JP2013085306A - Moving image encoding device, moving image decoding device, moving image encoding method, moving image decoding method, moving image encoding program, and moving image decoding program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance processing efficiency in encoding processing and decoding processing by suppressing increase of the number of codes in additional information.SOLUTION: A moving image decoding device 200 which generates a prediction signal by block unit, comprises: a low-resolution block decoding unit 151 which generates a low-resolution block that has less number of pixels than a prediction block by decoding encoded data; an expanded block generation unit 154 which expands the low-resolution block generated from decoded low-resolution block by using a decoded image, into a block that has the same number of pixels as the prediction block; a block division unit 162 which divides the expanded block on the basis of a predetermined division rule, and generates a plurality of small blocks; and a small block prediction unit 152 which generates prediction small blocks for the small blocks by using the decoded image and the plurality of small blocks.

Description

  The present invention relates to a moving picture decoding technique and a moving picture encoding technique for generating a prediction block in units of blocks.

  Conventionally, as an existing coding method for generating a prediction signal (prediction block) of a prediction target block in units of blocks, for example, a technique described in Patent Literature 1 has been proposed. This technology reduces redundancy in the time direction by entropy coding a motion vector indicating motion between frames and an error block between a prediction block generated from this motion vector and a prediction target block. To do. Here, the prediction target block is a divided block of the input frame on which the prediction process is performed next.

JP 2002-354486 A

  In order to improve the temporal prediction performance of an image region including a fine motion expression, information such as a large number of motion vectors (hereinafter referred to as “motion information”) is required. However, since the above-described conventional method requires coding of motion information, the amount of code of so-called additional information including motion information increases. This is a factor that makes it difficult to improve processing efficiency in encoding processing and decoding processing.

  Accordingly, an object of the present invention is to improve the processing efficiency in the encoding process and the decoding process by suppressing the increase in the code amount of the additional information described above.

  A moving image encoding apparatus according to the present invention is a moving image encoding apparatus that generates a prediction block in units of blocks, and includes a holding unit (memory) that stores a decoded image, and an input image to be encoded Are divided into a plurality of prediction target blocks, and the prediction target block is converted and encoded into a low resolution block having a smaller number of pixels than the prediction target block, and then the low resolution block Generating means for generating a decoded low-resolution block by decoding the encoded data, and an expanding means for expanding the generated decoded low-resolution block to a block having the same number of pixels as the prediction target block; A second dividing means for generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule; the generated small blocks; and the decoded image; Prediction means for generating, as a predicted small block, a block in the decoded image corresponding to a motion vector of the generated small block detected by block matching between the plurality of generated predicted small blocks, And integrating means for generating a prediction block by integrating based on the division rule.

  A moving picture decoding apparatus according to the present invention is a moving picture decoding apparatus that generates a prediction block in units of blocks, and stores a decoding unit that holds decoded images and decodes input encoded data. Decoding means for generating a low-resolution block having a smaller number of pixels than the prediction block based on the low-resolution block information and the decoded image obtained, and the generated low-resolution block as the prediction block Expanding means for expanding the block to the same number of pixels, dividing means for generating a plurality of small blocks by dividing the enlarged block based on a predetermined dividing rule, and the divided small blocks A block in the decoded image corresponding to a motion vector of the divided small block detected by block matching with the decoded image is predicted. Prediction means for generating a click, a plurality of predicted small blocks the generation, by integrating based on the division rule, and a merging unit for generating a prediction block.

  In order to perform predictive coding with high accuracy between the moving picture coding apparatus and the moving picture decoding apparatus, in the moving picture coding apparatus, a low resolution block having a smaller number of pixels than the prediction block is compressed and coded. Is preferably transmitted to the video decoding device. In this respect, according to these inventions, since the number of pixels of the low resolution block is smaller than the number of pixels of the prediction block, the encoded data in which the low resolution block is compressed is additional information than the conventional motion information. The amount of codes is reduced. Thereby, the increase in the code amount of the additional information is reduced, and as a result, the processing efficiency in the encoding process and the decoding process is improved.

  In particular, in the moving image decoding apparatus, first, an enlarged block is divided to obtain a small block, and a predicted small block is generated from the small block and the decoded image. As a result, a plurality of small prediction blocks capable of performing high-precision prediction of a fine area of the moving image are obtained. And a prediction block is produced | generated by integrating these. Therefore, simultaneously with the improvement of the processing efficiency described above, it is possible to perform highly accurate prediction using a low resolution block with a small code amount.

  The above-described moving image encoding apparatus subtracts the divided encoding target block and the prediction block in units of pixels by a third dividing unit that divides the encoding target image into encoding target blocks. Subtracting means for generating an error block; encoding means for encoding the generated error block; decoding means for decoding the encoded data and generating a decoded error block; and the prediction block It is desirable to further include output means for adding the decoded error block and the decoded error block in a pixel unit to generate a decoded block and outputting the decoded block to the holding means.

  Further, in the moving image encoding apparatus, the enlargement unit uses the generated decoded low-resolution block pixels and the decoded block pixels adjacent to the upper side and the left side of the prediction target block, and is the same as the prediction target block. It is desirable to generate a block of pixels.

  In the moving image encoding apparatus, the generation unit converts the prediction target block into a low resolution prediction target block having a smaller number of pixels than the prediction target block based on a predetermined conversion rule; A low-resolution prediction block generating means for generating a prediction block; means for generating a low-resolution error block by subtracting the low-resolution prediction target block and the low-resolution prediction block in units of pixels; Means for generating a decoded low-resolution error block by decoding encoded data of the low-resolution error block after encoding the image error block, the decoded low-resolution error block, and the low-resolution prediction block And reconstructing the low-resolution block by pixel by pixel, and the low-resolution prediction block generation unit includes a second prediction block having the same number of pixels as the prediction block. And means for generating the low-resolution prediction block from the second prediction block based on the conversion rule, and the generation means is further included in the low-resolution error block. In the case where all the difference pixel values of are zero, it is desirable to include means for outputting the second prediction block as the prediction block.

  The moving image decoding apparatus described above generates a decoding block by decoding the encoded data and generating an error block, adding the prediction block and the error block in units of pixels, and generating the decoding block. It is desirable to further include output means for outputting to the holding means.

  Further, in the moving image decoding apparatus, the enlargement unit uses a pixel of the generated low-resolution block and a pixel of the decoded block adjacent to the upper side and the left side of the prediction block, and has the same number of pixels as the prediction block It is desirable to generate

  Further, in the moving image decoding apparatus, the decoding unit decodes the input encoded data to decode a low resolution error block, and a low resolution prediction block generation unit generates a low resolution prediction block. Means for adding the low resolution error block and the low resolution prediction block in units of pixels to generate the low resolution block, and the low resolution prediction block generation means includes the prediction Means for generating a second prediction block having the same number of pixels as the block, and means for generating the low-resolution prediction block from the second prediction block based on a predetermined rule, the decoding means Further, it is preferable that the device further includes means for outputting the second prediction block as the prediction block when all the difference pixel values in the low resolution error block are zero.

  Furthermore, the moving image encoding / decoding technique according to the present invention can take the form of a method or a program as well as an apparatus, and a moving image encoding method, a moving image decoding method, and a moving image, whose features will be described below, will be described below. The image encoding program and the moving image decoding program have the same effects as the above-described effects related to the moving image encoding device and the moving image decoding device.

  That is, the video encoding method according to the present invention is a video encoding method for generating a prediction block in units of blocks using decoded images, and an input encoding target image is converted into a plurality of prediction targets. A first division step of dividing the block into blocks, and converting the prediction target block into a low resolution block having a smaller number of pixels than the prediction target block, and then decoding the encoded data of the low resolution block A generation step for generating a decoded low-resolution block, and an enlargement step for expanding the generated decoded low-resolution block to a block having the same number of pixels as the prediction target block using the decoded image. A second dividing step of generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule; and the generated small blocks A prediction step for generating a block in the decoded image corresponding to a motion vector of the generated small block detected by block matching between a block and the decoded image as a predicted small block; And integrating a plurality of predicted small blocks based on the division rule to generate a predicted block.

  A moving picture decoding method according to the present invention is a moving picture decoding method for generating a prediction block in units of blocks using decoded pictures, and includes low-resolution block information obtained by decoding input encoded data, A decoding step for generating a low-resolution block having a smaller number of pixels than a prediction block based on the decoded image, and using the decoded image, the generated low-resolution block is converted into the prediction block. An enlargement step for enlarging the block with the same number of pixels, a division step for generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule, the divided small block, and the A block in the decoded image corresponding to the motion vector of the divided small block detected by block matching with the decoded image is predicted. A prediction step of generating a small block, a plurality of predicted small blocks said generated by integrating based on the division rule, and a consolidation step of generating a prediction block.

  Further, the moving picture coding program according to the present invention is a coding that is input to a computer having holding means for holding a decoded picture and a moving picture coding program having a function of generating a prediction block in units of blocks. A first division process for dividing the target image into a plurality of prediction target blocks, and after converting and encoding the prediction target block into a low-resolution block having a smaller number of pixels than the prediction target block, A decoding process for generating a decoded low-resolution block by decoding the encoded data of the resolution block, and using the decoded image, the generated decoded low-resolution block is the same as the prediction target block. An enlargement process for enlarging the number of pixels into a block and a plurality of small blocks are generated by dividing the enlarged block based on a predetermined division rule. And a block in the decoded image corresponding to a motion vector of the generated small block detected by block matching between the generated small block and the decoded image as a predicted small block. A prediction process to be generated and an integrated process to generate a prediction block are performed by integrating the plurality of generated prediction small blocks based on the division rule.

  A moving picture decoding program according to the present invention decodes input encoded data to a computer having holding means for holding a decoded picture and a moving picture decoding program having a function of generating a prediction block in units of blocks. Based on the low-resolution block information obtained and the decoded image, a decoding process for generating a low-resolution block having a smaller number of pixels than the predicted block, and the generated low-resolution block using the decoded image. An enlargement process for enlarging the resolution block to a block having the same number of pixels as the prediction block; and a division process for generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule; The motion vector of the divided small block detected by block matching between the divided small block and the decoded image A prediction process for generating a corresponding block in the decoded image as a prediction small block, and an integration process for generating a prediction block by integrating the generated plurality of prediction small blocks based on the division rule; Is executed.

  In order to achieve the above object, a video decoding apparatus according to the present invention is a video decoding apparatus that decodes encoded data to be decoded in units of a predetermined block to generate a prediction block. By holding the held means and entropy decoding the input encoded data, the number of pixels is smaller than that of the decoded block, and the low-resolution block at the same position in time and space as the decoded block is encoded. Entropy decoding means for generating low-resolution block information, and low-resolution block for decoding the low-resolution block information generated by the entropy decoding means and generating the low-resolution block with reference to the decoded image A block decoding unit and a plurality of low resolution divisions by dividing the low resolution block generated by the low resolution block decoding unit based on a predetermined division rule A low resolution block dividing unit that generates a lock, a decoded image held in the holding unit, and a low resolution divided block generated by the low resolution block dividing unit. A predicted small block generating means for generating a predicted small block having a larger number of pixels than that of the low-resolution divided block at the same position in time and space, and a predicted small block generated by the predicted small block generating means, And a block integration unit that generates a prediction block by integrating based on a predetermined integration rule corresponding to the division rule.

  In order to achieve the above object, a moving picture encoding apparatus according to the present invention is a moving picture encoding apparatus that generates a prediction block by encoding an image to be encoded in units of a predetermined block. A holding unit that holds a completed image, an encoding block dividing unit that divides the input image into a plurality of encoding target blocks, and a number of pixels that is larger than that of the encoding target block divided by the encoding block dividing unit. The low-resolution block information is generated by encoding the low-resolution block at the same position in time and space as the encoding target block, and the decoded image is decoded. With reference to the decoded low-resolution block generating means for generating the decoded low-resolution block, and the decoded low-resolution block generated by the decoded low-resolution block generating means in accordance with a predetermined division rule. The low resolution block dividing means for generating a plurality of low resolution divided blocks, the decoded image held in the holding means, and the low resolution generated by the low resolution block dividing means. A predicted small block generating means for generating a predicted small block having a larger number of pixels than the low resolution divided block and at the same position temporally and spatially as the low resolution divided block from the image divided block; And a block integration unit that generates a prediction block by integrating the predicted small blocks generated by the block generation unit based on a predetermined integration rule corresponding to the division rule.

