four input sample points. FIG. 14 is a diagram illustrating an example of a reverse resampling process. It is a block diagram showing an example of composition of a picture decoding device by an embodiment of the present invention. FIG. 14 is a block diagram illustrating a configuration example of a conventional video encoding device. FIG. 29 is a block diagram illustrating a configuration example of a conventional video decoding device.

Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings.
[Video encoding device]
First, a video encoding device according to an embodiment of the present invention will be described. FIG. 1 is a block diagram illustrating a configuration example of a video encoding device according to an embodiment of the present invention. The video encoding apparatus 1 includes a block dividing unit 10, a difference unit 11, a resampling unit 12, an orthogonal transform unit 13, a quantization unit 14, an entropy encoding unit 15, an inverse quantization unit 16, and an inverse orthogonal transform unit 17. , An inverse resampling unit 18, an adding unit 19, a loop filter 20, a memory 21, a predicting unit 22, and an optimizing unit 24.

  Comparing the conventional video encoding device 100 shown in FIG. 6 with the video encoding device 1 according to the embodiment of the present invention shown in FIG. Means 11, orthogonal transformation means 13, quantization means 14, entropy coding means 15, inverse quantization means 16, inverse orthogonal transformation means 17, addition means 19, loop filter 20, memory 21, and prediction means 22 Are the same.

  On the other hand, the video encoding device 1 further includes a resampling unit 12 and an inverse resampling unit 18 in addition to the configuration of the video encoding device 100, and is different from the optimizing unit 23 provided in the video encoding device 100. It differs from the video encoding device 100 in that it has an optimizing unit 24. In FIG. 1, portions common to FIG. 6 are denoted by the same reference numerals as in FIG. 6, and detailed description thereof will be omitted.

  The resampling unit 12 is provided between the difference unit 11 and the orthogonal transformation unit 13, receives a sequence of residual values of a block from the difference unit 11, and calculates the number, arrangement, and residual value of the sample points for each block. Resample the sampling points by changing. Then, the resampling unit 12 outputs the residual value sequence of the block after the resampling to the orthogonal transformation unit 13.

  The resampling unit 12 may resample by changing only the number of sample points, may resample by changing only the arrangement of the sample points, Re-sampling may be performed by changing only the value. In short, the resampling unit 12 may resample the sample points by changing one, two, or all of the number, arrangement, and residual values of the sample points.

  FIG. 2 is a diagram illustrating an example of a resampling process by the resampling unit 12. In FIG. 2, a circle indicates a sample point (sample point before resampling (input sample point)) of a residual value sequence of a block input to the resampling unit 12, and a cross indicates resampling. 2 shows sample points (sample points after re-sampling (output sample points)) of the residual value sequence of the block output by the conversion means 12. The resampling rule (resampling rule) for converting the residual value sequence of the input sample points into the residual value sequence of the output sample points is based on the number and arrangement of the input sample points and the output sample points shown in FIG. Is defined, and is set in advance as a conversion rule using a mathematical expression for calculating a residual value of an output sample point.

  The resampling unit 12 receives the parameters of the optimal resampling mode from the optimizing unit 24, which will be described later, and based on the parameters, sets a plurality of preset resampling rules (for example, FIGS. One resampling rule is determined from rule d)). Then, the resampling unit 12 resamples the sampling points according to the determined resampling rule.

  FIG. 2A shows an example in which the sampling grid is not changed, and is equivalent to a case in which resampling is not performed. The resampling means 12 resamples the input sample points shown in FIG. 2A according to the preset resampling rule shown in FIG. Find the difference sequence. In this case, the resampling unit 12 outputs the input residual value sequence without performing resampling. As shown in FIG. 2A, this resampling rule is a rule that does not change the number and arrangement of input sample points and output sample points, and does not change the residual value.

  FIG. 2B shows an example in which the number of samples is reduced (１ ／ in the horizontal direction and 倍 in the vertical direction), and the arrangement after re-sampling is also uniform. The resampling means 12 resamples the input sampling points shown in FIG. 2B according to the preset resampling rule shown in FIG. Find the difference sequence. As shown in FIG. 2B, the resampling rule reduces the number of input sample points by a factor of 1/2 in the horizontal direction and by a factor of 1/2 in the vertical direction. This is a rule that sets the arrangement of the output sample points so as to have a uniform interval different from the (interval), and calculates the residual value of the output sample points using a predetermined mathematical formula.

