JP2008289005A - Image-predicting/encoding device, image prediction/encoding method, image prediction/encoding program, image-predicting/decoding device, image prediction/decoding method, and image prediction/decoding program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image-predicting/encoding device or an image-predicting/decoding device which selects a plurality of candidate prediction signals without increasing the amount of information. <P>SOLUTION: In a prediction signal generator 103c, a template-matching unit 201 searches for a plurality of prediction adjacent regions 404a-414c with high correlation with the already reproduced object adjacent region 403 adjacent to an object region 402 from search regions 401 and 417. Then, a combination setting unit 237 selects one or more prediction adjacent regions from among the plurality of searched prediction adjacent regions, at least using coordinate information indicating positions of the prediction adjacent regions, for example, reference image signals and motion vectors. A prediction region acquiring unit 204 generates one or more candidate prediction signals of object pixel signals, on the basis of the selected prediction adjacent regions, and a weighting unit 205 and an adder 206 process the candidate prediction signals by using a predetermined synthesis method to generate a prediction signal. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image predictive encoding device, an image predictive encoding method, an image predictive encoding program, an image predictive decoding device, an image predictive decoding method, and an image predictive decoding program. The present invention relates to an image predictive encoding device, an image predictive encoding method, an image predictive encoding program, an image predictive decoding device, an image predictive decoding method, and an image predictive decoding program.

  In order to efficiently transmit and store still images and moving image data, compression coding technology is used. In the case of moving images, MPEG-1 to 4 and ITU (International Telecommunication Union) H.264. 261-H. H.264 is widely used.

  In these encoding methods, encoding / decoding processing is performed after an image to be encoded is divided into a plurality of blocks. In predictive coding within a screen, a prediction signal is generated using an adjacent previously reproduced image signal (reconstructed compressed image data) in the same screen as the target block, and the generated A differential signal obtained by subtracting the prediction signal from the signal of the target block is encoded. In predictive coding between screens, the adjacent reproduced image signal in the screen different from the target block is referred to, the motion is corrected, the predicted signal is generated, and the predicted signal is subtracted from the signal of the target block. The difference signal is encoded.

  For example, H.M. H.264 intra-screen predictive encoding employs a method of generating a prediction signal by extrapolating already reproduced pixel values adjacent to a block to be encoded in a predetermined direction. FIG. 14 shows ITUH. 2 is a schematic diagram for explaining an intra-screen prediction method used for H.264. In FIG. 14A, a target block 1802 is a block to be encoded, and a pixel group 1801 composed of pixels A to M adjacent to the boundary of the target block 1802 is an adjacent region, and has already been obtained in past processing. It is the reproduced image signal.

  In this case, a prediction signal is generated by extending downward the pixel group 1801, which is an adjacent pixel immediately above the target block 1802. In FIG. 14B, the already reproduced pixels (I to L) on the left of the target block 1804 are stretched to the right to generate a prediction signal. Thus, the difference between each of the nine prediction signals generated by the method shown in FIGS. 14A to 14I and the pixel signal of the target block is taken, and the one with the smallest difference value is set as the optimum prediction signal. . As described above, a prediction signal can be generated by extrapolating pixels. The above contents are described in Patent Document 1 below.

  In normal inter-picture prediction encoding, a prediction signal is generated by a method in which a signal similar to the pixel signal is searched from an already reproduced screen for a block to be encoded. Then, a motion vector, which is a spatial displacement amount between the target block and a region formed by the searched signal, and a residual signal between the pixel signal and the prediction signal of the target block are encoded. Such a method of searching for a motion vector for each block is called block matching. FIG. 15 is a schematic diagram for explaining the block matching process. Here, a procedure for generating a prediction signal will be described using the target block 1402 on the encoding target screen 1401 as an example. A screen 1403 in FIG. 15A is a screen that has already been reproduced, and a region 1404 indicated by a broken line is a region in the same spatial position as the target block 1402. In block matching, a search range 1405 surrounding the region 1404 is set, and a region 1406 where the absolute value error sum with the pixel signal of the target block 1402 is minimized is detected from the pixel signals in this search range. A pixel signal in the region 1406 becomes a prediction signal, and a displacement amount from the region 1404 to the region 1406 is detected as a motion vector. ITUH. H.264 prepares a plurality of prediction types having different block sizes for encoding motion vectors in order to cope with changes in local features of an image. ITU H. H.264 prediction types are described in Patent Document 2, for example.

In the compression encoding of moving image data, the encoding order of each frame may be arbitrary. For this reason, there are three types of coding order methods for inter-screen prediction in which a prediction signal is generated with reference to a reproduced screen. The first method refers to forward prediction in which a prediction signal is generated with reference to a previously played screen in the playback order, and the second method refers to backward prediction with reference to a future played screen in the playback order. The method 3 is bidirectional prediction in which both forward prediction and backward prediction are performed and two prediction signals are averaged. About the kind of prediction between screens, it describes in patent document 3, for example.
US Pat. No. 6,765,964 US Patent Publication No. 7003035 US Pat. No. 6,259,739

  In the prior art, since a predicted signal for each pixel is generated by copying a reproduced pixel value including distortion due to encoding (for example, quantization noise), the predicted signal also includes encoding distortion. Thus, the prediction signal contaminated by the coding distortion becomes a factor of lowering the coding efficiency such as an increase in the code amount of the residual signal and deterioration in reproduction image quality.

  Although the influence of coding distortion can be suppressed by smoothing the prediction signal, additional information for indicating a plurality of candidate prediction signals to be smoothed is required, and the amount of information increases.

  Therefore, in the present invention, an image predictive encoding device, an image predictive encoding method, an image predictive encoding program, an image predictive decoding device, an image predictive decoding method, and an image that select a plurality of candidate prediction signals without increasing the amount of information An object is to provide a predictive decoding program.

  In order to solve the above-described problem, an image predictive coding apparatus according to the present invention includes an area dividing unit that divides an input image into a plurality of areas, and a processing target of the plurality of areas divided by the area dividing unit. Prediction signal generation means for generating a prediction signal for the target pixel signal in the target area, and residual signal generation means for generating a residual signal between the prediction signal generated by the prediction signal generation means and the target pixel signal; Encoding means for encoding the residual signal generated by the residual signal generating means, wherein the prediction signal generating means is an adjacent pixel that has already been reproduced and is adjacent to a target area composed of the target pixel signal. A plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of a signal are searched from a search area consisting of already reproduced images, and at least using the coordinate information indicating the position of the predicted adjacent area, One or more prediction adjacent regions are selected from the searched plurality of prediction adjacent regions, one or more candidate prediction signals of the target pixel signal are generated based on the selected prediction adjacent region, and the candidate prediction signal is preliminarily generated. A configuration is provided in which a prediction signal is generated by processing using a predetermined synthesis method.

  According to the present invention, a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from the search area made of the already-reproduced images, and at least using the coordinate information indicating the position of the predicted adjacent area, One or more prediction adjacent regions are selected from the prediction adjacent regions, one or more candidate prediction signals of the target pixel signal are generated based on the selected prediction adjacent regions, and a candidate prediction signal is determined using a predetermined synthesis method A prediction signal can be generated by processing. As a result, it is possible to select a combination of candidate prediction signals suitable for smoothing without increasing the amount of information using a target adjacent region composed of already reproduced adjacent pixel signals adjacent to the target region with diversity. Therefore, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics, and to perform efficient encoding.