  In order to perform highly accurate predictive coding between the moving picture coding apparatus and the moving picture decoding apparatus, in the moving picture coding apparatus, a low resolution block obtained by compressing and coding a low resolution block having a smaller number of pixels than the coding target block. It is desirable that the resolution block information is entropy encoded and transmitted to the video decoding device. The low-resolution block has a smaller number of pixels than the encoding target block, and the low-resolution block information has a smaller amount of additional information than conventional prediction information required for prediction of the encoding target block. An increase in the code amount of the additional information can be avoided and the processing efficiency can be improved.

  In the video decoding device, the low-resolution block is decoded from the low-resolution block information, the obtained low-resolution block is divided into a plurality of low-resolution divided blocks, and each low-resolution divided block is decoded. From the image, a predicted small block having a larger number of pixels than the low resolution divided block and the same position in time and space as the low resolution divided block is generated, and the predicted block is integrated by integrating the predicted small blocks. It is desirable to generate. In this case, the low-resolution block is further divided to obtain a low-resolution divided block, and a small predicted block is generated from the low-resolution divided block and the decoded image. A prediction block can be generated by obtaining a prediction small block capable of performing accurate prediction and integrating the prediction small blocks. That is, highly accurate prediction can be performed using low resolution block information with a small code amount, and at the same time, processing efficiency can be improved.

  In the moving picture decoding apparatus according to the present invention, the entropy decoding means performs error entropy decoding on the encoded data to obtain error block information representing an error between the encoding target block and the prediction block included in the encoded data. Further, the moving picture decoding apparatus decodes the error block information generated by the entropy decoding unit, thereby generating an error block decoding unit, and a prediction block generated by the block integration unit. It is preferable that the apparatus further comprises decoding block generation means for generating a decoded block based on the error block generated by the error block decoding means.

  The moving picture encoding apparatus according to the present invention obtains an error between the encoding target block divided by the encoding target block dividing unit and the prediction block generated by the block integration unit in units of pixels, Error block generating means for generating an error block representing an error, error block encoding means for generating error block information by encoding the error block generated by the error block generating means, and the error block encoding means Decoding error block information generated by the error block decoding means for generating a decoded error block, a prediction block generated by the block integration means, and a decoding error block generated by the error block decoding means, And a decoding block generating means for generating a decoding block based on It is desirable to.

  Further, the moving picture encoding apparatus according to the present invention is an entropy for entropy encoding the low resolution block information generated by the decoded low resolution block generation means and the error block information generated by the error block encoding means. It is desirable to further include encoding means.

  The present invention can also be understood as a moving image decoding method, a moving image encoding method, a moving image decoding program, and a moving image encoding program, and can be described as follows. These have the same effects as the above-described video decoding device and video encoding device.

  The moving picture decoding method according to the present invention includes a holding means for holding a decoded picture, and the moving picture decoding in a moving picture decoding apparatus that generates a prediction block by decoding encoded data to be decoded in a predetermined block unit. A method that encodes a low-resolution block at the same position temporally and spatially as the decoded block with fewer pixels than the decoded block by entropy decoding the input encoded data to be decoded An entropy decoding step for generating the low-resolution block information, a low-resolution block information generated by decoding the low-resolution block information generated in the entropy decoding step, and generating the low-resolution block by referring to the decoded image By dividing the low resolution block generated in the resolution block decoding step and the low resolution block decoding step based on a predetermined division rule, a plurality of low resolution blocks A low resolution block dividing step for generating a divided block, a decoded image held in the holding unit, and the low resolution divided block generated in the low resolution block dividing step, the low resolution A predicted small block generating step for generating a predicted small block having a larger number of pixels than the divided block and temporally and spatially the same position as the low resolution divided block, and the predicted small block generated in the predicted small block generating step And a block integration step of generating a prediction block by integrating the blocks based on a predetermined integration rule corresponding to the division rule.

  A moving image encoding method according to the present invention includes a holding unit that holds a decoded image, and a moving image in a moving image encoding apparatus that generates a prediction block by encoding an image to be encoded in a predetermined block unit. An encoding method, which is an encoding block dividing step for dividing an input encoding target image into a plurality of encoding target blocks, and pixels than the encoding target block divided in the encoding block dividing step A low-resolution block information is generated by encoding a low-resolution block in the same position temporally and spatially as the encoding target block with a small number, and the low-resolution block information is decoded and the decoded A decoded low-resolution block generating step for generating a decoded low-resolution block with reference to an image, and the decoded low-resolution block generated in the decoded low-resolution block generating step are divided into predetermined division rules. Generated in the low resolution block dividing step for generating a plurality of low resolution divided blocks, the decoded image held in the holding means, and the low resolution block dividing step. A predicted small block generating step for generating a predicted small block having a larger number of pixels than the low resolution divided block and temporally and spatially the same position as the low resolution divided block from the low resolution divided block; And a block integration step of generating a prediction block by integrating the prediction small block generated in the prediction small block generation step based on a predetermined integration rule corresponding to the division rule.

  A moving picture decoding program according to the present invention is provided in a moving picture decoding apparatus that includes a holding unit that holds a decoded image and that decodes encoded data to be decoded in a predetermined block unit to generate a prediction block. The computer encodes a low-resolution block in which the number of pixels is smaller than that of the decoded block and the same position in time and space as the decoded block by entropy decoding the input encoded data to be decoded. Entropy decoding means for generating resolution block information, and low resolution block for decoding the low resolution block information generated by the entropy decoding means and generating the low resolution block with reference to the decoded image By dividing the low resolution block generated by the decoding means and the low resolution block decoding means based on a predetermined division rule, a plurality of From the low resolution block dividing means for generating the resolution divided block, the decoded image held in the holding means, and the low resolution divided block generated by the low resolution block dividing means, the low resolution Predicted small block generating means for generating a predicted small block having a larger number of pixels than the image divided block and temporally and spatially the same position as the low resolution divided block, and the predicted small block generated by the predicted small block generating means The blocks are integrated based on a predetermined integration rule corresponding to the division rule to function as a block integration unit that generates a prediction block.

  A moving picture coding program according to the present invention is provided in a moving picture coding apparatus that includes a holding unit that holds a decoded image and generates a prediction block by coding a picture to be coded in a predetermined block unit. A coding block dividing unit that divides an input encoding target image into a plurality of encoding target blocks, and a smaller number of pixels than the encoding target block divided by the encoding target block dividing unit. By encoding a low-resolution block at the same position in time and space as the current block to be encoded, low-resolution block information is generated, the low-resolution block information is decoded, and the decoded image is With reference to a decoded low-resolution block generating means for generating a decoded low-resolution block, and a decoded low-resolution block generated by the decoded low-resolution block generating means By dividing based on the division rule, a low resolution block dividing unit that generates a plurality of low resolution divided blocks, a decoded image held in the holding unit, and a low resolution block dividing unit A predicted small block generating means for generating a predicted small block having the same number of pixels as the low resolution divided block temporally and spatially from the low resolution divided block; The prediction small block generated by the prediction small block generation unit is integrated based on a predetermined integration rule corresponding to the division rule to function as a block integration unit that generates a prediction block.

  According to the present invention, since the increase in the code amount of the additional information is suppressed, the processing efficiency in the encoding process and the decoding process can be increased.

It is a figure which shows the structure of the moving image encoder in 1st Embodiment. It is a figure which shows the structure of the moving image decoding apparatus in 1st Embodiment. It is a figure for demonstrating block matching. It is a figure for demonstrating a block reduction process. It is a 1st figure for demonstrating a block expansion process. It is a 2nd figure for demonstrating a block expansion process. It is a figure for demonstrating 4x4 division | segmentation of an expansion block. It is a figure for demonstrating 2x2 division | segmentation of an expansion block. It is a figure for demonstrating 4x4 small block motion estimation processing. It is a figure for demonstrating a 2 * 2 small block motion estimation process. It is a flowchart for demonstrating a block encoding process. It is a flowchart for demonstrating a low-resolution block encoding and decoding process. It is a flowchart for demonstrating a small block motion estimation process. It is a flowchart for demonstrating a block decoding process. It is a flowchart for demonstrating a low-resolution block decoding process. It is a figure explaining the data storage medium which stores the program for implement | achieving a moving image encoding process or a moving image decoding process with a computer system. It is a figure which shows the structure of the moving image encoding program of 1st Embodiment. It is a figure which shows the structure of the moving image decoding program of 1st Embodiment. It is a figure for demonstrating block matching with block division. It is a figure for demonstrating the 4th deformation | transformation aspect regarding block size. It is a figure for demonstrating the block expansion process in a 5th deformation | transformation aspect. It is a figure which shows the structure of the moving image encoder in the 5th deformation | transformation aspect. It is a figure which shows the structure of the moving image decoding apparatus in a 5th deformation | transformation aspect. It is a figure for demonstrating the block reduction process in a 6th deformation | transformation aspect. It is a figure which shows the structure of the moving image encoder in the 8th deformation | transformation aspect. It is a figure which shows the structure of the moving image decoding apparatus in an 8th deformation | transformation aspect. It is a figure for demonstrating the block expansion process in an 8th deformation | transformation aspect. It is a figure explaining the moving image decoding apparatus in 2nd Embodiment. It is a figure explaining the moving image encoder in 2nd Embodiment. It is the figure which showed the block decoding processing flow in 2nd Embodiment. It is the figure which showed the low-resolution block decoding processing flow in 2nd Embodiment. It is the figure which showed the small block prediction process flow in 2nd Embodiment. It is the figure which showed the block encoding process blow in 2nd Embodiment. It is the figure which showed the low-resolution block encoding / decoding processing flow in 2nd Embodiment. It is a figure for demonstrating a division | segmentation block prediction process. It is a figure for demonstrating block pixel arrangement | positioning. It is a block diagram of the moving image decoding program of 2nd Embodiment. It is a block diagram of the moving image encoding program of 2nd Embodiment.

[First Embodiment]
Hereinafter, a first embodiment of the present invention will be described with reference to the accompanying drawings. FIG. 1 is a diagram showing a configuration of a moving picture encoding apparatus (hereinafter simply referred to as “encoding apparatus 100”) capable of realizing predictive encoding according to the present invention. Similarly, FIG. 2 is a diagram showing a configuration of a video decoding device (hereinafter simply referred to as “decoding device 200”) that realizes predictive decoding according to the present invention. The components of these devices are connected to each other via a bus so that signals can be input and output. Moreover, in both figures, the same code | symbol is attached | subjected to the component which has the same function.

  First, the configuration of the encoding device will be described with reference to FIG. The encoding apparatus 100 divides the input frame 201 into 16 × 16 pixel encoding target blocks (prediction target blocks) 202 by the encoding block dividing unit 251. The apparatus is a device that predicts and encodes a plurality of encoding target blocks (prediction target blocks) 202 obtained by the division in the raster scan order.

  Here, the prediction target block is a block to be a next prediction target among blocks obtained by dividing the input frame 201 into blocks. The encoding target block is a block to be encoded next among blocks obtained by dividing the input frame 201 into blocks. In the present embodiment, the prediction target block and the encoding target block match in order to make the prediction processing target block and the encoding processing target block the same size. Moreover, a prediction block is a block produced | generated by the prediction process from the pixel decoded in the past.

  The encoding target block (prediction target block) 202 is first input to the decoded low-resolution block generation unit 252. The decoded low-resolution block generation unit 252 generates a decoded low-resolution block 110 and low-resolution block information 102 that have fewer pixels than the prediction target block 202. The prediction target block 202 is input to the motion estimation unit 211 in the order of raster scanning. The motion estimation unit 211 performs block matching between the input prediction target block 202 and a plurality of candidate prediction blocks 203 extracted from past decoded frames stored in the frame memory 153, A low-resolution block prediction motion vector 104 is detected.

  Here, block matching will be described with reference to FIG. Block matching is a technique for detecting a motion vector of a prediction target block. In block matching, the motion estimation unit 211 sets a search range 522 on the decoded frame 520 decoded in the past, and searches the search range. Next, the position of the block 105 at which the difference evaluation value from the prediction target block 202 is minimum is detected. The amount of movement from the virtual block 521 that is spatially the same position as the prediction target block 202 on the decoded frame to the detected block 105 is defined as a motion vector 104. In addition, as the difference evaluation value, a value obtained by adding a difference absolute value sum of corresponding pixels between the detection target block and the encoding target block, a converted value of a motion vector code amount, or the like Can be applied.