  FIG. 2C shows an example in which the positions of resampling are rearranged at non-uniform intervals (to change the density of the sample points) while maintaining the number of samples. The resampling unit 12 resamples the input sampling points shown in FIG. 2C according to the preset resampling rule shown in FIG. Find the difference sequence. As shown in FIG. 2C, the resampling rule sets the number of samples of the input sample points and the output sample points to be the same, and sets the output sample points to be the same or different from the input sample points and to have an uneven interval. This is a rule that sets the arrangement of the sample points and calculates the residual value of the output sample points using a predetermined mathematical formula.

  FIG. 2 (d) shows that the number of samples is reduced (1/4 times in the horizontal direction and 1/4 times in the vertical direction), and the resampling positions are set at uneven intervals (coarse and Is changed). The resampling means 12 resamples the input sampling points shown in FIG. 2D according to the preset resampling rule shown in FIG. Find the difference sequence. This resampling rule reduces the number of samples at the input sample points to 1/4 times in the horizontal direction and 1/4 times in the vertical direction as shown in FIG. This is a rule for setting the arrangement of output sample points so as to be at a position and non-uniform intervals, and calculating the residual value of the output sample points using a predetermined mathematical formula.

  As shown in FIGS. 2A to 2D, in the resampling rule, the position of the output sample point and the position of the input sample point do not necessarily need to overlap. In the resampling rule, the residual value of the output sampling point is calculated by calculating one or more input sampling points near the output sampling point (a predetermined number of input sampling points existing in a predetermined range starting from the output sampling point). (Sample point) is defined to be calculated based on the residual value.

  For example, the residual value at the position (x, y) of the output sample point is defined as the residual value of the input sample point whose distance (for example, the Euclidean distance) from the position (x, y) is closest. Is also good. Further, for example, the residual value at the position (x, y) of the output sample point is bilinear interpolation (bilinear interpolation) based on the residual values of the four input sample points closest to the position (x, y). May be defined as calculated by

The residual value at the position (x, y) of the output sample point is L (x, y), the residual value at the position (x, y) of the input sample point is H (x, y), and the position of the input sample point is When both the horizontal and vertical coordinates in are lattice points of integers, the following equation is used. This equation performs bilinear interpolation based on residual values of four input sample points.

  FIG. 3 is a diagram illustrating Expression (1) for performing bilinear interpolation based on residual values of four input sample points. In FIG. 3, a mark indicates an input sample point, and a mark indicates an output sample point. As shown in the above equation (1), bilinear interpolation is performed using the residual values H at the four input sample points a1 to a4 indicated by the circles in FIG. Is calculated.

  Note that the resampling unit 12 may apply a thinning filter (for example, a low-pass filter) as preprocessing for resampling. In this case, the resampling unit 12 receives the residual value sequence from the difference unit 11, performs a filtering process on the residual value sequence using a thinning filter, and resamples the sample points after the filtering process.

  Returning to FIG. 1, the inverse resampling unit 18 is provided between the inverse orthogonal transform unit 17 and the adding unit 19. The inverse resampling unit 18 receives the decoding residual value sequence of the block from the inverse orthogonal transforming unit 17, performs the inverse processing of the resampling unit 12, and performs the processing of the number, arrangement, and By changing the difference value, the sampling points are inversely resampled. Then, the inverse resampling unit 18 outputs the decoded residual value sequence of the block after the inverse sampling to the adding unit 19.

  The reverse resampling means 18 may perform reverse resampling by changing only the number of sample points, may perform reverse resampling by changing only the arrangement of sample points, May be inversely resampled by changing only the decoded residual value of. In short, the inverse re-sampling means 18 may inversely re-sample the sample points by changing one, two or all of the number, arrangement, and decoding residual value of the sample points.

  FIG. 4 is a diagram for explaining an example of the reverse resampling process by the reverse resampling unit 18. In FIG. 4, crosses indicate sample points (sample points before reverse re-sampling (input sample points)) of the decoded residual value sequence of the block input to the inverse re-sampling means 18, and circles indicate Shows sample points (sample points after reverse resampling (output sample points)) of the decoded residual value sequence of the block output by the inverse resampling means 18. The inverse resampling rule (inverse resampling rule) for converting the decoded residual value sequence at the input sample points into the decoded residual value sequence at the output sample points is the inverse of the resampling rule shown in FIG. The number and arrangement of the input sample points and the output sample points shown in FIG. 4 are defined, and are set in advance as a conversion rule using a predetermined mathematical expression for calculating the decoding residual value of the output sample points. Shall be.