  In addition, the image predictive coding apparatus according to the present invention includes a region dividing unit that divides an input image into a plurality of regions, and a target pixel of a target region that is a processing target among the plurality of regions divided by the region dividing unit. A prediction signal generating means for generating a prediction signal for the signal, a residual signal generating means for generating a residual signal between the prediction signal generated by the prediction signal generating means and the target pixel signal, and the residual signal generating means Encoding means for encoding the residual signal generated by the prediction signal generation means, and the prediction signal generation means includes a target adjacent area composed of already reproduced adjacent pixel signals adjacent to the target area composed of the target pixel signal; A plurality of predicted adjacent areas having a high correlation are searched from a search area composed of already reproduced images, the number of predicted adjacent areas to be selected is set, and coordinate information indicating at least the position of the predicted adjacent area is obtained. And selecting the set number of prediction adjacent regions from the search region, generating the set number of candidate prediction signals of the target pixel signal based on the selected prediction adjacent region, and generating the candidate prediction signal in advance A configuration is provided in which a prediction signal is generated by processing using a predetermined synthesis method.

  According to the present invention, a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from a search area composed of already reproduced images, the number of predicted adjacent areas to be selected is set, and at least the position of the predicted adjacent area is determined. The number of predicted adjacent areas set from the search area is selected using the coordinate information shown, and the number of candidate prediction signals set for the target pixel signal is generated based on the selected predicted adjacent area, and the candidate prediction signals are determined in advance. The prediction signal can be generated by processing using the synthesis method. As a result, it is possible to select a combination of candidate prediction signals suitable for smoothing without increasing the amount of information using a target adjacent region composed of already reproduced adjacent pixel signals adjacent to the target region with diversity. Therefore, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics, and to perform efficient encoding.

  In addition, the image predictive coding apparatus according to the present invention includes a region dividing unit that divides an input image into a plurality of regions, and a target pixel of a target region that is a processing target among the plurality of regions divided by the region dividing unit. A prediction signal generating means for generating a prediction signal for the signal, a residual signal generating means for generating a residual signal between the prediction signal generated by the prediction signal generating means and the target pixel signal, and the residual signal generating means Encoding means for encoding the residual signal generated by the prediction signal generation means, and the prediction signal generation means includes a target adjacent area composed of already reproduced adjacent pixel signals adjacent to the target area composed of the target pixel signal; A plurality of predicted adjacent areas having a high correlation are searched from a search area composed of already reproduced images, and at least different numbers of predicted adjacent areas are included by using coordinate information indicating the positions of the predicted adjacent areas. Two or more combinations of predicted adjacent regions are generated, and pixel signals of predicted adjacent regions belonging to the combination are processed using a predetermined synthesis method, thereby generating a comparison signal for each of the adjacent pixel signals, and comparing A combination of predicted adjacent areas having a high correlation between a signal and the adjacent pixel signal is selected from the generated combinations of predicted adjacent areas, and based on the predicted adjacent areas belonging to the selected combination, the target pixel signal A configuration is provided in which one or more candidate prediction signals are generated and the prediction signals are generated by processing the candidate prediction signals using a predetermined synthesis method.

  According to the present invention, a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from the search area made of the already reproduced images, and at least different numbers of predictions are made using the coordinate information indicating the positions of the predicted adjacent areas. Generate two or more combinations of predicted adjacent areas including adjacent areas, and process the pixel signals of the predicted adjacent areas belonging to the combination using a predetermined synthesis method, thereby generating a comparison signal for the adjacent pixel signals, A combination of prediction adjacent regions having a high correlation between the comparison signal and the adjacent pixel signal is selected, and one or more candidate prediction signals of the target pixel signal are generated based on the prediction adjacent regions belonging to the selected combination, and the candidate prediction signal Can be generated using a predetermined synthesis method to generate a prediction signal. As a result, it is possible to select a combination of candidate prediction signals suitable for smoothing without increasing the amount of information using a target adjacent region composed of already reproduced adjacent pixel signals adjacent to the target region with diversity. Therefore, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics, and to perform efficient encoding.

  In the image predictive coding apparatus according to the present invention, it is preferable that the prediction signal generation unit selects a prediction adjacent region using coordinate information including position information in either time or space or time and space.

  In addition, according to the present invention, it is possible to select a predicted adjacent region using coordinate information including position information in either time or space or time and space, which is suitable for smoothing without increasing the amount of information. Combinations of candidate prediction signals can be selected with diversity.

  In the image predictive coding apparatus according to the present invention, it is preferable that the prediction signal generation unit performs processing using coordinate information including an identification number of a reference image to which a prediction adjacent region belongs.

  Further, according to the present invention, it is possible to process using coordinate information including the identification number of the reference image to which the prediction adjacent region belongs, and to combine candidate prediction signals suitable for smoothing without increasing the amount of information. , You can choose with diversity.

  In the image predictive coding apparatus according to the present invention, it is preferable that the prediction signal generating unit selects a predicted adjacent area using coordinate information including spatial position information on a reference image to which the predicted adjacent area belongs.

  According to the present invention, a predicted adjacent area can be selected using coordinate information including spatial position information on a reference image to which the predicted adjacent area belongs, and candidate prediction suitable for smoothing without increasing the amount of information The combination of signals can be selected with diversity.

  Further, in the image predictive coding apparatus according to the present invention, the prediction signal generation means uses the coordinate information including a motion vector indicating a spatial positional relationship between the target region and the prediction adjacent region to predict adjacent It is preferable to select a region.

  According to the present invention, a prediction adjacent region can be selected using coordinate information including a motion vector of the prediction adjacent region, and various combinations of candidate prediction signals suitable for smoothing can be obtained without increasing the amount of information. You can choose with sex.

  In addition, the image predictive decoding apparatus according to the present invention includes a data decoding unit that decodes encoded data of a residual signal related to a target region that is a processing target in compressed data, and a signal obtained by decoding by the data decoding unit. Residual signal restoration means for restoring the reproduction residual signal, prediction signal generation means for generating a prediction signal for the target pixel signal in the target region, the prediction signal generated by the prediction signal generation means, and the residual signal recovery Reproduction image signal generation means for generating a reproduction image signal by adding the reproduction residual signal restored by the means, wherein the prediction signal generation means is adjacent to the target region composed of the target pixel signal. A plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signals to be searched from the search area consisting of the already reproduced image, and the plurality of searched predictions Selecting one or more predicted adjacent areas from the plurality of searched predicted adjacent areas, using coordinate information indicating positions of at least two predicted adjacent areas including at least one tangential area; A configuration is provided in which one or more candidate prediction signals of the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. ing.

  According to the present invention, a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from the search area made of the already-reproduced images, and at least using the coordinate information indicating the position of the predicted adjacent area, One or more prediction adjacent regions are selected from the prediction adjacent regions, one or more candidate prediction signals of the target pixel signal are generated based on the selected prediction adjacent regions, and a candidate prediction signal is determined using a predetermined synthesis method A prediction signal can be generated by processing. As a result, it is possible to select a combination of candidate prediction signals suitable for smoothing without increasing the amount of information using a target adjacent region composed of already reproduced adjacent pixel signals adjacent to the target region with diversity. Therefore, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, it is possible to decode the encoded data that is efficiently encoded by the prediction signal generated in this way.