  The motion compensation unit 133 generates a low-resolution prediction block 106 corresponding to the low-resolution block prediction motion vector 104. Specifically, based on the low-resolution block prediction motion vector 104, a pixel corresponding to a block of 8 × 8 pixels is acquired from a past decoded frame 520 stored in the frame memory 153, and this is reduced. The resolution prediction block 106 is assumed. FIG. 4 shows the correspondence between the block 10 having a 16 × 16 pixel size and the low resolution block 11 having an 8 × 8 pixel size. As shown in FIG. 4, the low-resolution block 11 having an 8 × 8 pixel size is a block obtained by extracting pixels from the block 10 having a 16 × 16 pixel size at intervals of one line in the vertical and horizontal directions.

  Here, generating a low-resolution block by pixel extraction has an effect of improving the prediction performance of the small block prediction process described later. That is, since each pixel of the low-resolution block holds a high-frequency component, the encoding device 100 can easily capture fine features when performing motion detection. As a result, the prediction performance is improved.

  Also for the encoding target block (prediction target block) 202, the reduction processing unit 134 performs the low resolution encoding target block (low resolution) with a smaller number of pixels than the encoding target block 202 in accordance with the rules shown in FIG. Prediction target block) 205. The subtracting unit 213 subtracts each pixel in the low-resolution encoding target block (low-resolution prediction target block) 205 and the low-resolution prediction block 106 to generate a low-resolution error block 206.

  The conversion unit 214 divides the low-resolution error block 206 into four 4 × 4 blocks, and then performs 4 × 4 discrete cosine transform (DCT) on each 4 × 4 block. Then, transform coefficients 207 for 16 low-resolution error blocks are generated. The quantization unit 215 quantizes the transform coefficient 207 to generate the quantized transform coefficient 107 of the low resolution error block. The integrating unit 216 collects the motion vector 104 and the quantized transform coefficient 107 of the low resolution error block, and outputs the resultant to the entropy encoding unit 255 as the low resolution block information 102.

  The inverse quantization unit 136 inversely quantizes the quantized transform coefficient 107 to generate a decoded quantized transform coefficient 108 of the low resolution error block. The inverse transform unit 137 performs inverse DCT transform on the decoded quantized transform coefficient 108 to generate a decoded low resolution error block 109. The adder 217 adds the pixels in the low resolution prediction block 106 and the decoded low resolution error block 109 to generate a decoded low resolution block 110.

  The enlarged block generation unit 154 expands the decoded low resolution block 110 to the enlarged block 127 shown in FIG. That is, the enlarged block generation unit 154 performs boundary decoding on the decoded pixels of the block adjacent to the upper side and the left side of the block position currently being predicted from the decoded pixels of the current frame stored in the frame memory 153. Obtained as the pixel group 123 and arranged around the decoded low-resolution block 110 as shown in FIG.

  Subsequently, the enlarged block generation unit 154 calculates the interpolation value of the pixel indicated by the black square by the interpolation calculation process shown in FIG. With this process, the decoded low-resolution block 110 having an 8 × 8 block size is expanded to an expanded block 127 having the same block size (16 × 16) as a prediction block generated by a prediction process described later. Note that when the prediction target block is the end block of the current frame, there is no adjacent block, and thus the enlarged block generation unit 154 cannot acquire the boundary decoded pixel group 123. In this case, the boundary pixel of the decoded low resolution block 110 is used instead.

  The block division unit 162 divides the enlarged block 127 into 16 4 × 4 small blocks 112a as shown in a block 127a in FIG. Further, the block dividing unit 162 divides the enlarged block 127 into 64 2 × 2 small blocks 112b as shown in a block 127b in FIG. Each of these divided small blocks is composed of one 4 × 4 small block 112a and four 2 × 2 small blocks 112b that are spatially at the same position as a set, and the small block prediction is sequentially performed. Input to the unit 152.

  The small block prediction unit 152 generates a prediction small block 116 of the small block 112 using the small block motion estimation unit 163 that detects the motion vector 115 of the small block 112 by block matching, and the detected small block motion vector 115. And a small block motion compensation unit 164. In the present embodiment, the small block prediction unit 152 generates a predicted small block 116b of each 2 × 2 small block 112b using these functions.

More specifically,
(1) The small block motion estimation unit 163 performs a motion vector detection process of a 4 × 4 small block. Hereinafter, among the 16 4 × 4 small blocks 112a, the function of the small block predicting unit 152 will be described by focusing on the upper left block 112a1 shown in FIG. 9, but the other 15 4 × 4 small blocks are described. Are processed according to the same procedure except that the spatial block positions are different. FIG. 9 is a diagram for explaining a detection method of the motion vector 115a1 of the 4 × 4 small block 112a1 located at the upper left corner of the 16 divided 4 × 4 small blocks. When the 4 × 4 small block 112a is input to the small block motion estimation unit 163, the small block motion estimation unit 163 sets the search range 181a1 on the previously decoded frame 520 (stored in the frame memory 153). Subsequently, the search range is searched according to a given rule, and the 4 × 4 small block 116a1 having the smallest difference evaluation value (sum of the absolute difference values of pixels in the small block) with the 4 × 4 small block 112a1 is obtained. The position is detected, and this is set as a 4 × 4 prediction small block 116a1. Then, the small block motion estimator 163 is between the 4 × 4 block 117a1 on the decoded frame 520 that is spatially the same position as the 4 × 4 small block 112a1 and the detected 4 × 4 predicted small block 116a1. The movement amount is set as a motion vector 115a1 of the 4 × 4 small block 112a1.

  The search range 181a1 has a predetermined size and is set at a position centered on the encoded low-resolution block prediction motion vector 104 used for generating the low-resolution prediction block. This is because the motion vector of the detected 4 × 4 small block is likely to be close to the low-resolution block prediction motion vector 104, and the effect of improving the search accuracy and the search range are reduced. There is an effect that can be set. Accordingly, the center point of the search range of 16 4 × 4 small blocks belonging to the same encoding target block is similarly set to the position indicated by the low resolution block prediction motion vector 104.

  (2) The small block motion estimation unit 163 detects the motion vectors of four 2 × 2 small blocks belonging to the same set as the 4 × 4 small block 112a1 from which the motion vector is detected in (1). In the following, description will be given focusing on the block 112b1 at the upper left end shown in FIG. 10 among the four 2 × 2 small blocks. FIG. 10 is a diagram for explaining a detection method of the motion vector 115b1 of the 2 × 2 small block 112b1 located at the upper left corner of the 64 divided 2 × 2 small blocks. When the 2 × 2 small block 112b1 is input to the small block motion estimation unit 163, the small block motion estimation unit 163 sets the search range 181b1 on the previously decoded frame 520 (stored in the frame memory 153). Subsequently, the search range is searched according to a predetermined rule, and the position of the 2 × 2 small block 116b1 at which the difference evaluation value from the 2 × 2 small block 112b1 (the sum of absolute difference values of pixels in the small block) is minimized. Is detected and set as a 2 × 2 prediction small block 116b1. Then, the small block motion estimator 163 includes a 2 × 2 block 117b1 on the decoded frame 520 that is spatially located at the same position as the 2 × 2 predicted small block 112b1, and the detected 2 × 2 small block 116b1. The movement amount is set as a motion vector 115b1 of the 2 × 2 small block 112b1. Further, with respect to the remaining three 2 × 2 small blocks, motion vectors of the 2 × 2 small blocks are detected in the same manner.

  The search range 181b1 has a predetermined size and is set at a position centered on the motion vector 115a1 of the 4 × 4 small block 112a1 belonging to the same set as the 2 × 2 small block 112b1. This is because the motion vectors of the detected 2 × 2 small blocks are spatially at the same position, that is, the motion vectors are likely to be close to the motion vectors of the 4 × 4 small blocks belonging to the same set. Therefore, the search range can be set small. Therefore, the small block motion estimation unit 163 similarly uses the motion vectors 115a1 of the 4 × 4 small blocks spatially located at the center point of the search range of four 2 × 2 small blocks belonging to the same set. Set to the indicated position.

  (3) The detected four 2 × 2 small block motion vectors 115 b are input to the small block motion compensation unit 164. The small block motion compensation unit 164 obtains a predicted small block 116b corresponding to the motion vector 115b of the 2 × 2 small block from the decoded frame 520 (stored in the frame memory 153). Such processing of the small block predicting unit 152 is executed using only data that can be generated from the encoded data by the decoding apparatus 200, and therefore, the decoding apparatus 200 can obtain the same processing result. That is, the decoding apparatus 200 can detect the same motion vector as that of the encoding apparatus 100. Therefore, it is not necessary to encode the motion vectors of each 2 × 2 small block.

  The block integration unit 155 integrates each of the 64 2 × 2 prediction small blocks by arranging them in the original positions before the division based on the division rule in the block division unit 162, thereby integrating 16 × 16 pixel. A prediction block 118 is generated. The subtraction unit 253 generates a difference block 208 by calculating a difference value of each pixel between the encoding target block 202 and the prediction block 118.

  The error block encoder 254 encodes the error block 208 to generate error block information 119 and outputs the error block information 119 to the entropy encoder 255. As shown in FIG. 1, the error block encoding unit 254 includes a transform unit 218 and a quantization unit 219 as its components. The transform unit 218 performs 4 × 4 discrete cosine transform (DCT) on the error block 208 in units of 4 × 4, and generates 16 transform coefficients 209 for each of the 16 4 × 4 blocks. The quantization unit 219 quantizes each transform coefficient 209 to generate error block information (for example, quantized transform coefficient) 119.

  The error block decoding unit 156 generates the decoded error block 121 by decoding the error block information 119. As illustrated in FIG. 1, the error block decoding unit 156 includes the inverse quantization unit 140 and the inverse transform unit 141 as its components. The inverse quantization unit 140 generates the decoded quantized transform coefficient 120 by inversely quantizing the 16 quantized transform coefficients 119 for each of the 16 4 × 4 blocks. The inverse transform unit 141 generates a decoded error block 121 by performing inverse DCT transform on the decoded quantization transform coefficient 120 of each 4 × 4 block.

  The adder 157 adds the pixels in the prediction block 118 and the decoding error block 121 to generate a decoding block 122. The generated decoded block 122 is stored in the frame memory 153, and as a result, the decoded image is updated. The entropy encoding unit 255 multiplexes the low resolution block information 102 and the error block information 119 into the encoded data 101 by entropy encoding.

  Next, the operation of the encoding apparatus 100 will be described. In addition, each step constituting the moving picture coding method according to the present invention will be described. FIG. 11 is a flowchart for explaining an encoding process of a target block for predictive encoding. In the block encoding process of FIG. 11, reference numerals of steps having the same contents and processing results as the block decoding process described later are shown in parentheses in the figure.

  First, when the input frame 201 is divided by the coding block division unit 251, the coding apparatus 100 obtains a coding target block 202. In S1, the decoded low-resolution block generation unit 252 performs the low-resolution block encoding / decoding process illustrated in FIG.

  In S101 of FIG. 12, the next encoding target block (prediction target block) is input. In S102, the motion estimator 211 detects a low-resolution block prediction motion vector from the prediction target block and a previously decoded frame stored in the frame memory. The motion vector detection method is already described in the description of the configuration of the decoded low-resolution block generation unit 252 and is therefore omitted.

  In S103, the motion compensation unit 133 acquires a low-resolution prediction block corresponding to the low-resolution block prediction motion vector from the decoded frame of the frame memory. In S104, the reduction processing unit 134 converts the prediction target block into the low resolution prediction target block based on the above-described conversion rule described with reference to FIG. In S105, the subtraction unit 213 subtracts each pixel between the low-resolution prediction target block and the low-resolution prediction block to generate a low-resolution error block.

  In S106, after the conversion unit 214 DCT-transforms the low resolution error block, the quantization unit 215 quantizes the DCT coefficient. As a result, a quantized transform coefficient of the low resolution error block is generated. Note that the quantization accuracy in S106 is set higher than the quantization accuracy in S8 of FIG. This is due to the effect that the 2 × 2 small prediction block detection accuracy in S3 and S4 can be increased by increasing the image quality of the low-resolution block by the encoding apparatus 100.

  In S107 of FIG. 12, the inverse quantization unit 136 and the inverse transform unit 137 decode (inverse quantization / inverse transform) the quantized transform coefficient of the low resolution error block. As a result, a decoded low resolution error block is generated. In S108, the addition unit 217 adds each pixel between the low-resolution prediction block and the decoded low-resolution error block to generate a decoded low-resolution block. Then, the integration unit 216 integrates the motion vector and the quantized transform coefficient of the low resolution error block to generate low resolution block information (S109).

  Returning to FIG. 11, in S <b> 2, the enlarged block generation unit 154 enlarges the decoded low-resolution block to an enlarged block 127 having the same size (16 × 16 pixels) as the prediction block, as shown in FIGS. 5 and 6. To do. The method for generating the enlarged block is the same as that described with reference to FIG. 5 and FIG.