  The inverse resampling unit 18 receives the parameters of the optimal inverse resampling mode from the optimizing unit 24 described later, and based on the parameters, sets a plurality of preset resampling rules (for example, FIG. 4A). To (d)), one reverse resampling rule is determined. Then, the reverse resampling means 18 resamples the sampling points in accordance with the determined reverse resampling rule.

  FIG. 4A corresponds to FIG. 2A and is an example in which the sampling grid is not changed, and is equivalent to the case where inverse resampling is not performed. The inverse resampling unit 18 resamples the input sample points shown in FIG. 4A according to the preset inverse resampling rule shown in FIG. 4A, and outputs the output sample points shown in FIG. Is obtained. In this case, the inverse resampling unit 18 outputs the input decoded residual value sequence without performing inverse resampling. As shown in FIG. 4A, this inverse resampling rule is a rule that does not change the number and arrangement of input sample points and output sample points, and does not change the decoding residual value.

  FIG. 4B corresponds to FIG. 2B, in which the number of samples is increased (doubled in the horizontal direction and doubled in the vertical direction), and the arrangement after the reverse resampling is made uniform. This is an example of restoration. The inverse resampling means 18 inversely resamples the input sampling points shown in FIG. 4B according to the preset inverse resampling rule shown in FIG. 4B, and outputs the output samples shown in FIG. Find a decoded residual value sequence for the point. As shown in FIG. 4B, this inverse resampling rule increases the number of samples at the input sampling points by a factor of two in the horizontal direction and by a factor of two in the vertical direction so that the original uniform intervals are obtained. This is a rule for setting the arrangement of output sample points and calculating the decoding residual value of the output sample points using a predetermined mathematical formula.

  FIG. 4 (c) corresponds to FIG. 2 (c), and is an example of restoring the arrangement after inverse re-sampling to uniform intervals (restoring the density of sampling points) while maintaining the number of samples. . The inverse resampling means 18 inversely resamples the input sampling points shown in FIG. 4C according to the preset inverse resampling rule shown in FIG. 4C, and outputs the output samples shown in FIG. Find a decoded residual value sequence for the point. In this inverse resampling rule, as shown in FIG. 4C, the arrangement of output sample points is set so that the number of input sample points and the number of output sample points are the same and the original uniform intervals are maintained. A rule for calculating the decoded residual value of the output sample point using a predetermined mathematical formula.

  FIG. 4 (d) corresponds to FIG. 2 (d), in which the number of samples is increased (4 times in the horizontal direction and 4 times in the vertical direction), and the positions after inverse re-sampling are evenly spaced. This is an example of restoring to (the original density of the sample points is restored). The inverse resampling means 18 inversely resamples the input sampling points shown in FIG. 4D according to the preset inverse resampling rule shown in FIG. 4D, and outputs the output samples shown in FIG. Find a decoded residual value sequence for the point. As shown in FIG. 4D, this inverse resampling rule increases the number of input sample points by four times in the horizontal direction and four times in the vertical direction so that the original uniform intervals are obtained. This is a rule for setting the arrangement of output sample points and calculating the decoding residual value of the output sample points using a predetermined mathematical formula.

  As shown in FIGS. 4A to 4D, in the reverse resampling rule, the position of the output sample point and the position of the input sample point do not necessarily have to overlap. In the inverse resampling rule, the decoding residual value of the output sample point is defined to be calculated based on the decoding residual value of one or more input sample points existing near the output sample point. I have.

  For example, the decoding residual value at the position (x, y) of the output sample point may be the decoding residual value of the input sample point having the closest distance (for example, the Euclidean distance) from the position (x, y). Good. Also, for example, the decoding residual value at the position (x, y) of the output sample point is calculated by bilinear interpolation based on the decoding residual values of the four input sample points closest to the position (x, y). May be performed.

  Note that the inverse resampling unit 18 may apply a thinning filter (for example, a low-pass filter) as preprocessing for inverse resampling. In this case, the inverse re-sampling unit 18 receives the decoded residual value sequence from the inverse orthogonal transform unit 17, performs a filtering process on the decoded residual value sequence using a thinning filter, and reverses the sample points after the filtering process. Resample.

  When calculating the decoding residual value of the output sample point, the inverse re-sampling unit 18 calculates the decoding residual value of one or more input sample points present in the vicinity of the output sample point or, if necessary, Based on the decoded residual values of the input sample points that have been processed previously, the high-frequency information lost when the re-sampling unit 12 performs re-sampling may be restored or synthesized by super-resolution technology. I do not care.