  The image predictive decoding apparatus according to the present invention also includes a data decoding unit that decodes encoded data of a residual signal related to a target region that is a processing target of compressed data, and a reproduction from a signal obtained by decoding by the data decoding unit. A residual signal restoring means for restoring a residual signal, a prediction signal generating means for generating a prediction signal for the target pixel signal in the target region, a prediction signal generated by the prediction signal generating means, and the residual signal restoring means Regenerated image signal generating means for generating a regenerated image signal by adding the regenerated residual signal restored in step (a), wherein the prediction signal generating means is adjacent to the already-regenerated image adjacent to the target pixel signal. A plurality of predicted adjacent areas having a high correlation with the target adjacent area made up of pixel signals are searched from a search area made up of already reproduced images and selected. The number of regions is set, and at least using the coordinate information indicating the position of the predicted adjacent region, the set number of predicted adjacent regions is selected from every position of the search region, and based on the selected predicted adjacent region , A number of the candidate prediction signals of the target pixel signal are generated, and the prediction signal is generated by processing the candidate prediction signals using a predetermined synthesis method.

  According to the present invention, a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from a search area composed of already reproduced images, the number of predicted adjacent areas to be selected is set, and at least the position of the predicted adjacent area is determined. The number of predicted adjacent areas set from the search area is selected using the coordinate information shown, and the number of candidate prediction signals set for the target pixel signal is generated based on the selected predicted adjacent area, and the candidate prediction signals are determined in advance. The prediction signal can be generated by processing using the synthesis method. As a result, it is possible to select a combination of candidate prediction signals suitable for smoothing without increasing the amount of information using a target adjacent region composed of already reproduced adjacent pixel signals adjacent to the target region with diversity. Therefore, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, it is possible to decode the encoded data that is efficiently encoded by the prediction signal generated in this way.

  In addition, the image predictive decoding apparatus according to the present invention includes a data decoding unit that decodes encoded data of a residual signal related to a target region that is a processing target in compressed data, and a signal obtained by decoding by the data decoding unit. Residual signal restoration means for restoring the reproduction residual signal, prediction signal generation means for generating a prediction signal for the target pixel signal in the target region, the prediction signal generated by the prediction signal generation means, and the residual signal recovery A reproduction image signal generation unit that generates a reproduction image signal by adding the reproduction residual signal restored by the unit, and the prediction signal generation unit includes a reproduced image signal adjacent to the target pixel signal. Prediction in which a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from the search area including the already-reproduced image and selected for the target adjacent area including the adjacent pixel signal The number of contact areas is set, and at least two combinations of predicted adjacent areas including different numbers of predicted adjacent areas are generated using coordinate information representing the positions of the predicted adjacent areas, and predicted adjacent areas belonging to the combination By generating a comparison signal for the adjacent pixel signal by processing the pixel signal using a predetermined synthesis method, and generating a combination having a high correlation between the comparison signal and the adjacent pixel signal. And generating one or more candidate prediction signals of the target pixel signal based on a prediction adjacent region belonging to the selected combination, and using the predetermined synthesis method It has a configuration for generating a prediction signal by processing.

  According to the present invention, a plurality of predicted adjacent areas having a high correlation with the target adjacent area are searched from the search area made of the already reproduced images, and at least different numbers of predictions are made using the coordinate information indicating the positions of the predicted adjacent areas. Generate two or more combinations of predicted adjacent areas including adjacent areas, and process the pixel signals of the predicted adjacent areas belonging to the combination using a predetermined synthesis method, thereby generating a comparison signal for the adjacent pixel signals, A combination of prediction adjacent regions having a high correlation between the comparison signal and the adjacent pixel signal is selected, and one or more candidate prediction signals of the target pixel signal are generated based on the prediction adjacent regions belonging to the selected combination, and the candidate prediction signal Can be generated using a predetermined synthesis method to generate a prediction signal. As a result, it is possible to select a combination of candidate prediction signals suitable for smoothing without increasing the amount of information using a target adjacent region composed of already reproduced adjacent pixel signals adjacent to the target region with diversity. Therefore, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, it is possible to decode the encoded data that is efficiently encoded by the prediction signal generated in this way.

  In the image predictive decoding device of the present invention, it is preferable that the prediction signal generation unit selects a prediction adjacent region using coordinate information including a time or space position, or a time-space position.

  Further, according to the present invention, it is possible to process using coordinate information including the identification number of the reference image to which the prediction adjacent region belongs, and to combine candidate prediction signals suitable for smoothing without increasing the amount of information. , You can choose with diversity.

  In the image predictive decoding device of the present invention, it is preferable that the prediction signal generating unit selects a predicted adjacent area using coordinate information including an identification number of a reference image to which the predicted adjacent area belongs.

  According to the present invention, a predicted adjacent area can be selected using coordinate information including spatial position information on a reference image to which the predicted adjacent area belongs, and candidate prediction suitable for smoothing without increasing the amount of information The combination of signals can be selected with diversity.

  In the image predictive decoding device of the present invention, it is preferable that the prediction signal generation unit selects a predicted adjacent area using coordinate information including a spatial position on a reference image to which the predicted adjacent area belongs.

  According to the present invention, a predicted adjacent area can be selected using coordinate information including spatial position information on a reference image to which the predicted adjacent area belongs, and candidate prediction suitable for smoothing without increasing the amount of information The combination of signals can be selected with diversity.

  Further, in the image predictive decoding device of the present invention, the prediction signal generation means uses the coordinate information including a motion vector indicating a spatial positional relationship between the target region and the prediction adjacent region to predict the adjacent region. Is preferably selected.

  According to the present invention, a prediction adjacent region can be selected using coordinate information including a motion vector of the prediction adjacent region, and various combinations of candidate prediction signals suitable for smoothing can be obtained without increasing the amount of information. You can choose with sex.

  By the way, the present invention can be described as the invention of the image predictive encoding device or the image predictive decoding device as described above, and as described below, the image predictive encoding method, the image predictive decoding method, the image predictive encoding program, It can also be described as an invention of an image predictive decoding program. These are substantially the same inventions only in different categories, and exhibit the same actions and effects.

  That is, the image predictive coding method of the present invention includes a region dividing step for dividing an input image into a plurality of regions, and a target pixel of a target region that is a processing target of the plurality of regions divided by the region dividing step. A prediction signal generation step for generating a prediction signal for the signal, a residual signal generation step for generating a residual signal between the prediction signal generated by the prediction signal generation step and the target pixel signal, and the residual signal generation step An encoding step for encoding the residual signal generated by the step, wherein the prediction signal generating step includes a target adjacent region composed of an already reproduced adjacent pixel signal adjacent to the target region composed of the target pixel signal, and A plurality of predicted adjacent areas having a high correlation with each other, searched from a search area consisting of already reproduced images, and at least coordinate information indicating the position of the predicted adjacent area And selecting one or more predicted adjacent areas from the plurality of searched predicted adjacent areas, generating one or more candidate predicted signals of the target pixel signal based on the selected predicted adjacent areas, and The prediction signal is generated by processing the prediction signal using a predetermined synthesis method.