  In S3, as shown in FIGS. 7 and 8, the block dividing unit 162 divides the enlarged block into 4 × 4 small blocks and 2 × 2 small blocks. Each of the divided small blocks is input to the small block motion estimation unit 163 as one set of 4 × 4 small blocks and four 2 × 2 small blocks spatially located at the same position. The

  In S4, the small block motion estimator 163 detects the motion vectors of four 2 × 2 small blocks in units of the set. The small block motion estimation process is executed for one 4 × 4 small block and four 2 × 2 small blocks. In this order, processing for 4 × 4 small blocks is executed first, and then processing for four 2 × 2 small blocks is executed.

  Here, the small block motion estimation process will be described with reference to FIG. First, in S41, a small block is input, and the evaluation value Emin is set as the maximum integer value (initial value) (S42). Next, in S43, the size of the small block is determined. Thereafter, processing S441 or S442 according to the determination result is executed. That is, the size of the search range and the center position (search center point) of the search range are set on the past decoded frame 520 (stored in the frame memory 153).

  Subsequently, prediction small block candidates within the search range of the decoded frame 520 are searched in a spiral order from the search center point, and as a result, the prediction small block having the minimum evaluation value Emin is specified. Specifically, the small block motion estimation unit 163 obtains the next candidate predicted small block 113 from the decoded frame in the frame memory 153 (S45), and calculates an evaluation value E (S46). The evaluation value E is the sum of absolute values of pixel difference values between the small block and the predicted small block.

  In S47, it is determined whether or not the evaluation value E is smaller than the evaluation value Emin. When the evaluation value E is smaller than the evaluation value Emin (S47; YES), the evaluation value Emin is updated to the evaluation value E, and the motion vector of the small block is changed from the position of the small block to the current candidate block of the predicted small block. The amount of movement to the position is updated (S48). On the other hand, when the evaluation value E is equal to or higher than the evaluation value Emin (S47; NO), the process of S48 is omitted and the process proceeds to S49.

  In S49, it is determined whether or not the search for all candidate predicted small blocks within the search range has been completed. Thereafter, a series of processes of S45 to S49 are repeatedly executed until it is determined in S49 that the search for all candidate small predicted blocks within the search range has been completed. Then, when it is determined that the search is finished, the small block motion estimation process is finished.

  Returning to FIG. 11, in S5, the small block motion compensation unit 164 converts the 2 × 2 predicted small block 116 corresponding to the motion vectors of the four 2 × 2 small blocks detected in S3 and S4 into the frame memory 153. Respectively obtained from the decoded frames. Each process of S3-S5 is performed about 16 sets of small blocks.

  In S6, based on the division rule used in S3, the block integration unit 155 arranges the 64 2 × 2 prediction small blocks at the position before division, thereby integrating them into a prediction block of 16 × 16 pixels. In S7, the subtraction unit 253 subtracts each pixel between the encoding target block and the prediction block to generate an error block. In S8, the error block encoder 254 encodes (transforms / quantizes) the error block to generate error block information (for example, quantized transform coefficients).

  In S9, the error block decoding unit 156 decodes (inverse quantization / inverse transform) the error block information to generate a decoded error block. In S10, the adding unit 157 adds each pixel between the prediction block and the decoding error block to generate a decoding block. When the generated decoded block is stored in the current decoded frame in the frame memory 153 (S11), the entropy encoding unit 255 entropy encodes the low-resolution block information and the error block information (S12), The encoding process ends.

  Next, the configuration of the decoding device 200 shown in FIG. 2 will be described. The decoding device 200 is a device that executes prediction processing in units of blocks and decoding processing of the encoded data 101 and reproduces the decoding blocks 122 in raster scan order in units of blocks. In this embodiment, the prediction block generated by the prediction process and the decoded block have the same size, and both are 16 × 16 pixels.

  When receiving the input of the encoded data 101 necessary for generating a decoding block to be decoded, the entropy decoding unit 150 converts the encoded data 101 into low-resolution block information 102 and error block information 119 by entropy decoding. And to separate. The low-resolution block decoding unit 151 decodes the low-resolution block information 102 to generate a decoded low-resolution block 110 having 8 × 8 pixels that has a smaller number of pixels than the prediction block.

  The separation unit 131 separates the low-resolution block information 102 into low-resolution block prediction motion vectors 104 and low-resolution error block information (for example, quantized transform coefficients of the low-resolution error block) 107. The motion compensation unit 133 acquires the low-resolution prediction block 106 corresponding to the low-resolution block prediction motion vector 104 from the decoded frame 520 stored in the frame memory 153.

  The inverse quantization unit 136 inversely quantizes the quantized transform coefficient 107 of the low resolution error block to generate the decoded quantized transform coefficient 108 of the low resolution error block. The inverse transform unit 137 generates a decoded low-resolution error block 109 by performing inverse DCT transform on the decoded quantized transform coefficient 108. The adder 135 calculates an addition value of each pixel in the low resolution prediction block 106 and the decoded low resolution error block 109, and generates a decoded low resolution block 110.

  Note that the functions and processing results of the expanded block generation unit 154, the block division unit 162, the small block prediction unit 152, and the block integration unit 155 are the same as those described in the encoding apparatus 100. Therefore, the explanation and illustration for explanation are omitted.

  Further, the error block decoding unit 156 generates error decoding block 121 by decoding error block information (for example, quantized transform coefficient) 119. The error block decoding unit 156 includes an inverse quantization unit 140 and an inverse transform unit 141. The inverse quantization unit 140 inversely quantizes the quantized transform coefficient 119 to generate a decoded quantized transform coefficient 120. The inverse transform unit 141 performs inverse DCT transform on the decoded quantized transform coefficient 120 to generate a decoded error block 121.

  The adder 157 calculates, for each pixel of the prediction block 118, an addition value between a pixel at a position spatially coincident with the decoding error block 121 and the pixel, and thereby generates a decoding block 122. The decoding block 122 is stored in the frame memory 153 and the decoded image is updated.

  Subsequently, the operation of the decoding device 200 will be described, and each step constituting the moving image decoding method according to the present invention will be described. In the block decoding process of FIG. 14, the reference numerals of steps having the same contents and processing results as the block encoding process shown in FIG. 11 are shown in parentheses in the figure.

  When the decoding process for the next block is started, first, at T1, encoded data for one block necessary for decoding the block is input to the entropy decoding unit 150 (T1). In T2, the entropy decoding unit 150 performs entropy decoding on the input encoded data to obtain low-resolution block information and error block information. Among these pieces of information, the low resolution block information is input to the low resolution block decoding unit 151 by the low resolution block decoding process 404 and decoded into the decoded low resolution block.

  Here, the low-resolution block decoding process in T3 will be described with reference to FIG. At T31, the low resolution block information is input to the separation unit 131. In T32, the separation unit 131 separates the low resolution block information into the motion vector and the low resolution error block information. In T33, the motion compensation unit 133 acquires the low-resolution block prediction block 106 corresponding to the low-resolution block prediction motion vector 104 from the previously decoded frame 520 (stored in the frame memory 153).

  At T34, the inverse quantization unit 136 inversely quantizes the quantized transform coefficient 107 as the low-resolution error block information, and generates a decoded quantized transform coefficient 108. Further, the inverse transform unit 137 performs inverse DCT transform on the decoded quantized transform coefficient 108 to generate a decoded low resolution error block 109. At T35, the adding unit 135 adds the pixels of the low-resolution prediction block 106 and the decoded low-resolution error block 109 to generate a decoded low-resolution block 110.

  Returning to FIG. 14, as illustrated in FIGS. 5 and 6, the expanded block generation unit 154 expands the decoded low-resolution block 110 to the expanded block 127 having the same size (16 × 16 pixels) as the prediction block (T4). ). The method for generating the enlarged block is the same as that described with reference to FIG. 5 and FIG.

  In T5 to T7, the block dividing unit 162 and the small block predicting unit 152 divide the expanded block into 64 2 × 2 small blocks, and generate 64 2 × 2 predicted small blocks. Note that the processes at T5 and T6 in FIG. 14 are the same as the processes at S3 and S4 shown in FIG. 11, and thus detailed description thereof is omitted. Further, the process of T7 is the same as the process of S5 in FIG.

  In T8, the block integration unit 155 arranges the 64 2 × 2 prediction small blocks generated in T6 and T7 at the original positions before the division based on the division rule in T5. As a result, the small prediction block is integrated into a 16 × 16 pixel prediction block.

  In T9, the error block decoding unit 156 decodes (inverse quantization / inverse transform) error block information (for example, quantized transform coefficients) to generate a decoded error block. In T10, the addition unit 157 adds each pixel of the prediction block and the decoding error block to generate a decoding block. At T11, the decoded block is stored in the frame memory 153, and as a result, the decoded image is updated.

  FIG. 16A shows the physical structure of a flexible disk in which a moving image encoding program for realizing the image encoding process and a moving image decoding program for realizing the image decoding process in the present embodiment are stored. The format is illustrated. FIG. 16B shows a front appearance and a cross-sectional structure of the flexible disk. The flexible disk FD is built in the case F, and a plurality of tracks Tr are formed concentrically from the outer periphery toward the inner periphery on the surface of the disk. Each of these tracks is divided into 16 sectors Se in the angular direction. In the area allocated on the flexible disk FD, the above-mentioned program is recorded as data.

  When the program is recorded on the flexible disk FD, as shown in FIG. 16C, the computer system Cs passes the flexible disk drive FDD. When the encoding device 100 or the decoding device 200 is built in the computer system Cs by the program in the flexible disk FD, the flexible disk drive FDD reads the program from the flexible disk FD and transfers it to the computer system Cs.

  FIG. 17 is a diagram showing a configuration of the moving image encoding program 261 according to the present invention. The moving image encoding program 261 is recorded on the recording medium 260. The recording medium 260 is, for example, the flexible disk described above, but may be an optical disk, an IC card, a CD-ROM, a DVD, or a semiconductor memory. The computer system Cs is not limited to a PC (Personal Computer), and includes a DVD player, a set-top box, a mobile phone, and the like that have a CPU and perform software information processing and control.

  Next, the configuration of the moving image encoding program and the moving image decoding program according to the present invention will be described. As shown in FIG. 17, the moving image encoding program 261 is stored in a program storage area 260a formed in the recording medium 260. The moving image encoding program 261 includes a main module 261a that comprehensively controls moving image encoding processing, an encoding target block division module 261b, a decoded low-resolution block generation module 261c, and a low-resolution block enlargement module 216d. And an expanded block division module 261e, a small block prediction module 261f, and a block integration module 261g. The functions realized by executing these modules are the encoding block dividing unit 251, the decoded low-resolution block generating unit 252, the enlarged block generating unit 154, the block dividing unit 162, the small unit, and the like of the moving image encoding apparatus 100 described above. The functions of the block prediction unit 152 and the block integration unit 155 are the same.

  The moving picture decoding program 161 is stored in a program storage area 160a formed on the recording medium 160, as shown in FIG. The moving picture decoding program 161 includes a main module 161a that comprehensively controls moving picture decoding processing, an entropy decoding module 161b, a low resolution block decoding module 161c, a low resolution block enlargement module 161d, and an enlargement block division module. 161e, a small block prediction module 161f, and a block integration module 161g are included as structural units. The functions realized by executing these modules are the entropy decoding unit 150, the low-resolution block decoding unit 151, the expanded block generation unit 154, the block division unit 162, and the small block prediction unit 152 of the video decoding device 200 described above. The functions of the block integration unit 155 are the same.

  Note that a part or all of the moving picture encoding program 261 may be transmitted via a transmission medium such as a communication line, received by another device, and recorded (including installation). Similarly, a part or all of the moving picture decoding program 161 may be transmitted via a transmission medium such as a communication line and received and recorded (including installation) by another device.

  The aspect described in the above embodiment is a preferred example of the moving picture encoding / decoding technique according to the present invention, and the present invention is not limited to this aspect.

(First variant)
This aspect relates to a small block motion vector detection method. In the above embodiment, the search range when detecting a motion vector of a small block is ± 32 for a 4 × 4 small block and ± 1 for a 2 × 2 small block, but may be a different search range. Further, the accuracy of the motion vector component of the small block can be not only an integer value but also a real value. The real value includes ½ pixel, ¼ pixel, and the like. As a prediction pixel calculation method when the vector component is a real value, for example, H.264 is used. H.264 (“Text of ISO / IEC 14496-10 Advanced Video Coding 3rd Edition”, September 2004) can be applied.