  Returning to FIG. 1, the optimizing means 24 includes the block dividing means 10, the orthogonal transform means 13 (corresponding inverse orthogonal transform means 17), the quantizing means 14 (corresponding inverse quantizing means 16), the entropy. In addition to the encoding means 15, the loop filter 20, and the prediction means 22, the parameters used in the resampling means 12 (corresponding to the inverse resampling means 18) are variously changed and the experiment is performed. For example, the optimization unit 24 tries the resampling unit 12 by changing the parameters corresponding to FIGS. In addition, the optimizing unit 24 tries the reverse resampling unit 18 corresponding to the resampling unit 12 by changing the parameters corresponding to FIGS.

  The optimizing unit 24 is, like the optimizing unit 23 shown in FIG. 6, a parameter for which the rate distortion characteristic is the highest or approximates the highest (the value representing the rate distortion characteristic from the highest value to a value lower than the predetermined value by a predetermined amount). Parameters if they exist between them). The specified parameter is data capable of achieving the highest or a coding efficiency that approximates the highest.

  That is, the optimization unit 24 variously changes the parameters of the block division unit 10, the resampling unit 12, and the like, and evaluates the coding efficiency (for example, the degree of good image quality and the degree of small code amount). Is calculated, and a parameter when the evaluation value falls within a predetermined range is specified. The specified parameter is, for example, a parameter that requires the least amount of code when realizing a predetermined image quality and that provides the best image quality when a predetermined amount of code is used.

  The optimizing unit 24 sets the specified parameters and the parameters of the optimal mode, outputs the parameters of the optimal mode to the block dividing unit 10, the resampling unit 12, and the like, and operates according to the parameters. Thereby, the block dividing unit 10, the resampling unit 12, and the like operate according to the parameters of the optimal mode, so that a desired encoding efficiency that is balanced as a whole can be realized.

  For example, the optimization unit 24 specifies, for the resampling unit 12, a parameter of the resampling mode in which the rate distortion characteristic is the highest or approximates the highest as a parameter of the optimal resampling mode. In addition, the optimization unit 24 specifies, for the inverse resampling unit 18, a parameter of the inverse resampling mode in which the rate distortion characteristic is the highest or approximates the maximum as a parameter of the optimal inverse resampling mode.

  The block dividing unit 10, the resampling unit 12, and the like perform processing according to the parameters input from the optimizing unit 24. For example, the resampling unit 12 inputs the parameters of the optimal resampling mode from the optimizing unit 24, and executes the resampling rule indicated by the parameter (for example, any one of the resampling rules shown in FIGS. 2A to 2D). (Sampling rules). The inverse resampling unit 18 receives the parameter of the optimal inverse resampling mode from the optimizing unit 24, and executes the inverse resampling rule indicated by the parameter (for example, any one of the rules shown in FIGS. 4A to 4D). (Re-sampling rule).

  The optimizing unit 24 outputs the parameters of the optimal mode to the entropy encoding unit 15 as optimal mode information.

  As a result, the optimal mode information is entropy-encoded by the entropy encoding unit 15 and output to the video decoding device 2 described later as an encoded bit sequence.

  As described above, the encoding process of the video encoding device 1 converts the quantization index sequence and the optimal mode information into an encoded bit sequence and outputs the encoded bit sequence to the video decoding device 2 described later.

  The optimizing unit 24 includes the resampling unit 12 and the inverse resampling unit 18 in the optimization target, but may not include them. In this case, the resampling unit 12 performs resampling according to a resampling rule indicated by a parameter of a preset resampling mode. The reverse resampling unit 18 performs reverse resampling according to a reverse resampling rule indicated by a preset reverse resampling mode parameter.

  As described above, according to the video encoding device 1 of the embodiment of the present invention, the optimizing unit 24 trials by changing various parameters used in the block dividing unit 10, the resampling unit 12, and the like, A parameter whose rate-distortion characteristic is the highest or approximates the highest is specified, and the specified parameter (parameter in the optimal mode) is output to the block dividing means 10, the resampling means 12, and the like, and the operation is performed according to the parameter. The parameters of the optimal mode specified by the optimizing unit 24 are entropy-encoded by the entropy encoding unit 15 together with the quantization index sequence, and output to the video decoding device 2 described later as an encoded bit sequence.

  The re-sampling unit 12 inputs the residual value sequence of the block, which is the result of subtracting the pixel value sequence of the predicted image from the pixel value sequence of the block, and inputs the parameters (optimal resampling mode According to the resampling rule indicated by the parameter (parameter), the input sample points are resampled to obtain a residual value sequence of the output sample points. The number, arrangement, and residual value of the sampling points are changed by the resampling unit 12 for each block.