  The image predictive decoding method according to the present invention includes a data decoding step for decoding encoded data of a residual signal related to a target region to be processed in compressed data, and a signal obtained by decoding by the data decoding step. A residual signal restoration step for restoring a reproduction residual signal from the prediction signal, a prediction signal generation step for generating a prediction signal for the target pixel signal in the target region, a prediction signal generated by the prediction signal generation step, and the residual signal A reproduction image signal generation step of generating a reproduction image signal by adding the reproduction residual signal restored in the restoration step, and the prediction signal generation step is performed on a target region composed of the target pixel signal. A plurality of predicted adjacent regions having a high correlation with the target adjacent region made up of adjacent pre-reproduced adjacent pixel signals are searched for from the pre-reproduced image From the plurality of predicted adjacent areas searched using coordinate information indicating the position of at least one of the two or more arbitrary predicted adjacent areas including at least one of the searched predicted adjacent areas. One or more prediction adjacent regions are selected, one or more candidate prediction signals of the target pixel signal are generated based on the selected prediction adjacent regions, and the candidate prediction signals are processed using a predetermined synthesis method Thus, a configuration for generating a prediction signal is provided.

  The image predictive coding program according to the present invention includes an area division module that divides an input image into a plurality of areas, and a target pixel of a target area that is a processing target of the plurality of areas divided by the area division module. A prediction signal generation module that generates a prediction signal for the signal, a residual signal generation module that generates a residual signal between the prediction signal generated by the prediction signal generation module and the target pixel signal, and the residual signal generation module An encoding module that encodes the residual signal generated by the prediction signal generation module, and the prediction signal generation module includes a target adjacent region including a reconstructed adjacent pixel signal adjacent to the target region including the target pixel signal; A plurality of predicted adjacent areas having a high correlation of One or more predicted adjacent areas are selected from the searched plurality of predicted adjacent areas using coordinate information indicating a position, and one candidate predicted signal of the target pixel signal is selected based on the selected predicted adjacent area. A configuration for generating a prediction signal by generating the above and processing the candidate prediction signal using a predetermined synthesis method is provided.

  The image predictive decoding program of the present invention includes a data decoding module that decodes encoded data of a residual signal related to a target region that is a processing target in compressed data, and a signal obtained by decoding by the data decoding module. A residual signal restoration module for restoring a reproduction residual signal from the prediction signal, a prediction signal generation module for generating a prediction signal for a target pixel signal in the target region, a prediction signal generated by the prediction signal generation module, and the residual signal A reproduction image signal generation module that generates a reproduction image signal by adding the reproduction residual signal restored by the restoration module, and the prediction signal generation module is provided in a target region including the target pixel signal. A plurality of predicted adjacent areas having a high correlation with the target adjacent area composed of adjacent adjacent pixel signals that have been reproduced are Searching from a search area composed of raw images, and using the coordinate information indicating the position of at least one predicted at least one predicted adjacent area including at least one of the searched predicted adjacent areas, the plurality of searched One or more predicted adjacent areas are selected from the predicted adjacent areas, one or more candidate predicted signals of the target pixel signal are generated based on the selected predicted adjacent area, and the candidate predicted signal is synthesized in advance It has the structure which produces | generates a prediction signal by processing using the method.

  According to the present invention, by using the target adjacent region composed of the already reproduced adjacent pixel signals adjacent to the target region, the combination of candidate prediction signals suitable for smoothing without increasing the amount of information is provided with diversity. Since the selection can be made, it is possible to efficiently generate a prediction signal in consideration of local noise characteristics.

The present invention can be readily understood by considering the following detailed description with reference to the accompanying drawings shown for the embodiments. Subsequently, embodiments of the present invention will be described with reference to the accompanying drawings. Where possible, the same parts are denoted by the same reference numerals, and redundant description is omitted.
<First Embodiment>

  FIG. 1 is a block diagram illustrating an image predictive coding apparatus 100 that can execute an image predictive coding method according to an embodiment of the present invention. The image predictive coding apparatus 100 includes an input terminal 101, a block divider 102 (region dividing means), a prediction signal generator 103 (predicted signal generating means), a frame memory 104, a subtractor 105 (residual signal generating means), It includes a converter 106 (encoding means), a quantizer 107 (encoding means), an inverse quantizer 108, an inverse transformer 109, an adder 110, an entropy encoder 111, and an output terminal 112. Yes. The converter 106 and the quantizer 107 function as encoding means.

  The configuration of the image predictive encoding device 100 configured as described above will be described below.

  The input terminal 101 is a terminal for inputting a moving image signal composed of a plurality of images.

  The block divider 102 divides an image to be encoded, which is a moving image signal input from the input terminal 101, into a plurality of regions. In the embodiment according to the present invention, the block is divided into 8 × 8 pixels, but may be divided into other block sizes or shapes.

  The prediction signal generator 103 is a part that generates a prediction signal for a target region (target block) to be encoded. Specific processing of the prediction signal generator 103 will be described later.

  The subtractor 105 uses the prediction signal generator 103 input via the line L103 from the pixel signal indicated by the target region obtained by the division by the block divider 102 input via the line L102. This is a part for generating a residual signal by subtracting the generated prediction signal. The subtractor 105 outputs the residual signal obtained by the subtraction to the converter 106 via the line L105.

  The converter 106 is a part that performs discrete cosine transform on the residual signal obtained by subtraction. The quantizer 107 is a part that quantizes the transform coefficient that has been discrete cosine transformed by the transformer 106. The entropy encoder 111 compresses and encodes the transform coefficient quantized by the quantizer 107, and outputs the compressed and compressed data via the line L111. The output terminal 112 outputs compressed data that is information input from the entropy encoder 111 to the outside.

  The inverse quantizer 108 inversely quantizes the quantized transform coefficient, and the inverse transformer 109 performs inverse discrete cosine transform to restore the encoded residual signal. The adder 110 adds the restored residual signal and the prediction signal sent from the line L103, reproduces the signal of the target block, obtains a reproduced image signal, and stores the reproduced image signal in the frame memory 104. Remember. In this embodiment, the converter 106 and the inverse converter 109 are used. However, other conversion processes may be used in place of these converters, and the converter 106 and the inverse converter 109 are not necessarily essential. Thus, in order to perform intra prediction or inter prediction for the subsequent target region, the compressed pixel signal of the target region is restored by inverse processing and stored in the frame memory 104.

  Next, the prediction signal generator 103 will be described. The prediction signal generator 103 is a part that generates a prediction signal for a target region (hereinafter referred to as a target block) to be encoded. In this embodiment, three types of prediction methods are used. That is, the prediction signal generator 103 generates a prediction signal using at least one or both of an inter-screen prediction method and an intra-screen prediction method described later.

  Processing of the prediction signal generator 103 in this embodiment will be described. FIG. 2 is a block diagram showing the prediction signal generator 103 used in the image predictive coding apparatus 100 according to the present embodiment. The template matching unit 201, the coordinate information memory 202, the candidate prediction signal combination determiner 203, and the prediction The area acquisition unit 204, the weighting unit 205, and the adder 206 are included.

  In the template matching unit 201, an image signal (reproduced image signal) that has already been reproduced in the past processing is input from the frame memory 104 via the line L104, and prediction signal candidates (candidate prediction signals) for the target pixel signal in the target region are obtained. A plurality of searches are performed by template matching described later, and coordinate information for accessing the searched candidate prediction signal is output to the coordinate information memory 202 via the line L201a. At the same time, difference value data indicating the relationship between the target region and each candidate prediction signal (corresponding to a sum of absolute error values (SAD (sumof absolute difference) described later)) is combined with a candidate prediction signal combination determiner 203 via a line L201b. Output to.