  In the above embodiment, the center point of the search range when detecting the motion vector of the small block is the low resolution block prediction motion vector in the 4 × 4 small block and spatially the same position in the 2 × 2 small block. The motion vector is a 4 × 4 small block. However, the present invention is not limited to this as long as the value can be uniquely determined on the encoding side and the decoding side. For example, a prediction vector selected from motion vectors of adjacent small blocks by a method such as a zero vector or median prediction can be set as the center point of the search range.

  Furthermore, although the evaluation value when detecting the motion vector of the small block is the sum of the absolute values of the pixel difference values, the evaluation value is not limited to this, and any evaluation method that can be uniquely determined on the encoding side and the decoding side may be used. . For example, it is possible to transform the pixel difference value into a square sum or a weighted absolute value sum (square sum) of the pixel difference value according to the pixel position. Although the decoding frame to be predicted is limited to one frame, a prediction small block may be generated from a decoding frame that is different for each small block with respect to a plurality of decoded frames. At this time, if the same evaluation value is used on the encoding side and the decoding side, such as the sum of absolute values of pixel difference values as the evaluation value, it is not necessary to encode the selection information of the decoded frame, which is efficient.

(Second modification)
This aspect relates to a prediction method. Although the prediction method has been limited to prediction in the temporal direction, the same method can be applied to prediction in the spatial direction using a decoded image that has already been decoded in the same frame as the current block or decoding block. can do. That is, the above-described decoded low-resolution block can be expanded using intra prediction. More specifically, the decoding apparatus 200 divides the low-resolution block obtained by re-dividing the decoded low-resolution block, and uses adjacent decoded pixels in the same frame. H.264 ("Text of ISO / IEC 14496-10 Advanced Video Coding 3rd Edition", September 2004) The case where a block of a signal is predicted is also included in the present invention. In this case, since it is not necessary to explicitly transmit the prediction mode of each reduced divided block, there is an excellent effect that the target block can be predicted with high accuracy with a small amount of auxiliary information. Moreover, you may perform highly accurate prediction combining the prediction of the time direction mentioned above, and the prediction of a spatial direction. For example, it is also effective to execute an expansion process of a decoded low-resolution block by intra prediction in the expanded block generation unit 154 and divide the expanded block into small blocks to realize the processing of the small block prediction unit 152. is there.

(Third Modification)
Such an aspect relates to encoding of error blocks. The process of encoding the error block between the prediction block and the encoding target block is not essential. That is, the error block encoding unit 254 (FIG. 1) and the error block decoding unit 156 (FIGS. 1 and 2) can be omitted. Also, the presence or absence of encoding may be determined for each encoding target block or decoding block. The specific encoding / decoding method of the error block and the low resolution error block is not limited to the above-described method. Furthermore, although the encoding target block and the prediction target block have the same block size in the above embodiment, the prediction method of the present invention can be realized even with different sizes. Further, the encoding of the error signal is not limited to the block unit, and may be performed on a screen basis. However, in this case, in order to prepare decoded pixels at the block boundary used for the enlargement process of the low resolution block, before executing the enlargement process of the low resolution block, the generation process of the adjacent prediction block and the prediction error signal It is necessary to execute a decryption process.

(Fourth modification)
Such an aspect relates to block size. In the above embodiment, one motion vector for predicting low-resolution blocks is set for each of the encoding target block and the decoding block. However, as shown in FIG. 19, the prediction method according to the present invention can also be applied to a method of dividing a block to be encoded into a plurality of blocks and performing block matching in units of divided blocks. In this case, the size of the small block is determined for each block to be encoded, and the size information of the divided blocks and a plurality of motion vectors (104a, 104b, 104c, 104d) are encoded.

  Furthermore, in the example of FIG. 19, the center point of the search range when detecting the motion vector of the 4 × 4 small block is indicated by the motion vector at the same spatial position from the four motion vectors for low resolution block prediction. Can be set to position.

  In FIG. 19, the division method is 8 × 8 units, but is not limited to this. H.264 (“Text of ISO / IEC 14496-10 Advanced Video Coding 3rd Edition”, September 2004) can be used adaptively. Also, the encoding target block size and the decoding block size are not limited to 16 × 16. That is, even if the block sizes are different such as 32 × 32, 8 × 8, and 4 × 4, the present invention can be applied by changing the size of the small block according to the size.

  The size of the small block in the small block prediction unit 152 (S3 and S4 in FIG. 6) is not limited to 4 × 4 and 2 × 2. For example, the small block size may be enlarged by causing adjacent small blocks to overlap each other like a 5 × 5 small block 112c or a 3 × 3 small block 112d shown in FIG. A small block size tends to decrease the prediction performance when there is no feature in the block. Therefore, by adopting such a mode, it is expected that the local prediction performance is improved. It is also effective to adaptively switch the small block size between 5 × 5 and 4 × 4, and 3 × 3 and 2 × 2. Similarly, the block size at the time of executing transformation / quantization is not limited. That is, the encoding device 100 and the decoding device 200 may transform and quantize the error block, for example, in units of 8 × 8 blocks or 16 × 16 blocks.

(5th modification)
This aspect relates to a method for enlarging a decoded low resolution block. In the above embodiment, the enlarged block generation unit 154 acquires the boundary decoded pixel group 123 referred to for enlarging the decoded low-resolution block from the decoded pixels of the blocks adjacent to the upper side and the left side of the prediction block. did. However, as shown in FIG. 21, the boundary decoded pixel group 123 can be acquired from the decoded pixels of the decoded low-resolution block of the adjacent block to generate an enlarged block.

An encoding device 300 and a decoding device 400 for realizing the above-described modification are shown in FIGS. 22 and 23, respectively. The encoding device 300 and the decoding device 400 are different from the encoding device 100 and the decoding device 200 in the following points.
(I) The point where the decoded low-resolution block 110 is once stored in the frame memory 153 (ii) The decoded low-resolution block spatially corresponding to the adjacent block is integrated into the decoded low-resolution block as a boundary decoded pixel (Iii) A point that is input to the enlarged block generation unit 154 in units of a low resolution block 124 (see FIG. 21) including boundary decoding pixels. By configuring the encoding device 300 and the decoding device 400 in this way, It becomes possible to change the boundary decoding pixel.

  Note that the features of each device shown in FIGS. 1 and 2 and the features of each device shown in FIGS. 22 and 23 may be adaptively combined according to conditions. For example, each device uses the decoded pixel of the decoded low-resolution block as the boundary decoded pixel when the motion vector for the low-resolution block of the adjacent block is the same as the current prediction target block. If they are different, the decoded pixel of the decoded frame is used as the boundary decoded pixel. Further, in an encoding method in which a plurality of prediction methods are adaptively switched in units of blocks, boundary decoding pixels may be changed on the condition that the prediction method of a block adjacent to the current prediction target block is used as a condition.

  Furthermore, an enlarged block may be generated only from pixels in the low-resolution prediction block, as in the processing at the frame end, without using boundary decoding pixels. In the above embodiment, the boundary decoded pixels are extracted from the decoded image of the current frame. However, the decoded pixels of the past decoded frames at the same spatial position can also be used. The boundary decoding pixel is a pixel adjacent to the prediction target block. However, the boundary decoding pixel is not limited to this as long as it is a decoded pixel, and is generated from a decoding pixel separated from the block boundary or a plurality of decoding pixels. Pixels that have been used may also be used.

(Sixth modification)
This aspect relates to a reduction method. In the above embodiment, as shown in FIG. 4, the block thinning process employs a method of thinning out at intervals of one line in the vertical and horizontal directions, but is not limited to this method. For example, the number of pixels of the low-resolution block may be determined in advance between the encoding device 100 and the decoding device 200, and can be an arbitrary shape having a smaller number of pixels than the decoding block (encoding target block). . For example, a block in which each pixel of the low resolution prediction block 106 shown in FIG. 4 is shifted by one pixel in the vertical and horizontal directions, or a block in which the number of pixels is halved, such as the low resolution prediction block 1061 in FIG. The present invention is applicable.

  In the above embodiment, the block reduction is realized only by the pixel thinning process. However, the thinning process may be performed after the low-pass filter process is performed on the pixels of the block. However, in order to increase the accuracy of the motion search, it is necessary to perform the same low-pass filter processing on the candidate prediction block 113 that is an input to the small block motion estimation unit 163.

(Seventh modification)
Such an aspect relates to low resolution block prediction. In each of the above embodiments, each device directly acquires the low-resolution prediction block from the frame memory 153. However, the data 15 necessary for generating the low-resolution prediction block 106 is stored in the frame memory 153. It may be obtained from the past decoded frame 520. Thereafter, the motion compensation unit 133 may generate the prediction block 105 having the size of 16 × 16 pixels illustrated in FIG. 3 and convert the prediction block 105 into the low-resolution prediction block 106 having the size of 8 × 8 pixels. .

  In each of the above embodiments, each device uses a 16 × 16 pixel prediction target block as an input to the motion estimation unit 211. However, low-resolution block prediction is not limited to the method described in this embodiment. For example, a low-resolution block obtained by applying the conversion rule of FIG. 4 to the prediction target block may be used as the input. In this case, the low-resolution block prediction motion vector is selected so that the prediction error between the low-resolution prediction block and the low-resolution encoding target block (low-resolution prediction target block) becomes small. There is an effect that the code amount of the image error block is reduced.

(Eighth modification)
This aspect relates to encoding of multiple prediction modes. In the above-described embodiment, only the prediction method according to the present invention is used as the prediction method for the encoding target block (decoded block). However, one of the conventional block prediction method for encoding a plurality of motion vectors as described in the background art and the prediction method according to the invention is appropriately selected in units of encoding target blocks (decoding blocks). It may be a thing.

  Furthermore, in the above-described prediction method, all pixels in the 16 × 16 block are predicted by the prediction method according to the present invention. However, a 16 × 16 block may be divided and the prediction method of the present invention may be adaptively switched between the prediction method of the present invention and the inter-frame prediction method using the low-resolution block prediction motion vector for each divided block. For example, in a case where a 16 × 16 encoding target block (decoding block) is divided into four 8 × 8 blocks and adaptive prediction is performed in units of 8 × 8, an encoding apparatus 500 that enables the implementation is illustrated. 25. A similar decoding device 600 is shown in FIG. FIG. 27 shows a decoded low-resolution block and an enlarged block in this aspect.

  The encoding apparatus 500 in FIG. 25 is different from the encoding apparatus 100 shown in FIG. 1 in the following points. First, the enlarged block generating unit 1541 executes the decoding process of the decoded low-resolution block in units of 8 × 8. That is, the expanded block generation unit 1541 expands the decoded low-resolution block to an 8 × 8 expanded block in units of 4 × 4, and outputs it to the block dividing unit 162. As shown in FIG. 27, the enlarged block generation unit 1541 divides the 8 × 8 decoded low-resolution block 110 into four 4 × 4 low-resolution blocks 110a, 110b, 110c, and 110d. Then, the 4 × 4 block unit is integrated with the boundary decoded pixel groups 128 a to 128 d and expanded to an 8 × 8 expanded block 1271. The expanded 8 × 8 enlarged block is divided by the block dividing unit 165 into 4 × 4 small blocks and 2 × 2 small blocks as shown in blocks 1271a and 1271b. Then, it is input to the small block prediction unit 152.

  Note that since the decoded pixels are used for the boundary decoded pixel groups 128a to 128d, in this adaptive prediction method, it is necessary to execute a prediction error encoding process in units of 8 × 8. However, in the existing encoding method, since the prediction error encoding process is normally executed in units of 8 × 8, it is possible to cope with the adaptive prediction method only by switching the execution unit of the prediction process and the error encoding process. be able to.

  Second, the determination unit 256 selects a prediction method and a prediction block in units of 8 × 8 blocks. The determination unit 256 receives the prediction target block 202 and the 16 × 16 prediction block 105 read from the frame memory 153 by the motion compensation unit 1331 using the low-resolution block prediction motion vector 104. Thereafter, these blocks are each divided into four 8 × 8 encoding target blocks and a low-resolution prediction block selection 8 × 8 prediction block obtained by dividing the 16 × 16 prediction block into four.

  On the other hand, the 8 × 8 prediction block 118 generated by the small block prediction method of the present invention is input from the block integration unit 155 to the determination unit 256. The determination unit 256 calculates the pixel difference absolute value sum of the two types of prediction blocks with the encoding target block in units of 8 × 8, selects the smaller 8 × 8 prediction block, and has 16 × 16 pixels. Are integrated into the prediction block 126. The selection information 125 of the selected prediction method is output to the integration unit 2161.