  Thereby, the parameter of the optimal mode specified by the optimizing means 24 is data capable of realizing the highest encoding efficiency. Therefore, the resampling rule indicated by the parameter of the optimal mode also has the highest encoding efficiency. It is a feasible rule. In other words, the re-sampling unit 12 performs the encoding process on the blocks divided by the block dividing unit 10 so as to realize the highest encoding efficiency. You can change the value.

  For example, when the residual value of the video signal is a similar value in a predetermined wide range, the optimizing unit 24 specifies a parameter for performing block division in a predetermined wide range for the block dividing unit 10 as an optimal mode parameter. Then, it is assumed that the parameter of the optimal resampling mode for reducing the number of sampling points for the resampling unit 12 is specified. In this case, the resampling unit 12 obtains a residual value sequence in which the number of sampling points is reduced for a predetermined wide range of blocks. As a result, the number of sample points is reduced, so that the calculation load of the subsequent processing can be reduced, and the coding efficiency can be improved.

  Further, for example, when the optimizing unit 24 specifies a parameter of the optimal resampling mode for increasing the number of sampling points for a predetermined position in the block for the resampling unit 12 as the parameter of the optimal mode, The converting means 12 obtains a residual value sequence in which the number of sample points is increased for a predetermined location in the block. As a result, as shown in FIGS. 2C and 2D, a large number of sample points are arranged at necessary positions in the block, or a small number of sample points are arranged at other positions in the block. Since the density of the sample points can be changed, the approximation accuracy of the video signal can be improved, and the coding efficiency can be improved.

  According to the embodiment of the present invention, the processing of the resampling unit 12 is inserted into the encoding process, and the resampling unit 12 performs block division processing by the block division unit 10, orthogonal transformation processing by the orthogonal transformation unit 13, and In conjunction with the quantization processing by the quantization means 14, the number, arrangement, and residual value of sample points are changed in block units.

  In the embodiment of the present invention, unlike the related art in which the pre-processing and the post-processing are added outside the existing coding processing, the re-sampling can be performed in block units. In other words, the resampling process in the embodiment of the present invention is performed in conjunction with other processes (orthogonal transform process, quantization process, etc.) after the block division and other blocks of the same frame and blocks of other frames. A balance can be achieved in consideration of coding efficiency, and an optimum coding process can be realized. As a result, the encoding efficiency of the video signal can be improved, and the video signal can be compression-encoded and decoded with high image quality.

[Video decoding device]
Next, a video decoding device according to an embodiment of the present invention will be described. FIG. 5 is a block diagram illustrating a configuration example of the video decoding device according to the embodiment of the present invention. The video decoding device 2 includes an entropy decoding unit 39, an inverse quantization unit 32, an inverse orthogonal transformation unit 33, an inverse resampling unit 34, an addition unit 35, a loop filter 36, a memory 37, and a prediction unit 38.

  Comparing the conventional video decoding device 200 shown in FIG. 7 with the video decoding device 2 according to the embodiment of the present invention shown in FIG. 5, both video decoding devices 2 and 200 have the inverse quantization means 32 and the inverse orthogonal transform. They are the same in that they have a means 33, an adding means 35, a loop filter 36, a memory 37 and a predicting means 38.

  On the other hand, the video decoding device 2 further includes an inverse resampling unit 34 in addition to the configuration of the video decoding device 200, and an entropy decoding unit 39 different from the entropy decoding unit 31 provided in the video decoding device 200. This is different from the video decoding device 200. In FIG. 5, portions common to FIG. 7 are denoted by the same reference numerals as in FIG. 7, and detailed description thereof will be omitted.

  The video decoding device 2 receives a quantization index sequence and an encoded bit sequence such as optimal mode information from the video encoding device 1, performs decoding processing including inverse re-sampling according to parameters of the optimal mode information, and Restore the original video signal frame from the column. Then, the video decoding device 2 outputs a decoded video signal.

  The entropy decoding unit 39 receives the encoded bit sequence from the video encoding device 1, performs a process reverse to that of the entropy encoding unit 15 shown in FIG. 1, and performs entropy decoding on the encoded bit sequence to obtain the original information. Reversibly decode. The original information includes, in addition to the quantization index sequence, the motion vector information obtained by the prediction unit 22 shown in FIG. 1, the optimal mode information obtained by the optimization unit 23, the information such as the size of the video, and the like. , And information added by the user or the like. The optimal mode information includes parameters of the optimal inverse resampling mode.