  The candidate prediction signal combination determiner 203 uses the difference value data input via the line L201b to set a plurality of candidate prediction signal combinations. Then, a candidate prediction signal combination is determined using the pixel signal input from the frame memory via the line L104 according to the coordinate information input from the coordinate information memory 202 via the line L202a, and the candidate prediction signal combination information is obtained. It outputs to the prediction area | region acquisition device 204 via the line L203.

  The prediction area acquisition unit 204 acquires coordinate information of candidate prediction signals belonging to this combination via line L202b according to the combination information of candidate prediction signals input via line L203. Then, the prediction region acquisition unit 204 acquires a candidate prediction signal corresponding to the acquired coordinate information from the frame memory 104 via the line L104 and outputs it to the weighting unit 205 as needed. The weighting unit 205 multiplies each candidate prediction signal input via the line L204 by a weighting coefficient, and outputs the result to the adder 206 via the line L205. The adder 206 sequentially adds the weighted candidate prediction signals and outputs the result as a prediction signal to the subtracter 105 in FIG. 1 via the line L103. As for the operation of the weighting unit, for example, there is a method of multiplying each candidate prediction signal by 1 / N when the number of candidate prediction signals is N, but other methods may be used.

  Furthermore, each configuration of the template matching unit 201, the candidate prediction signal combination determining unit 203, and the prediction region acquiring unit 204 will be described in detail. First, processing contents in the template matching unit 201 will be described. The template matching unit 201 accesses the reproduced image signal stored in the frame memory 104 via the line L104 and performs matching processing. Here, this matching process will be described. 3 and 4 are schematic diagrams for explaining the template matching process according to the embodiment of the present invention. Here, a process of generating a candidate prediction signal for the target block 402 will be described.

  First, a “target adjacent region” (also referred to as a template) and a “search region” are set for a target block by a predetermined method. In FIG. 3, a part (or all) of the reproduced images that are in contact with the target block 402 and are reproduced before and within the same screen are set as the search area 401. Here, the target block 402 is a small block of 4 × 4 pixels obtained by dividing an encoding target block made up of 8 × 8 pixels, but may be divided into other block sizes or shapes, or may not be divided. Also good.

  Further, in FIG. 4, a part of the reproduced image shown on the screen 411 different from the target block 402 is set as the search area 417. Further, a search area may be provided on each of a plurality of screens different from the target block 402. As the “target adjacent region”, the already reproduced pixel group (reverse L character region) 403 adjacent to the target block 402 is used. In this embodiment, the search area is set in two screens (FIGS. 3 and 4), but it may be set only in the same screen as the target block (FIG. 3). You may set only in a different screen (FIG. 4).

  As shown in FIG. 3, the search area 401 and the target block 402 do not need to touch each other, and the search area 401 may not touch the target block 402 at all. In addition, as shown in FIG. 4, it is not necessary to limit the search area to be set on one screen different from the target block (only the screen 411). Each of the search areas may be set in order to include future frames in order.

  In addition, the target adjacent area 403 may be in contact with the target block 402 even at least one pixel. Here, the shape of the target adjacent region is an inverted L character, but is not limited thereto. Therefore, if the target adjacent region is composed of reproduced pixels around the target block 402, the shape and the number of pixels may be arbitrary as long as they are determined in advance. The shape and size of the template in sequence units, frame units, or block units. (Number of pixels) may be encoded.

  The template matching unit 201 calculates the sum of absolute error values between corresponding pixels (SAD) between the search region 401 and the search region 417 or a pixel group having the same shape as the target adjacent region 403 in one of the search regions. Value), M areas having a small SAD value are searched, and these are set as “predicted adjacent areas”. The search accuracy may be in units of integer pixels, or a pixel with decimal precision such as 1/2 pixel or 1/4 pixel may be generated, and the search may be performed with decimal pixel precision. The value of M may be an arbitrary value as long as it is set in advance. As shown in FIGS. 3 and 4, M = 6, and the predicted adjacent regions 404a, 404b, 404c, 414a, 414b, and 414c are searched. Alternatively, the value of M may be determined by searching a region where the SAD value is smaller than a certain threshold without determining the number of predicted adjacent regions to be searched, or the smaller of the number of regions smaller than the threshold and the set number is M It is good also as the value of. At this time, the threshold value may be applied not to the SAD value itself but to a difference value between the minimum SAD value and the second and subsequent SAD values. In the latter case, the template matching unit 201 can search many predicted adjacent regions without changing the threshold value even when the minimum SAD value is large. The upper limit value and threshold value of M may be set in advance, but an appropriate value may be encoded in sequence units, frame units, or block units.

  The regions 405a, 405b, 405c, 415a, 415b, and 415c adjacent to the prediction adjacent regions 404a, 404b, 404c, 414a, 414b, and 414c are determined as the prediction regions for the target block 402, and the pixel signals in the prediction region are candidate predictions Determined to be a signal. Here, the positional relationship between the prediction adjacent region and the prediction region indicating the candidate prediction signal is the same as the target region and the target adjacent region, but this need not be the case. In the present embodiment, as the access information for acquiring each predicted adjacent area (and predicted area) from the frame memory 104, the difference coordinates between the target adjacent area (and target area) and the predicted adjacent area (and predicted area) The identification numbers of the screens (reference images) to which 406a, 406b, 406c, 416a, 416b and 416c and the respective prediction adjacent areas (and prediction areas) belong are stored in the coordinate information memory 202 as coordinate information.

  A configuration of the template matching unit 201 for performing the operation as described above will be described. The template matching unit 201 includes a target adjacent region acquisition unit 211, a predicted adjacent region acquisition unit 212, a comparator 213, and a switch 214. First, the target adjacent area acquisition unit 211 acquires the target adjacent area 403 from the frame memory 104 via the line L104.

  The predicted adjacent area acquisition unit 212 acquires a pixel signal of an area having the same shape as the target adjacent area from the search area in the frame memory 104 via the line L104, and is obtained from the target adjacent area acquisition unit 211 via the line L211. The SAD value is calculated between the pixel signal (adjacent pixel signal) in the target adjacent region. The comparator 213 inputs the calculated SAD value via the line L212b and compares it with the Mth smallest SAD value among the SAD values acquired so far. When the comparator 213 determines that the input SAD value is smaller, the comparator 213 temporarily stores the SAD value that is within the Mth, and the MADth SAD value. Erase. The comparator 213 sets a sufficiently large value as an initial value at the start of processing as an initial value of the SAD value as compared with a normal SAD value.

  In addition to performing this process, the predicted adjacent area acquisition unit 212 obtains coordinate information as information for accessing the predicted adjacent area (and the predicted area) via the line L212a by controlling the switch 214 by the comparator 213. To 202. At this time, the coordinate information with the M + 1th SAD value becomes unnecessary, and may be overwritten and stored in the coordinate information memory 202. When the search within the search region by the prediction adjacent region acquisition unit 212 is completed, the comparator 213 outputs the temporarily stored M SAD values to the candidate prediction signal combination determiner 203 via the line L201b. To do.

  Next, the operation of the candidate prediction signal combination determiner 203 will be described. The candidate prediction signal combination determiner 203 includes a combination setting unit 231, a prediction adjacent region acquisition unit 232, a target adjacent region acquisition unit 233, a weighting unit 234, an adder 235, and a comparison / selection unit 236.

  The combination setting unit 231 sets a combination of a plurality of predicted adjacent areas based on the M SAD values input via the line L201b. As will be described in detail in the description of FIG. 6 with respect to the processing in the predictive adjacent region combination determiner 203, the N predictive adjacent regions having small input SAD values are set as one combination. The value of N is a multiplier of 2 smaller than M, and when M = 6, three combinations of N = 1, 2, and 4 are created.