  Third, the integration unit 2161 integrates the selection information 125 with the low resolution block information 1021. When the selection information 125 selects a prediction method using the low-resolution block prediction motion vector 104, information on the corresponding 4 × 4 block area among the decoded low-resolution blocks of 8 × 8 block size. Is not necessary. For this reason, in such a case, the quantization transform coefficients of the corresponding 4 × 4 low resolution error block are all zero, and are not integrated into the low resolution block information 1021.

  26 is different from the decoding device 200 shown in FIG. 1 in the following points. First, the enlarged block generating unit 1541 executes the decoding process of the decoded low-resolution block in units of 8 × 8. Since the method is the same as described with reference to FIG. 27, the details are omitted, but the decoding side also decodes the corresponding 8 × 8 prediction error block when the 8 × 8 prediction block is generated. To decode the 8 × 8 decoding block.

  Second, the switch 159 switches between two types of prediction methods according to the selection information 125 separated from the low resolution block information 1021. However, when the selection information 125 selects the prediction method using the motion vector 104, the corresponding 4 × 4 block region information among the 8 × 8 decoded low resolution blocks is the low resolution block information. 1021 is not included. That is, the quantization coefficients of the low resolution error block are all zero. For this reason, the corresponding 4 × 4 low-resolution block decoding process and the small block prediction process of the present invention are not executed for the region.

  Note that the block expansion method described in the fifth modification described above can also be applied in the eighth modification, and the block expansion method is adaptive according to the prediction method of adjacent blocks of the prediction target block. You can switch to

[Second Embodiment]
Embodiments according to the present invention will be described below with reference to the drawings.

[Device configuration]
FIG. 28 shows the configuration of a moving picture decoding apparatus (hereinafter referred to as “decoding apparatus”) that implements predictive decoding according to the present invention, and FIG. 29 shows a moving picture coding apparatus that implements predictive coding according to the present invention. (Hereinafter referred to as “encoding device”). It should be noted that the processing units with the same numbers in FIGS. 28 and 29 have the same functions.

[Configuration of Encoding Device]
29 divides the input frame 201 into 8 × 8 encoding target blocks 202 by the encoding block dividing unit 251 and encodes a plurality of encoding target blocks 202 obtained by the division in the raster scan order. It is a device to convert. Similarly to the first embodiment, in this embodiment, since the block size of the prediction target block to be predicted is the same as that of the encoding target block to be encoded, the prediction target block and the encoding target block match. Moreover, a prediction block is a block produced | generated by the prediction process from the pixel decoded in the past. Note that the block sizes of the prediction target block and the encoding target block may be different. The encoding target block 202 is first input to the decoded low resolution block generation unit 252.

  The decoded low-resolution block generation unit 252 is a processing unit that generates the “decoded low-resolution block 110 having a smaller number of pixels than the encoding target block (prediction target block) 202” and “low-resolution block information 102”. . Here, the operation of the decoded low-resolution block generation unit 252 will be described. The motion estimation unit 211 receives the encoding target block (prediction target block) 202 in the raster scan order, and the motion estimation unit 211 is held in the input encoding target block (prediction target block) 202 and the frame memory 153. Block matching is performed with a plurality of candidate prediction blocks 203, and one motion vector 104 is detected. The motion vector prediction / encoding unit 212 receives the detected motion vector 104 and calculates a difference motion vector 103 obtained by subtracting the value of each component of the predicted motion vector from the value of each component. The predicted motion vector is calculated by median prediction (each component) for the motion vectors of three adjacent blocks on the left, directly above, and upper right (upper left in the case of outside the frame) of the current block. The motion compensation unit 133 acquires the block 105 corresponding to the motion vector 104 from the decoded frame 520 that is already stored in the frame memory 153. The block 105 is converted into a low resolution prediction block 106 in which pixels are extracted every other vertical and horizontal lines, as shown in FIG. Further, the encoding target block (prediction target block) 202 is similarly reduced by the reduction processing unit 134 to a low resolution encoding target block (low resolution prediction target) having a smaller number of pixels than the encoding target block (prediction target block) 202. Block) 205. For each pixel of the low-resolution encoding target block (low-resolution prediction target block) 205, the subtraction unit 213 calculates a difference value between the pixel at a position spatially matching the low-resolution prediction block 106 and each pixel. The low resolution error block 206 is generated by calculation. The conversion unit 214 performs 4 × 4 discrete cosine transform (DCT) on the low resolution error block 206 to generate conversion coefficients 207 of 16 low resolution error blocks. The quantization unit 215 quantizes the transform coefficient 207 to generate a quantized transform coefficient 107 of a low resolution error block. The integrating unit 216 collects the difference motion vector 103 and the quantized transform coefficient 107 of the low resolution error block, and outputs them to the entropy encoding unit 255 as the low resolution block information 102. The inverse quantization unit 136 inversely quantizes the quantized transform coefficient 107 to generate a decoded quantized transform coefficient 108 of the low resolution error block. The inverse transform unit 137 performs inverse DCT transform on the decoded quantized transform coefficient 108 to generate a decoded low resolution error block 109. The adder 217 calculates, for each pixel of the low resolution prediction block 106, an addition value between the pixel at a position spatially coincident with the decoded low resolution error block 109 and each pixel, and the decoded low resolution block 110. Is generated.

  Next, the low-resolution block dividing unit 166 divides the decoded low-resolution block 110 into 2 × 2 low-resolution divided blocks 112 in the same manner as the block 106 in FIG. Each low-resolution divided block 112 is input to the small block prediction unit 158 (corresponding to 152 in FIGS. 1 and 2).

  The small block prediction unit 158 is a processing unit that generates a 4 × 4 prediction small block 116 for each low resolution divided block 112 using the decoded image stored in the frame memory 153. Since the process of the small block predicting unit 158 is performed using only the decoded image as an input, the same process can be performed in the decoding apparatus. Here, the operation of the small block prediction unit 158 will be described. A 4 × 4 block (referred to as a “candidate prediction small block”) 113 input to the small block prediction unit 158 is converted into a candidate reduced prediction small block 114 having a small number of pixels by the reduction processing unit 134, and is supplied to the expansion motion estimation unit 138. Is output. The expanded motion estimation unit 138 has a 2 × 2 low-resolution divided block 112 (112a to 112d) illustrated in FIG. 35 and four pixels spatially corresponding to the four pixels of the low-resolution divided block (candidate reduced prediction small block 114). And the motion vector 115 (115a to 115d) of the low resolution divided block is detected. The difference evaluation value at this time spatially corresponds to 4 pixels of the low resolution divided block among 16 pixels of “4 pixels of the low resolution divided block” and “4 × 4 block to be searched (candidate prediction small block 113)”. The sum of absolute differences is 4 pixels (candidate reduced prediction small block 114) ”(details will be described later using FIG. 32). Since a difference evaluation value using only such decoded images as input values can be calculated by the decoding device, the same motion vector as that of the encoding device can be detected by the decoding device. The predicted small block acquisition unit 139 acquires the 4 × 4 predicted small blocks 116 (116 a to 116 d) corresponding to the motion vectors 115 (115 a to 115 d) of the low resolution divided blocks from the decoded frame 520 of the frame memory 153.

  Next, the block integration unit 155 integrates the four prediction small blocks 116a to 116d as shown on the left side of FIG. 35, and generates an 8 × 8 prediction block 118 (FIG. 29). The subtraction unit 253 calculates, for each pixel of the encoding target block 202, a difference value between the pixel at a position spatially coincident with the prediction block 118 and each pixel, and generates an error block 208.

  The error block encoding unit 254 is a processing unit that encodes the error block 208 to generate error block information 119 and outputs the error block information 119 to the entropy encoding unit 255. Here, the operation of the error block encoding unit 254 will be described. The transform unit 218 performs 8 × 8 discrete cosine transform (DCT) on the error block 208 to generate 64 transform coefficients 209. The quantization unit 219 quantizes the transform coefficient 209 to generate error block information (quantized transform coefficient) 119.

  The error block decoding unit 156 is a processing unit that generates the decoded error block 121 by decoding the error block information 119. Here, the operation of the error block decoding unit 156 will be described. The inverse quantization unit 140 inversely quantizes the quantized transform coefficient 119 to generate a decoded quantized transform coefficient 120. The inverse transform unit 141 performs inverse DCT transform on the decoded quantized transform coefficient 120 to generate a decoded error block 121.

  Next, the addition unit 157 calculates the addition value of the pixel at the position spatially coincident with the decoding error block 121 and each pixel for each pixel of the prediction block 118 and generates a decoding block 122. The generated decoded block 122 is held in the current decoded frame 610 of the frame memory 153. Finally, the entropy encoding unit 255 performs entropy encoding on the low-resolution block information 102 and the error block information 119 and multiplexes the encoded data 101.

[Configuration of Decoding Device]
The decoding device in FIG. 28 is a device that decodes the encoded data 101 and reproduces the 8 × 8 decoding block 122 in raster scan order. The encoded data 101 corresponding to one 8 × 8 decoded block is first input to the entropy decoding unit 150. The entropy decoding unit 150 generates low resolution block information 102 and error block information 119 by entropy decoding.

  The low-resolution block decoding unit 151 is a processing unit that decodes the low-resolution block information 102 and generates a decoded low-resolution block 110 having a smaller number of pixels than the decoded block. Here, the operation of the low-resolution block decoding unit 151 will be described. The separation unit 131 separates the low-resolution block information 102 into the differential motion vector 103 and the low-resolution error block information (quantized transform coefficient of the low-resolution error block) 107. The motion vector prediction / decoding unit 132 calculates the motion vector 104 by adding the value of each component of the motion vector predictor to the value of each component of the differential motion vector 103. The predicted motion vector is calculated, for example, by median prediction (for each component) for the motion vectors of three adjacent blocks on the left, immediately above, and upper right (upper left in the case of outside the frame) of the current block. The motion compensation unit 133 acquires the block 105 corresponding to the motion vector 104 from the decoded frame 520 that is already stored in the frame memory 153. The block 105 is converted into a low resolution prediction block 106 in which pixels are extracted every other vertical and horizontal lines, as shown in FIG. The inverse quantization unit 136 inversely quantizes the quantized transform coefficient 107 of the low resolution error block, and generates a decoded quantized transform coefficient 108 of the low resolution error block. The inverse transform unit 137 performs inverse DCT transform on the decoded quantized transform coefficient 108 to generate a decoded low resolution error block 109. The adder 135 calculates, for each pixel of the low resolution prediction block 106, an addition value between the pixel at a position spatially coincident with the decoded low resolution error block 109 and each pixel, and the decoded low resolution block 110. Is generated.

  Next, the low-resolution block dividing unit 166 divides the decoded low-resolution block 110 into 2 × 2 low-resolution divided blocks 112, similarly to the block 106 in FIG. Each low-resolution divided block 112 is input to the small block prediction unit 158.

  The small block prediction unit 158 (corresponding to 152 in FIGS. 1 and 2) generates a 4 × 4 prediction small block 116 for each low resolution divided block 112 using the decoded image held in the frame memory 153. Is a processing unit. Here, the operation of the small block prediction unit 158 will be described. The expanded motion estimation unit 138 has a 2 × 2 low-resolution divided block 112 (112a to 112d) illustrated in FIG. 35 and four pixels spatially corresponding to the four pixels of the low-resolution divided block (candidate reduced prediction small block 114). And the motion vector 115 (115a to 115d) of the low resolution divided block is detected. The difference evaluation value at this time spatially corresponds to 4 pixels of the low resolution divided block among 16 pixels of “4 pixels of the low resolution divided block” and “4 × 4 block to be searched (candidate prediction small block 113)”. The sum of absolute differences is 4 pixels (candidate reduced prediction small block 114) ”(details will be described later using FIG. 32). Since a difference evaluation value using only such decoded images as input values can be calculated by the decoding device, the same motion vector as that of the encoding device can be detected by the decoding device. The predicted small block acquisition unit 139 acquires the 4 × 4 predicted small blocks 116 (116 a to 116 d) corresponding to the motion vectors 115 (115 a to 115 d) of the low resolution divided blocks from the decoded frame 520 of the frame memory 153.

  Next, the block integration unit 155 integrates the four prediction small blocks 116a to 116d as shown on the left side of FIG. 35, and generates an 8 × 8 prediction block 118 (FIG. 28).

  The error block decoding unit 156 is a processing unit that generates the decoded error block 121 by decoding the error block information (quantized transform coefficient) 119. Here, the operation of the error block decoding unit 156 will be described. The inverse quantization unit 140 inversely quantizes the quantized transform coefficient 119 to generate a decoded quantized transform coefficient 120. The inverse transform unit 141 performs inverse DCT transform on the decoded quantized transform coefficient 120 to generate a decoded error block 121.