  The entropy decoding unit 39 outputs the parameters of the optimal mode, which are the entropy-decoded optimal mode information, to the inverse quantization unit 32, the inverse orthogonal transform unit 33, the inverse resampling unit 34, the loop filter 36, and the prediction unit 38, Operate according to the parameters. In this case, the entropy decoding unit 39 outputs the parameters of the optimal inverse resampling mode to the inverse resampling unit 34.

  Accordingly, the inverse quantization means 32, the inverse orthogonal transform means 33, the inverse resampling means 34, the loop filter 36, and the prediction means 38 perform processing according to the parameters of the optimal mode input from the entropy decoding means 39.

  Further, the entropy decoding unit 39 outputs the quantization index sequence of the block subjected to the entropy decoding to the inverse quantization unit 32.

  The inverse resampling means 34 is provided between the inverse orthogonal transform means 33 and the adding means 35. The inverse re-sampling unit 34 receives the decoded residual value sequence of the block from the inverse orthogonal transform unit 33 and performs the same processing as the inverse re-sampling unit 18 shown in FIG. , The sample points are inversely resampled by changing the number, arrangement, and decoding residual value of the sample points in block units. Then, the inverse resampling unit 34 outputs the decoded residual value sequence of the block after the inverse sampling to the adding unit 35.

  The reverse resampling means 34 may perform reverse resampling by changing only the number of sample points, may perform reverse resampling by changing only the arrangement of sample points, May be inversely resampled by changing only the decoded residual value of. In short, the inverse re-sampling unit 34 may inversely re-sample the sample points by changing one, two, or all of the number, arrangement, and decoding residual value of the sample points.

  As described above, by the decoding process of the video decoding device 2, the coded bit sequence is converted into the quantization index sequence and the optimal mode information and the like, and the frame of the original video signal is restored.

  Note that the case has been described where the video decoding device 2 inputs a coded bit sequence including the parameters of the optimal inverse resampling mode from the video encoding device 1, but the coded bit sequence including the parameters of the optimal inverse resampling mode is input. It may not be. In this case, the inverse resampling unit 34 performs inverse resampling according to the inverse resampling rule indicated by the parameter of the preset inverse resampling mode.

  As described above, according to the video decoding device 2 of the embodiment of the present invention, the entropy decoding unit 39 performs entropy decoding on the coded bit sequence input from the video coding device 1, and performs the quantization index sequence and the parameter of the optimal mode. Is generated, and the parameters are output to the inverse quantization means 32, the inverse resampling means 34, and the like, and are operated according to the parameters.

  The inverse re-sampling unit 34 inputs the decoded residual value sequence of the block obtained by subtracting the pixel value sequence of the prediction image from the pixel value sequence of the block, and inputs the parameter (optimum inverse re-sampling) input from the entropy decoding unit 39. According to the inverse resampling rule shown by the sampling mode parameter), the input sample points are inversely resampled to obtain a decoded residual value sequence of the output sample points. The inverse resampling unit changes the number, arrangement, and decoding residual value of the sample points in block units, and returns to the original. Then, the original video signal is decoded.

  Thereby, the parameters of the optimal mode input from the video encoding device 1 and decoded by the entropy decoding means 39 are data capable of realizing the highest encoding efficiency. The coding rule is also a rule that can achieve the highest coding efficiency. That is, the inverse re-sampling unit 34 calculates the number, arrangement, and decoding residual value of the sample points in units of blocks in the decoding process in the video decoding device 2 so as to realize the highest coding efficiency. Can be changed to return to

  For example, when the residual value of the video signal has a similar value in a predetermined wide range, the optimizing unit 24 of the video encoding device 1 uses the block dividing unit 10 as a parameter of the optimal mode to block the block in the predetermined wide range. It is assumed that a parameter to be divided is specified, and a parameter of the optimal resampling mode for the resampling unit 12 that reduces the number of sampling points is specified. In this case, the entropy decoding unit 39 of the video decoding device 2 decodes, as the parameter of the optimal mode, the parameter for the block that has been divided into blocks in a predetermined wide range, and determines the number of sampling points for the inverse re-sampling unit 34. Decode the increasing optimal inverse resampling mode parameters. Then, the inverse re-sampling unit 34 decodes the blocks having a larger number of sample points corresponding to the re-sampling unit 12 and the inverse re-sampling unit 18 of the video encoding device 1 for a predetermined wide range of blocks. A sequence of residual values will be obtained. As a result, the number of sample points can be restored, and coding efficiency can be improved.