  Thus, by using the candidate prediction signal combination processing of the present invention, it is possible to determine the smoothing strength of the candidate prediction signal appropriate for each target block, that is, the number of prediction candidate signals to be averaged. It becomes. In other words, the difference SAD value between the target signal and the adjacent pixel signal obtained by averaging the pixel signals of the N predicted adjacent regions having a small difference SAD value that is an absolute error value between the adjacent pixel signals is the smallest N. Can be selected from the M candidate prediction signals searched for, without additional information, candidate prediction signals suitable for generating a prediction signal. The reason why the value of N is a multiplier of 2 is that it is considered that the signal averaging process is performed only by addition and shift calculation.

  Further, the combination of the M value, the N value, and the N predicted signal regions is not limited to this. The number of predicted adjacent regions included in one combination can be arbitrarily set from 1 to M. For example, when creating a combination composed of N predicted adjacent regions smaller than M, it is possible to select N from M and set all combinations. At this time, the value of N may be fixed, or a combination may be set by selecting two or more of 1 to M. However, in order to automatically select the same prediction adjacent region combination in the image predictive encoding device 100 which is an encoder and the image predictive decoding device 300 which will be described later as a decoder, the setting method of the combination of both is made to match. It is necessary to keep.

  When one piece of prediction adjacent region combination information is input via the line L231a, the prediction adjacent region acquisition unit 232 acquires coordinate information for the prediction adjacent region included in the combination from the coordinate information memory 202 via the line L202a. . Then, the predicted adjacent area acquisition unit 232 acquires the pixel signal of the predicted adjacent area corresponding to the coordinate information via the line L104 and outputs it to the weighting unit 234 as needed.

  The weighter 234 multiplies the pixel signal of each prediction adjacent region input via the line L232 and outputs the result to the adder 235 via the line L234.

  The adder 235 generates a comparison signal to be compared with the pixel signal (adjacent pixel signal) of the target adjacent region by accumulating and accumulating the weighted pixel signals of the predicted adjacent region, and performing the generated comparison. The signal is output to the comparator / selector 236 via the line L235. Note that the operation of the weighter 234 includes, for example, a method of multiplying the pixel signal of each prediction adjacent region by 1 / N when the number of prediction adjacent regions is N, but other methods may be used. For example, a method of calculating a difference value (absolute error sum, square error sum, variance, etc.) between pixel signals of N predicted adjacent regions and adjacent pixel signals and determining a weighting coefficient for each predicted adjacent region based on the ratio Can be considered.

  Here, by making the weighting methods of the weighting unit 205 and the weighting unit 234 the same, an appropriate combination of candidate prediction signals (pixel signals of the prediction region) can be estimated using the prediction adjacent region. Since the predictive adjacent region can be shared by the encoder and the decoder, this is a decoder using this technique. There is an effect that a combination of prediction signals can be acquired without additional information. Note that the weighting methods of the weighter 205 and the weighter 234 are not necessarily the same. For example, in order to reduce the amount of calculation, it is also effective to apply a simple weighting process to the weighter 234 and to apply an adaptive weighting process according to local signal characteristics to the weighter 205. is there.

  The target adjacent area acquisition unit 233 acquires the pixel signal (adjacent pixel signal) of the target adjacent area from the frame memory 104 via the line L104 and outputs the acquired pixel signal to the comparison / selection unit 236.

  The comparator / selector 236 calculates the SAD value between the comparison signal corresponding to the combination of a plurality of prediction adjacent regions and the adjacent pixel signal, and selects the combination of the target adjacent regions having the minimum value as a combination of candidate prediction signals. Choose as. The selected combination of candidate prediction signals is output to the prediction region acquisition unit 204 via line L203 as candidate prediction signal combination information.

  As described above, according to the present embodiment, for each target block, it is possible to select a combination of candidate prediction signals effective for prediction signal generation from a plurality of candidate prediction signals without additional information.

  FIG. 5 shows a search for a plurality (M) of candidate prediction signals for a pixel signal (target pixel signal) in the target region (target block) by the template matching unit 201 according to the present embodiment, and accesses the searched candidate prediction signals. It is a flowchart which shows the method of acquiring the coordinate information for doing. First, the target adjacent area acquisition unit 211 acquires the target adjacent area (template signal) for the target block from the frame memory 104 (S502).

  Next, the template matching unit 201 initializes a threshold value for selecting M candidate prediction signals to a sufficiently large value (S503). The predicted adjacent area acquisition unit 212 obtains a pixel group having the same shape as the target adjacent area and the target adjacent area in the search area and the sum of absolute differences (SAD) (S504). The SAD is compared with the threshold value by the comparator 213 in the template matching unit 201 (S505). If it is determined that the SAD value is smaller than the threshold value, the process proceeds to S506. If not, the process proceeds to S508.

  The comparator 213 in the template matching unit 201 compares the obtained SAD value with the previous SAD value, and when the obtained SAD value is included in order from the Mth, the searched pixel group is used as a candidate prediction signal ( And the candidate signal is updated. In this embodiment, space-time coordinate information (spatial position and searched pixels) for accessing a candidate prediction signal (and a pixel signal of a prediction adjacent region) instead of the candidate prediction signal (and a pixel signal of the prediction adjacent region) The coordinate information is updated by storing the identification number of the screen including the group in the coordinate information memory 202 based on the switching control by the switch 214 (S506). At the same time, the template matching unit 201 updates the threshold value to the Mth smallest SAD value (S507).

  Thereafter, the predicted adjacent area acquisition unit 212 confirms whether or not all search areas have been searched (S508). When it is determined that not all have been searched, the process returns to S504, and the predicted adjacent area acquisition unit 212 and the next pixel group having the same shape as the target adjacent area and the target adjacent area in the search area and the sum of absolute differences (SAD) Is required.

  When all the search has been completed, the template matching process for one target block ends (S509).

  In this way, the processing by the template matching unit 201 can identify the top M predicted adjacent areas including pixel signals having a high correlation with the pixel signal in the target adjacent area.

  Next, processing of the candidate prediction signal combination determiner 203 will be described with reference to the drawings. FIG. 6 selects a combination of N candidate prediction signals suitable for generating a prediction signal by smoothing (weighted average) a plurality of candidate prediction signals in the candidate prediction signal combination determiner 203 according to the present embodiment. 3 is a flowchart illustrating a method. First, the combination setter 231 of the combination determiner 203 sets the number N of candidate prediction signals to 1 (S602). Next, the target adjacent area acquisition unit 233 acquires the target adjacent area (template signal) for the target block from the frame memory 104 (S603).

  Then, the predicted adjacent area acquisition unit 232 acquires N predicted adjacent areas belonging to the combination set by the combination setting unit 231 via the line L104. That is, the predicted adjacent area acquisition unit 232 coordinates the coordinate information corresponding to N predicted adjacent areas having a small SAD value that is a difference value between the pixel signal of the target block and the adjacent pixel signal of the target adjacent area for the target block. Obtained from the information memory 202. Then, N predicted adjacent areas corresponding to the acquired coordinate information are acquired from the frame memory 104 (S604).