  Next, the addition unit 157 calculates the addition value of the pixel at the position spatially coincident with the decoding error block 121 and each pixel for each pixel of the prediction block 118 and generates a decoding block 122. The decoding block 122 is held in the current decoding frame 610 of the frame memory 153.

[Description of processing flow]
FIG. 30 shows a moving picture decoding process flow for realizing the predictive decoding of the present invention, and FIG. 33 shows a moving picture coding process flow for realizing the predictive coding of the present invention. It should be noted that the processes with the same numbers in FIGS. 30 and 33 lead to the same process results.

[Video decoding process flow]
A moving picture decoding process flow for realizing the predictive coding of the present invention will be described with reference to FIG. When the decoding process for the next decoding block is started (process 401), first, the encoded data for one block required for decoding the next decoding block is sent to the entropy decoding unit 150 in FIG. Entered. In step 403, the input encoded data is entropy-decoded by the entropy decoding unit 150, and low-resolution block information and error block information are obtained. Of these, the low-resolution block information is subjected to low-resolution block decoding processing by the low-resolution block decoding unit 151 in process 404 to generate a decoded low-resolution block.

  Here, the low-resolution block decoding process flow of the process 404 will be described with reference to FIG. When the low-resolution block decoding process is started (process 421), first, the separation unit 131 acquires low-resolution block information (process 422), and the low-resolution block information is converted into a differential motion vector and a low-resolution error. Separated into block information (process 423). In the process 424, the motion vector prediction / decoding unit 132 calculates (decodes) the motion vector 104 by adding the value of each component of the motion vector predictor to the value of each component of the differential motion vector 103. The predicted motion vector is calculated, for example, by median prediction (for each component) for the motion vectors of three adjacent blocks on the left, immediately above, and upper right (upper left in the case of outside the frame) of the current block. The obtained motion vector is retained for the motion vector prediction process in the subsequent decoding of the decoded block.

  In the next process 425, the motion compensation unit 133 acquires the block 105 corresponding to the motion vector 104 from the decoded frame 520 that has been stored in the frame memory 153. In the next process 426, the acquired block 105 is converted into a low-resolution prediction block 106 in which pixels are extracted every other vertical and horizontal lines as shown in FIG.

  In the next process 427, the inverse quantization unit 136 inversely quantizes the quantized transform coefficient 107 of the low resolution error block to generate the decoded quantized transform coefficient 108 of the low resolution error block. Then, the inverse transform unit 137 performs inverse DCT transform on the decoded quantized transform coefficient 108 to generate a decoded low resolution error block 109. In process 428, the adding unit 135 calculates, for each pixel of the low resolution prediction block 106, an addition value between the pixel at a position spatially coincident with the decoded low resolution error block 109 and each pixel, A decoded low-resolution block 110 is generated, and the low-resolution block decoding process is terminated (process 429).

  Returning to FIG. In the process 405, the low resolution block dividing unit 166 divides the decoded low resolution block 110 into 2 × 2 low resolution division blocks 112. In processing 406, the small block prediction unit 158 performs small block prediction processing on the four low-resolution divided blocks 112.

  Here, the small block prediction processing flow will be described with reference to FIG. When the small block prediction process is started (process 430), first, in process 431, as an initial value, the evaluation value Emin is set to the maximum integer value and set to the divided motion vector 0 vector. Next, in process 432, the low resolution divided block 112 is input to the small block prediction unit 158. Subsequently, in the expanded motion estimation process of process 434, the expanded motion estimation unit 138 detects the motion vector of the low resolution divided block.

  In this expanded motion estimation process, candidate prediction small blocks are extracted in the spiral order from the center (0 vector) within the search range (± 16 pixels in the vertical and horizontal directions) of the decoded frame in the frame memory 153, and the evaluation value Emin is minimized. The predicted small block position is detected. More specifically, the next candidate predicted small block is acquired from the decoded frame of the frame memory 153 in the process 4341, and the acquired candidate predicted small block is processed in the process 4342, for example, with the block 106 in FIG. Similarly, it is converted into the dimension (number of pixels) of the low resolution prediction block. In the process 4343, the difference between the evaluation value E (“4 pixels of the low resolution divided block” and “4 pixels spatially corresponding to 4 pixels of the low resolution divided block among the 16 pixels of the 4 × 4 block to be searched”) In step 4344, it is determined whether or not the evaluation value E is smaller than the evaluation value Emin. If the evaluation value E is smaller than the evaluation value Emin, the evaluation value Emin is updated to the evaluation value E in processing 4345, and the divided motion vector is updated to the divided motion vector corresponding to the evaluation value E. . Thereafter, processing 4341 to processing 4345 are repeated until the search for all candidate prediction small blocks within the search range is completed. Finally, in process 435, the predicted small block acquisition unit 139 acquires the predicted small block corresponding to the divided motion vector having the smallest evaluation value E from the decoded frame in the frame memory 153, thereby performing the small block prediction process. Is terminated (process 436). By the above small block prediction process, four predicted small blocks are generated.

  Returning to FIG. The four predicted small blocks generated in the process 406 are integrated into a predicted block by the block integration unit 155 in the process 407. In the next process 408, the error block decoding unit 156 decodes (inverse quantization / inverse transform) error block information (quantized transform coefficient) to generate a decoded error block. In the next process 409, the addition unit 157 adds the prediction block and the decoding error block for each pixel to generate a decoding block. Finally, in process 410, the decoded block is held in the current decoded frame in the frame memory 153, and the block decoding process ends (process 411).

[Video coding process flow]
Next, an encoding process flow of an encoding target block that realizes predictive encoding according to the present invention will be described with reference to FIG. First, the coding block dividing unit 251 in FIG. 29 divides the input frame 201, whereby the coding target block 202 is obtained. When the encoding process of the encoding target block 202 is started (process 441), first, in process 442, the decoded low-resolution block generation unit 252 executes the low-resolution block encoding / decoding process of FIG. .

  Hereinafter, the low-resolution block encoding / decoding process will be described with reference to FIG. When the low-resolution block encoding / decoding process is started (process 451), first, in process 452, the next encoding target block is input. In process 453, the encoding target block and the decoded frame in the frame memory are input. From these, a motion vector is detected. Note that the motion vector detection method here is omitted because it has already been described in the configuration of the decoded low-resolution block generation unit 252 in FIG. In the next process 454, a block corresponding to the motion vector is obtained from the decoded frame in the frame memory. In the next process 455, after the predicted motion vector is generated, a subtraction process is performed on each component between the motion vector and the predicted motion vector to generate a differential motion vector. The motion vector and the generated differential motion vector are retained for motion vector prediction in subsequent encoding target blocks. The predicted motion vector generation method here is omitted because it has already been described in the configuration of the motion vector prediction / encoding unit 212 in FIG. In the next process 456, the encoding target block (prediction target block) is converted into a low resolution encoding target block (low resolution prediction target block), for example, similarly to the block 106 in FIG. The block corresponding to the motion vector is converted into a low resolution prediction block in the same manner as the encoding target block (prediction target block).

  In the process 458, subtraction is performed on each pixel between the low-resolution encoding target block (low-resolution prediction target block) and the low-resolution prediction block to generate a low-resolution error block. In the next process 459, the low resolution error block is encoded (transformed / quantized), and the quantized transform coefficient of the low resolution error block is generated. In the next process 427, the quantized transform coefficient is decoded (inverse quantization / inverse transform), and a decoded low-resolution error block is generated. In the next process 428, addition is performed for each pixel between the low-resolution prediction block and the decoded low-resolution error block to generate a decoded low-resolution block. Finally, in process 460, the differential motion vector and the quantized transform coefficient are integrated, low-resolution block information is generated, and the low-resolution block encoding / decoding process ends (process 461).

  Returning to FIG. In process 405, the low-resolution block dividing unit 166 divides the generated decoded low-resolution block into low-resolution divided blocks. In the process 406, the small block prediction unit 158 performs a small block prediction process on the divided low-resolution divided block to generate a predicted small block. Here, since the small block prediction process (FIG. 32) has already been described, the description thereof is omitted.

  In process 407, the block integration unit 155 integrates the generated prediction small block into the prediction block. In the next process 443, the subtraction unit 253 performs a subtraction process on each pixel between the encoding target block and the prediction block to generate an error block. In the next processing 444, the error block encoding unit 254 encodes (transforms / quantizes) the error block, and generates error block information (quantized transform coefficient). In the next process 408, the error block decoding unit 156 decodes (inverse quantization / inverse transform) the error block information (quantized transform coefficient) to generate a decoded error block. In the next process 409, the adder 157 performs an addition process for each pixel between the prediction block and the decoded error block to generate a decoded block. Then, after the generated decoded block is held in the current decoded frame of the frame memory 153 in process 410, in process 445, the entropy encoding unit 255 entropy converts the low-resolution block information and the error block information. The block encoding process is completed (process 446).

[Configuration example of computer system]
FIG. 16 is a diagram for explaining a case where the present invention is implemented by a computer system using a storage medium such as a flexible disk storing a moving image encoding program related to the image encoding process and a moving image decoding program related to the image decoding process of the above embodiment. FIG.

  FIG. 16B shows the appearance, cross-sectional structure, and flexible disk as seen from the front of the flexible disk, and FIG. 16A shows an example of the physical format of the flexible disk that is the recording medium body. The flexible disk FD is built in the case F, and a plurality of tracks Tr are formed concentrically on the surface of the disk from the outer periphery toward the inner periphery, and each track is divided into 16 sectors Se in the angular direction. Has been. Therefore, in the flexible disk storing the program, data as the program is recorded in an area allocated on the flexible disk FD.

  FIG. 16C shows a configuration for recording and reproducing the program on the flexible disk FD. When the program is recorded on the flexible disk FD, data as the program is written from the computer system Cs via the flexible disk drive. When the encoding device or the decoding device is constructed in a computer system by a program on a flexible disk, the program is read from the flexible disk by a flexible disk drive and transferred to the computer system.

  In the above description, a flexible disk is used as the data recording medium, but the same can be done using an optical disk. Further, the recording medium is not limited to this, and any recording medium such as an IC card or a ROM cassette capable of recording a program can be similarly implemented. In addition, the computer includes a DVD player, a set top box, a mobile phone and the like that have a CPU and perform processing and control by software.

  Next, the configuration of the moving image decoding program and the moving image encoding program will be described.

  As shown in FIG. 37, the moving picture decoding program 161 is stored in a program storage area 160a formed on the recording medium 160. The moving picture decoding program 161 includes a main module 161a that comprehensively controls the moving picture decoding process, an entropy decoding module 161b that executes the operation of the entropy decoding unit 150 in FIG. 28, and the low-resolution block decoding in FIG. 28, the low-resolution block decoding module 161c for executing the operation of the unit 151, the low-resolution block dividing module 161h for executing the operation of the low-resolution block dividing unit 166 of FIG. 28, and the small block prediction of FIG. The small block prediction module 161i for executing the operation of the unit 158 and the block integration module 161g for executing the operation of the block integration unit 155 of FIG.

  As shown in FIG. 38, the moving image encoding program 261 is stored in a program storage area 260a formed on the recording medium 260. The moving picture coding program 261 includes a main module 261a that controls the moving picture coding process in an integrated manner, a coding block division module 261b for executing the operation of the coding block division unit 251 in FIG. 29, and FIG. The decoded low-resolution block generation module 261c for executing the operation of the decoded low-resolution block generation unit 252 and the low-resolution block division module 261h for executing the operation of the low-resolution block division unit 166 of FIG. And a small block prediction module 261i for executing the operation of the small block prediction unit 158 of FIG. 29 and a block integration module 261g for executing the operation of the block integration unit 155 of FIG.

  A part or all of the moving picture decoding program 161 may be transmitted via a transmission medium such as a communication line, received by another device, and recorded (including installation). Similarly, a part or all of the moving image encoding program 261 may be transmitted via a transmission medium such as a communication line, received by another device, and recorded (including installation).

[Description of various modifications]
As mentioned above, although embodiment of this invention was described, the following modifications may be performed and any form is also included in this invention.

(1) Modified Mode Related to Prediction Method In the above description, the prediction method is limited to prediction in the time direction, but a decoded image that has already been decoded in the same frame as the encoding target block (prediction target block) or the decoding block. The same method can also be applied to the prediction of the spatial direction using. That is, based on the above-described decoded low-resolution block, for the low-resolution divided block obtained by re-dividing the decoded low-resolution block, using adjacent decoded pixels in the same frame, for example, H.264 (“Text of ISO / IEC 14496-10 Advanced Video Coding 3rd Edition”, September 2004.) After determining the optimal mode from multiple intra prediction modes, the high-resolution signal corresponding to the decoded low-resolution block The present invention also includes the case of predicting a block of the same. In this case, since it is not necessary to explicitly transmit the prediction mode of each low-resolution divided block, there is an excellent effect that the target block can be predicted with high accuracy with a small amount of auxiliary information. Moreover, you may perform highly accurate prediction combining the prediction of the time direction mentioned above, and the prediction of a spatial direction.