  Further, for example, the optimizing unit 24 of the video encoding device 1 sets the parameter of the optimal resampling mode for increasing the number of sampling points for a predetermined location in the block for the resampling unit 12 as the parameter of the optimal mode. Suppose that it is specified. In this case, the entropy decoding unit 39 of the video decoding device 2 decodes the parameter of the optimal re-resampling mode for the inverse re-sampling unit 34 to reduce the number of sampling points for a predetermined position in the block. Then, the inverse resampling unit 34 corresponds to the resampling unit 12 and the inverse resampling unit 18 of the video encoding device 1 and, as shown in FIGS. For each predetermined location, a sequence of decoded residual values in which the number of sample points is reduced is obtained. As a result, the number of sample points can be restored, and the coding efficiency can be improved.

  In the embodiment of the present invention, the processing of the inverse re-sampling unit 34 is inserted into the decoding process, and the inverse re-sampling unit 34 performs the inverse quantization process by the inverse quantization unit 32 and the inverse orthogonal transform unit 33. In conjunction with the inverse orthogonal transform processing, the number, arrangement, and decoding residual value of the sample points are changed back to the original values in block units.

  In the embodiment of the present invention, unlike the related art, inverse resampling can be performed in block units. That is, the inverse resampling process in the embodiment of the present invention interlocks with other processes (inverse quantization process, inverse orthogonal transform process, and the like) to interoperate with other blocks of the same frame and blocks of other frames. Thus, a balance can be achieved in consideration of the encoding efficiency, and a decoding process corresponding to an optimal encoding process can be realized. As a result, the encoding efficiency of the video signal can be improved, and the video signal can be compression-encoded and decoded with high image quality.

  As described above, the present invention has been described with reference to the embodiment. However, the present invention is not limited to the above embodiment, and can be variously modified without departing from the technical idea thereof. For example, in the above embodiment, the video encoding device 1 illustrated in FIG. 1 includes the loop filter 20, but does not necessarily need to include the loop filter 20. When the video encoding device 1 does not include the loop filter 20, the decoded pixel value sequence obtained by the adding unit 19 is stored in the memory 21. Here, the adding unit 19 stores the decoded pixel value sequence in the memory 21 while considering the position of the block. Thus, decoded images are sequentially formed in the memory 21.

  Although the video decoding device 2 shown in FIG. 5 includes the loop filter 36, the video decoding device 2 does not necessarily need to include the loop filter 36. When the video decoding device 2 does not include the loop filter 36, the decoded pixel value sequence obtained by the adding unit 35 is sequentially stored in the memory 37 at the position of the corresponding block. Thus, decoded images are sequentially formed in the memory 37.

  In the above embodiment, the resampling unit 12, the orthogonal transform unit 13, the quantization unit 14, the entropy encoding unit 15, the inverse quantization unit 16, and the inverse orthogonal transform unit of the video encoding device 1 shown in FIG. 17 and the inverse resampling unit 18 perform processing on each sample point of the block with respect to a residual value (decoding residual value) obtained by subtracting the pixel value of the predicted image from the pixel value of the video signal. I did it. On the other hand, the resampling unit 12, the orthogonal transformation unit 13, the quantization unit 14, the entropy encoding unit 15, the inverse quantization unit 16, the inverse orthogonal transformation unit 17, and the inverse resampling unit 18 perform The processing may be performed on the pixel value (decoded pixel value) of the sample point instead of the difference value (decoded residual value). The same applies to the inverse quantization means 32, the inverse orthogonal transform means 33, the inverse resampling means 34, etc. of the video decoding device 2 shown in FIG.

  In the embodiment, the optimizing unit 24 of the video encoding device 1 shown in FIG. 1 includes the block dividing unit 10, the resampling unit 12 (the corresponding inverse resampling unit 18), the orthogonal transform unit. 13 (corresponding inverse orthogonal transform means 17), quantizing means 14 (corresponding inverse quantizing means 16), entropy coding means 15, loop filter 20 and prediction means 22 vary various parameters. And tried to identify the optimal parameters. On the other hand, the optimizing unit 24 may change the parameters used in one or more of these components to perform the experiment. The component that is not the target of the trial performs processing according to the preset parameters.

  As a hardware configuration of the video encoding device 1 and the video decoding device 2 according to the embodiment of the present invention, an ordinary computer can be used. The video encoding device 1 and the video decoding device 2 are configured by a computer including a CPU, a volatile storage medium such as a RAM, a nonvolatile storage medium such as a ROM, an interface, and the like.