  Thereafter, the weighting unit 234 and the adder 235 generate a comparison signal by averaging the pixel signals of the N predicted adjacent regions (may be weighted average) (S605), and the comparison / selector 236 generates the comparison signal. An SAD value that is a difference value between the comparison signal and the adjacent pixel signal is calculated (S606). At the same time, the comparator / selector 236 compares the calculated SAD value with the minimum SAD value so far (S607), and if the SAD value is determined to be the minimum value, the process proceeds to S608, and if not, , The process proceeds to S609. In S607, if the calculated SAD value and the minimum SAD value so far are the same, the process proceeds to S609, but may proceed to S608.

  When the calculated SAD value is the minimum SAD value so far, the combination of coordinate information acquired in S604 is stored in the comparator / selector 236 (S608).

  Here, the combination determiner 203 updates the value of N twice (S609). Then, the updated N is compared with the magnitude of M (S610). If the updated N value is smaller than M, the process returns to S604. If the updated N value is larger than M, the coordinate information stored in S608 is stored. Is determined as a candidate prediction signal combination, and the candidate prediction signal combination selection process is terminated (S611).

  As described above, by using the candidate prediction signal combination processing of this embodiment, it is possible to determine the smoothing strength of the candidate prediction signal appropriate for each target block, that is, the number of prediction candidate signals to be averaged. It becomes possible. In other words, N having the smallest difference SAD value between the target signal and the neighboring pixel signal obtained by averaging the pixel signals of the N predicted neighboring regions having a small SAD value that is a difference value between the neighboring pixel signals is determined. Thus, a candidate prediction signal suitable for generation of a prediction signal can be selected from the searched M candidate prediction signals without additional information.

  The reason why the value of N is a multiplier of 2 is that it is considered that the signal averaging process is performed only by addition and shift calculation. In the present embodiment, the value of N is not limited to a multiplier of 2. Moreover, the method of setting the combination of a some prediction adjacent area | region is not limited to the method of FIG. The number of predicted adjacent regions included in one combination can be arbitrarily set from 1 to M. For example, when creating a combination composed of N predicted adjacent regions smaller than M, it is possible to set all combinations of selecting N from M. At this time, the value of N may be fixed, or a combination may be set by selecting two or more of 1 to M. In addition, in order to automatically select the same combination of prediction adjacent regions in the encoder and the decoder, it is necessary to match the setting method of the combination of both.

  FIG. 7 is a flowchart showing a prediction signal generation method in smoothing (weighted average) of a plurality of candidate prediction signals according to the present embodiment.

  The prediction region acquisition unit 204 acquires one or more candidate prediction signals corresponding to the target block from the frame memory 104 according to the selected coordinate information (S702). Then, the weighting unit 205 and the adder 206 perform weighted averaging of the obtained candidate prediction signals to generate a prediction signal for the target block (S703). This completes the process for one target block (S704).

  FIG. 8 is a flowchart showing an image predictive encoding method in the image predictive encoding device 100 according to the present embodiment. First, the prediction signal generator 103 of FIG. 2 generates a prediction signal of the target block (S102). Next, a residual signal indicating a difference between the signal of the target block and the prediction signal of the target block is encoded by the transformer 106, the quantizer 107, and the entropy encoder 111 (S103). The encoded residual signal is output via the output terminal 112 (S105).

  Thereafter, the encoded residual signal is decoded by an inverse quantizer 108 and an inverse transformer 109 in order to predictively encode a subsequent target block. Then, the prediction signal is added to the decoded residual signal by the adder 110, and the signal of the target block is reproduced and stored as a reference image in the frame memory 104 (S106). If all the target blocks have not been processed, the process returns to S102, the process for the next target block is performed, and if it has been completed, the process ends (S107).

  As described above, in the image predictive coding device 100 of the present embodiment, a prediction signal smoothed using a plurality of prediction signals can be obtained without using additional information.

  Next, an image predictive decoding method according to this embodiment will be described. FIG. 9 is a block diagram showing an image predictive decoding apparatus 300 according to this embodiment. The image predictive decoding apparatus 300 includes an input terminal 301, an entropy decoder 302 (data decoding means), an inverse quantizer 303 (residual signal restoration means), an inverse transformer 304 (residual signal restoration means), and an adder 305. (Reproduced image signal generation means), output terminal 306, frame memory 307, prediction signal generator 308 (prediction signal generation means). As the residual signal restoration means by the inverse quantizer 303 and the inverse transformer 304, other means may be used. Further, the inverse converter 304 may not be provided. Each configuration will be described below.

  The input terminal 301 inputs compressed data that has been compression-encoded by the above-described image predictive encoding method. The compressed data includes a residual signal obtained by predicting and encoding a target block obtained by dividing an image into a plurality of blocks.

  The entropy decoder 302 decodes the encoded data of the residual signal of the target block by entropy decoding the compressed data input at the input terminal 301.

  The inverse quantizer 303 inputs the residual signal of the decoded target block via the line L302 and performs inverse quantization. The inverse transformer 304 performs inverse discrete cosine transform on the inversely quantized data. Inverse quantizer 303 and inverse transformer 304 output signals obtained by inverse quantization and inverse discrete cosine transform, respectively, as difference signals (reproduced residual signals).

  The prediction signal generator 308 basically has the same configuration as that shown in FIG. 2 or a function corresponding thereto, and generates a prediction signal by the same processing as the prediction signal generator 103 of FIG. Since the prediction signal generator 308 generates a prediction signal only from the reproduced signal stored in the frame memory 307, by managing the input data to the frame memory 307 and the frame memory 104 in FIG. The same prediction signal as the prediction signal generator 103 in FIG. 1 can be generated. Details of the configuration of the prediction signal generator 308 are omitted because they have already been described with reference to FIG. The prediction signal generator 308 operating in this way outputs the generated prediction signal to the adder 305 via the line L308.

  The adder 305 adds the prediction signal generated by the prediction signal generator 308 to the difference signal (reproduction residual signal) restored by the inverse quantizer 303 and the inverse transformer 304, and reproduces the reproduced image of the target block. The signal is output to the output terminal 306 and the frame memory 307 via the line L305. The output terminal 306 outputs to the outside (for example, a display).

  The frame memory 307 stores the reproduction image output from the adder 305 as a reference image as a reference reproduction image for the next decoding process. At this time, the reproduced image is managed by the same method as that of the image predictive coding apparatus 100 of FIG.

  Next, the image predictive decoding method in the image predictive decoding apparatus 300 according to the present embodiment will be described with reference to FIG. First, the compressed compressed data is input via the input terminal 301 (S302). Then, the entropy decoder 302 performs entropy decoding on the compressed data and extracts quantized transform coefficients (S303). Here, a prediction signal is generated by the prediction signal generator 308 (S304). The process of S304 is basically the same as the process S102 of FIG. 8, and the processes of FIGS.

  On the other hand, the quantized transform coefficient is inversely quantized using the quantization parameter in the inverse quantizer 303, and inverse transform is performed in the inverse transformer 304, thereby generating a reproduction difference signal (S305). Then, the generated prediction signal and the reproduction difference signal are added to generate a reproduction signal, and the reproduction signal is stored in the frame memory to reproduce the next target block (S306). If there is next compressed data, this process is repeated again (S307), and all data is processed to the end (S308). If necessary, the process may return to S302 and take in compressed data.

  The image predictive encoding method and the image predictive decoding method according to the present embodiment can be provided by being stored in a recording medium as a program. Examples of the recording medium include a recording medium such as a floppy disk (registered trademark), CD-ROM, DVD, or ROM, or a semiconductor memory.