(2) Modified mode related to encoding of error block It is not essential to encode an error block between a prediction block and a block to be encoded. That is, the processing of the error block encoding unit 254 (FIG. 29) and the error block decoding unit 156 (FIGS. 28 and 29) may be omitted. Also, the presence or absence of encoding may be determined for each encoding target block or decoding block. Further, the specific encoding / decoding method of the error block and the low resolution error block is not limited to the above-described method.

(3) Modified Mode Related to Encoding in Multiple Prediction Mode In the above-described embodiment, the prediction method applied to the encoding process of the encoding target block is only the prediction method of the present invention, but the encoding target block is not reduced. The prediction method of the present invention may be applied to a conventional encoding method.

(4) Modifications related to block size In the above description, the size of the encoding target block (prediction target block) and the decoding block is set to 8x8, the size of the low resolution block is set to 4x4, and the low resolution divided block is set to 2x2. The predictive coding is not limited to these sizes. For example, the size of the encoding target block and the decoding block is 16x16 which is the same as H.264, the low resolution block is 8x8, and the low resolution divided block is 2x2. Even in this case, the present invention is effective. Similarly, the block size for performing transformation / quantization is not limited. For example, as in H.264, an error block may be divided into 4 × 4 blocks, and conversion and quantization may be performed in units of 4 × 4 blocks.

(5) Modified Mode Related to Reduction Method In the above description, as shown in FIG. 36, the reduction processing is a method of thinning out one line vertically and horizontally, but the block reduction processing is not limited to this method. For example, the low resolution block may have an arbitrary shape having a smaller number of pixels than the decoding block (the encoding target block and the prediction target block) as long as it is determined in advance between the encoding device and the decoding device. A block in which each pixel of the block 106 is shifted by one pixel in the vertical and horizontal directions (a block 1060 in FIG. 36), a block 1062 having both the pixel in the block 106 and the pixel in the block 1060 in FIG. In addition, various methods applied to image reduction, such as a method of down-sampling vertically and horizontally after applying low-pass filter processing to a block, and a method of averaging four adjacent pixels and converting them to one pixel are applied. can do.

(6) Modification concerning the implementation procedure of the reduction processing The reduction processing unit may be omitted. For example, when the function of the reduction processing unit is included in the motion compensation unit or the expanded motion estimation unit, or when the processing can be performed without including pixels that do not belong to the low resolution block, the reduction processing unit Is not required. In addition, different methods may be used for the reduction process by the low-resolution block decoding unit and the reduction process by the low-resolution prediction block generation unit as long as they are defined in advance between the decoding apparatus and the encoding apparatus.

(7) Modified Mode Related to Expanded Motion Estimation The search range, search procedure, and search start point in the expanded motion estimation process are not limited to the method described above. The search range can be set to an arbitrary range as long as it is defined in advance between the encoding device and the decoding device. The same applies to the search procedure. For example, instead of a spiral from the center, a method of searching the search range from the upper left to the lower right in the raster scan order can be applied. As for the search start point, it is also possible to set a search range centering on the prediction vector, for example, instead of the zero vector. In addition, the motion estimation / compensation method is not limited to block matching for detecting a motion vector, and a method for calculating a predicted value of a pixel in a block using a motion parameter such as an affine parameter or a perspective projection transformation parameter may be used. The invention is applicable. Also in this case, it is not necessary to transmit the motion parameter detected by the small block prediction unit. Similarly, the small block prediction of the first embodiment is not limited to block matching.

(8) Modifications Regarding Separation Unit and Integration Unit The separation unit 131 in FIG. 28 and the integration unit 216 in FIG. 29 are not necessarily required. For example, in FIG. 28, the motion vector prediction / decoding unit 132 and the inverse quantization unit 136 may directly receive data from the entropy decoding unit 150. In FIG. 29, data may be directly output from the motion vector prediction / encoding unit 212 and the quantization unit 215 to the entropy encoding unit 255.

  DESCRIPTION OF SYMBOLS 100, 300, 500 ... Moving image encoding apparatus, 131 ... Separation part, 132 ... Vector prediction and decoding part, 133 ... Compensation part, 134 ... Reduction process part, 135 ... Addition part, 136 ... Dequantization part, 137 ... inverse transform unit, 138 ... estimation unit, 139 ... predicted small block acquisition unit, 140 ... inverse quantization unit, 141 ... inverse transform unit, 150 ... entropy decoding unit, 151 ... low resolution block decoding unit, 152, 158 ... Small block prediction unit, 153 ... frame memory, 154 ... enlarged block generation unit, 155 ... block integration unit, 156 ... error block decoding unit, 157 ... addition unit, 159 ... switch, 160 ... recording medium, 162 ... block division unit, 163 ... Estimating unit, 164 ... Compensating unit, 165 ... Block dividing unit, 166 ... Low resolution block dividing unit, 200, 400, 600 ... Video decoding device, 21 Estimator 212, vector predictor / encoder 213, subtractor 214, converter 215, quantizer 216, integrator 217, adder 218, converter 219, quantizer 251... Encoded block dividing unit, 252... Decoded low resolution block generating unit, 253... Subtracting unit, 254... Error block encoding unit, 255 ... Entropy encoding unit, 256.

Claims (12)

A video encoding device that generates a prediction block in block units,
Holding means for holding the decoded image;
A first dividing unit that divides the input encoding target image into a plurality of prediction target blocks;
After the prediction target block is converted and encoded into a low resolution block having a smaller number of pixels than the prediction target block, a decoded low resolution block is generated by decoding the encoded data of the low resolution block Generating means for
Expansion means for expanding the generated decoded low-resolution block to a block having the same number of pixels as the prediction target block;
A second dividing unit that generates a plurality of small blocks by dividing the enlarged block based on a predetermined division rule;
Prediction means for generating a block in the decoded image corresponding to a motion vector of the generated small block detected by block matching between the generated small block and the decoded image as a predicted small block; ,
A moving image encoding apparatus comprising: an integration unit that generates a prediction block by integrating the plurality of generated prediction small blocks based on the division rule. Third dividing means for dividing the encoding target image into encoding target blocks;
Subtracting means for generating an error block by subtracting the divided encoding target block and the prediction block in units of pixels;
Encoding means for encoding the generated error block;
Decoding means for decoding the encoded data and generating a decoding error block;
The moving image encoding apparatus according to claim 1, further comprising: an output unit that adds the prediction block and the decoding error block in units of pixels to generate a decoded block, and outputs the decoded block to the holding unit. The enlargement unit generates a block having the same number of pixels as the prediction target block using the pixel of the generated decoded low-resolution block and the pixel of the decoded block adjacent to the upper side and the left side of the prediction target block. The moving picture coding apparatus according to claim 1, wherein: The generating means is
Means for converting the prediction target block into a low resolution prediction target block having a smaller number of pixels than the prediction target block based on a predetermined conversion rule;
Low-resolution prediction block generation means for generating a low-resolution prediction block;
Means for subtracting the low-resolution prediction target block and the low-resolution prediction block in units of pixels to generate a low-resolution error block;
Means for generating a decoded low-resolution error block by decoding the encoded data of the low-resolution error block after encoding the low-resolution error block;
Means for adding the decoded low-resolution error block and the low-resolution prediction block in units of pixels to restore the low-resolution block;
Comprising
The low-resolution prediction block generating means is
Means for generating a second prediction block having the same number of pixels as the prediction block;
Means for generating the low-resolution prediction block from the second prediction block based on the conversion rule;
Comprising
The generating means further includes:
When all the difference pixel values in the low-resolution error block are zero, the second prediction block is configured to output as the prediction block.
The moving picture coding apparatus according to claim 1, wherein: A video decoding device that generates a prediction block in block units,
Holding means for holding the decoded image;
Decoding means for generating a low-resolution block having a smaller number of pixels than a prediction block based on low-resolution block information obtained by decoding input encoded data and the decoded image;
Enlarging means for enlarging the generated low-resolution block to a block having the same number of pixels as the prediction block;
Dividing means for generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule;
Prediction means for generating a block in the decoded image corresponding to a motion vector of the divided small block detected by block matching between the divided small block and the decoded image as a predicted small block; ,
A moving picture decoding apparatus provided with the integrated means which produces | generates a prediction block by integrating the produced | generated several prediction small block based on the said division | segmentation rule. Generating means for decoding the encoded data and generating an error block;
6. The moving picture decoding apparatus according to claim 5, further comprising: an output unit that adds the prediction block and the error block in units of pixels to generate a decoded block, and outputs the decoded block to the holding unit. The expansion means generates a block having the same number of pixels as the prediction block, using the pixel of the generated low resolution block and the pixel of the decoded block adjacent to the upper side and the left side of the prediction block. The moving picture decoding apparatus according to claim 5. The decoding means comprises:
Means for decoding the input encoded data to decode a low-resolution error block;
Low-resolution prediction block generation means for generating a low-resolution prediction block;
Means for adding the low resolution error block and the low resolution prediction block in units of pixels to generate the low resolution block;
Comprising
The low-resolution prediction block generating means is
Means for generating a second prediction block having the same number of pixels as the prediction block;
Means for generating the low-resolution prediction block from the second prediction block based on a predetermined rule;
Comprising
The decoding means further includes
When all the difference pixel values in the low-resolution error block are zero, the second prediction block is configured to output as the prediction block.
The moving picture decoding apparatus according to claim 5, wherein: A video encoding method for generating a prediction block in units of blocks using a decoded image,
A first dividing step of dividing the input encoding target image into a plurality of prediction target blocks;
After the prediction target block is converted and encoded into a low resolution block having a smaller number of pixels than the prediction target block, a decoded low resolution block is generated by decoding the encoded data of the low resolution block Generating step to
An expansion step of expanding the generated decoded low-resolution block to a block having the same number of pixels as the prediction target block;
A second division step of generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule;
A prediction step of generating a block in the decoded image corresponding to a motion vector of the generated small block detected by block matching between the generated small block and the decoded image as a predicted small block; ,
A moving picture encoding method comprising: an integrating step of generating a prediction block by integrating the plurality of generated prediction small blocks based on the division rule. A video decoding method for generating a prediction block in block units using a decoded image,
A decoding step for generating a low-resolution block having a smaller number of pixels than a prediction block, based on low-resolution block information obtained by decoding input encoded data and the decoded image;
Expanding the generated low-resolution block to a block having the same number of pixels as the prediction block;
A division step of generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule;
A prediction step of generating a block in the decoded image corresponding to a motion vector of the divided small block detected by block matching between the divided small block and the decoded image as a predicted small block; ,
A moving picture decoding method comprising: an integrating step of generating a prediction block by integrating the plurality of generated prediction small blocks based on the division rule. In a computer comprising holding means for holding a decoded image and a moving picture coding program having a function of generating a prediction block in units of blocks,
A first division process for dividing the input encoding target image into a plurality of prediction target blocks;
After the prediction target block is converted and encoded into a low resolution block having a smaller number of pixels than the prediction target block, a decoded low resolution block is generated by decoding the encoded data of the low resolution block Generation process to
An expansion process for expanding the generated decoded low-resolution block to a block having the same number of pixels as the prediction target block;
A second division process for generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule;
A prediction process for generating a block in the decoded image corresponding to a motion vector of the generated small block detected by block matching between the generated small block and the decoded image as a predicted small block; ,
A moving picture encoding program that performs an integration process for generating a prediction block by integrating the plurality of generated prediction small blocks based on the division rule. In a computer comprising holding means for holding a decoded image and a moving picture decoding program having a function of generating a prediction block in block units,
A decoding process for generating a low-resolution block having a smaller number of pixels than the prediction block, based on the low-resolution block information obtained by decoding the input encoded data and the decoded image;
An expansion process for expanding the generated low-resolution block to a block having the same number of pixels as the prediction block;
A division process for generating a plurality of small blocks by dividing the enlarged block based on a predetermined division rule;
A prediction process for generating a block in the decoded image corresponding to a motion vector of the divided small block detected by block matching between the divided small block and the decoded image as a predicted small block; ,
A moving picture decoding program that performs an integration process for generating a prediction block by integrating the plurality of generated prediction small blocks based on the division rule.

Images