  Block dividing means 10, difference means 11, resampling means 12, orthogonal transform means 13, quantization means 14, entropy encoding means 15, inverse quantization means 16, inverse orthogonal transform means 17 provided in video encoding apparatus 1. , The inverse resampling unit 18, the adding unit 19, the loop filter 20, the memory 21, the predicting unit 22, and the optimizing unit 23 are realized by causing a CPU to execute a program describing these functions. .

  Also, each function of the entropy decoding means 39, the inverse quantization means 32, the inverse orthogonal transform means 33, the inverse resampling means 34, the addition means 35, the loop filter 36, the memory 37 and the prediction means 38 provided in the video decoding device 2 Is realized by causing a CPU to execute a program describing these functions.

  These programs are stored in the storage medium, and are read and executed by the CPU. These programs can also be stored and distributed on storage media such as magnetic disks (floppy (registered trademark) disks, hard disks, etc.), optical disks (CD-ROMs, DVDs, etc.), semiconductor memories, etc., and can be distributed via a network. You can also send and receive.

1,100 video encoding device 2,200 video decoding device 10 block dividing means 11 difference means 12 resampling means 13 orthogonal transform means 14 quantization means 15 entropy encoding means 16, 32 inverse quantization means 17, 33 inverse orthogonal Conversion means 18, 34 Inverse resampling means 19, 35 Addition means 20, 36 Loop filter 21, 37 Memory 22, 38 Prediction means 23, 24 Optimization means 31, 39 Entropy decoding means

Claims (4)

A frame of a video signal is divided into blocks, and a pixel value of each sample point of the block or a residual value that is a difference between the pixel value and a pixel value of a predicted image is subjected to orthogonal transformation and quantization. , A video encoding device that outputs an encoded bit sequence entropy encoded,
Re-sampling means for re-sampling the sample points by changing at least one of the number, arrangement, and the pixel value or the residual value of the sample points in units of the block,
Changing the parameters used in the resampling unit, and a plurality of parameters including the parameters used in the process of dividing the block, the process of the orthogonal transform, and the process of the quantization, and representing the coding efficiency Optimizing means for determining a value, respectively, the plurality of parameters when the evaluation value is a value in a predetermined range, as an optimal plurality of parameters,
The resampling means resampled, the orthogonal index and the quantized index sequence subjected to the quantization, and the plurality of parameters specified by the optimizing means are entropy-encoded, and the encoded bit sequence And entropy encoding means for generating
The resampling means,
A video encoding device, wherein the sampling points are resampled in accordance with a resampling rule corresponding to the parameter specified by the optimizing means. The video encoding device according to claim 1 ,
When the sample points input by the resampling unit are input sample points, and the sample points resampled by the resampling unit are output sample points,
The resampling rule sets the arrangement of the output sample points so as to reduce the number of the input sample points or to change the density of the input sample points in units of the block, A rule for calculating the pixel value or the residual value of the output sample point based on the pixel value or the residual value of a predetermined number of the input sample points existing in a predetermined range based on the sample point. , A video encoding device. A video decoding apparatus for receiving a coded bit string output by the video coding apparatus according to claim 1 , entropy decoding the coded bit string, performing inverse quantization and inverse orthogonal transform, and restoring a frame of an original video signal. ,
Entropy decoding means for entropy decoding the coded bit sequence and generating a quantization index sequence and a plurality of parameters;
In a process reverse to the process in which the video encoding device resamples the sample points in units of blocks obtained by dividing the frame of the video signal, a reverse recoding corresponding to the parameter generated by the entropy decoding unit is performed. According to the sampling rule, the number of the sample points, the arrangement, and the quantization index sequence by changing at least one of the inverse orthogonal transform and the inverse orthogonal transformed decoded pixel value or decoding residual value, Reverse resampling means for reverse resampling the sample points,
A video decoding apparatus, wherein a frame of the original video signal is restored based on the decoded pixel value or the decoded residual value inversely resampled by the inverse resampling unit. The video decoding device according to claim 3 ,
When the sample points input by the inverse re-sampling means are input sample points, and the sample points inversely re-sampled by the inverse re-sampling means are output sample points,
The inverse resampling rule sets the arrangement of the output sample points so as to increase the number of the sample points or to change the density of the input sample points in units of the block, and Calculating the decoded pixel value or the decoded residual value of the output sample point based on the decoded pixel value or the decoded residual value of a predetermined number of the input sample points existing in a predetermined range based on the sample point. A video decoding device, characterized in that:

Images