  FIG. 11 is a block diagram illustrating modules of a program that can execute the image predictive coding method. The image predictive coding program P100 includes a block division module P102, a prediction signal generation module P103, a storage module P104, a subtraction module P105, a transformation module P106, a quantization module P107, an inverse quantization module P108, an inverse transformation module P109, and an addition module P110. The entropy encoding module P111 is included. As shown in FIG. 12, the prediction signal generation module P103 includes a template matching module P201, a candidate prediction signal combination determination module P202, and a prediction signal synthesis module P203.

  The functions realized by executing the modules are the same as those of the constituent elements of the image predictive coding apparatus 100 described above. That is, the function of each module of the image predictive coding program P100 includes a block divider 102, a prediction signal generator 103, a frame memory 104, a subtractor 105, a converter 106, a quantizer 107, an inverse quantizer 108, and an inverse quantizer. The functions of the converter 109, the adder 110, and the entropy encoder 111 are the same, and the functions of each module of the prediction signal generation module P103 are the template matching unit 201, the candidate prediction signal combination determiner 203, and signal synthesis. This is similar to the functions of the prediction region acquisition unit 204 to the addition unit 206.

  FIG. 13 is a block diagram showing modules of a program that can execute the image predictive decoding method. The image predictive decoding program P300 includes an entropy decoding module P302, an inverse quantization module P303, an inverse transform module P304, an addition module P305, a storage module P307, and a prediction signal generation module P308.

  The functions realized by executing the modules are the same as those of the components of the image predictive decoding apparatus 300 described above. That is, the function of each module of the image predictive decoding program P300 is the same as the functions of the entropy decoder 302, the inverse quantizer 303, the inverse transformer 304, the adder 305, and the frame memory 307. The prediction signal generation module P308 has a function equivalent to that of the prediction signal generation module P103 in the image prediction encoding program P100, and includes a template matching unit 201, a candidate prediction signal combination determination unit 203, and signal synthesis. Functions of the prediction region acquisition unit 204 to the adder 206 are provided.

  The image predictive encoding program P100 or the image predictive decoding program P300 configured as described above is stored in the recording medium 10 and executed by a computer described later.

  FIG. 16 is a diagram illustrating a hardware configuration of a computer for executing a program recorded in a recording medium, and FIG. 17 is a perspective view of the computer for executing a program stored in the recording medium. Examples of the computer include a DVD player, a set-top box, a mobile phone, and the like that have a CPU and perform processing and control by software.

  As shown in FIG. 16, a computer 30 includes a reading device 12 such as a floppy disk drive device, a CD-ROM drive device, a DVD drive device, a working memory (RAM) 14 in which an operating system is resident, and a recording medium 10. A memory 16 for storing the program stored in the memory, a display device 18 such as a display, a mouse 20 and a keyboard 22 as input devices, a communication device 24 for transmitting and receiving data and the like, and a CPU 26 for controlling execution of the program. And. When the recording medium 10 is inserted into the reading device 12, the computer 30 can access the image predictive coding program P100 or the image predictive decoding program P300 stored in the recording medium 10 from the reading device 12, and image predictive coding. The program P100 or the image predictive decoding program P300 can operate as the image predictive encoding device 100 or the image predictive decoding device 300 of the present embodiment.

  As shown in FIG. 17, the image predictive coding program P100 or the image predictive decoding program P300 may be provided via a network as a computer data signal 40 superimposed on a carrier wave. In this case, the computer 30 stores the image predictive encoding program P100 or the image predictive decoding program P300 received by the communication device 24 in the memory 16, and executes the image predictive encoding program P100 or the image predictive decoding program P300. it can.

  The embodiment described so far may be modified as follows. In the candidate prediction signal combination determiner 203 of FIG. 2, the SAD value (absolute error) that is a difference value between the pixel signal of the target adjacent region of the target block and the target signal obtained by weighted averaging the pixel signals of the plurality of adjacent prediction regions. (Sum of values) is calculated to determine the optimal combination of candidate prediction signals, but the combination is determined using the sum of squared errors (SSD) and variance (VAR) of the difference signal instead of the SAD value. Is possible. In the three evaluation criteria, the calculation amount increases in the order of SAD, SSD, and VAR. On the other hand, the evaluation accuracy is improved, and the effect that the code amount of the error signal can be reduced is expected.

  It is also effective to determine the final combination using SSD or VAR when a plurality of combinations having the same SAD value as the difference value between the target adjacent signal and the target signal are obtained. That is, when the SAD value calculated by the predicted adjacent region acquisition unit 232 matches the minimum value calculated so far, the comparator / selector 236 further selects either SSD or VAR as a comparison target. Compare if the value is small. Here, the combination determined to have a small SSD or VAR is stored in the comparator / selector 236 as a combination having the minimum value. In this case, the comparator / selector 236 calculates SSD or VAR together with SAD, and temporarily stores it.

  Furthermore, it is also possible to use a method for calculating the variance of combinations of a plurality of candidate prediction signals instead of the pixel signals in the adjacent prediction region and using them for determining the combination. Specifically, the prediction adjacent region acquisition unit 232, the weighting unit 234, and the adder 235 in FIG. 2 can be realized by replacing with the prediction region acquisition unit 204, the weighting unit 205, and the adder 206, respectively. In this modified example, the processing of the target adjacent region acquisition unit 233 is not necessary, and a prediction signal can be output from the comparison / selection unit 236 to the subtracter 105 in FIG. 1, so that the circuit scale can be reduced. There is an effect.

  The comparator 213 of the template matching unit 201 also uses the SAD value, which is a difference value, for evaluating the pixel group having the same shape as the target adjacent area of the target block and the searched target adjacent area. However, the SSD or VAR is changed. The same effect as the case of the candidate prediction signal combination determiner 203 may be expected.

  The prediction signal generator 103 in FIG. 1 is not limited to the configuration in FIG. For example, in FIG. 2, coordinate information for accessing a plurality of candidate prediction signals searched by the template matcher is stored in the coordinate information memory 202. A configuration in which pixel signals are stored may be used. Although the amount of memory in FIG. 2 increases, there is an effect of reducing access to the frame memory 104. The template matching unit 201 outputs the SAD value, which is the difference value between the target adjacent region and the M predicted adjacent regions, to the candidate prediction signal combination determiner 203 via the line L201b. When the SAD value that is the difference value is not used for the setting, the line L201b is unnecessary.

  In the present embodiment, the target adjacent area is composed of reproduced pixels, but may be composed of prediction signals of adjacent blocks. This is effective when it is desired to make the prediction area smaller than the coding block unit of the error signal, or when it is desired to generate the signal of the target adjacent area without performing the encoding / decoding process.

  This embodiment can be applied to a method of searching for an indefinite number of candidate signals and selecting an appropriate number of candidate signals by template matching.

  Next, the operational effects of the image predictive encoding device 100 and the image predictive decoding device 300 of this embodiment will be described.

  In the image predictive coding device 100 according to the present embodiment, the template matching unit 201 has a high correlation with the target adjacent region 403 that is composed of the already reproduced adjacent pixel signals that are adjacent to the target region (target block) 402 that is composed of the target pixel signal. A plurality of predicted adjacent areas 404a to 404c and 414a to 414c are searched from search areas 401 and 417 made of already-reproduced images. The combination setting unit 231 in the candidate prediction signal combination determiner 203 derives two or more combinations of arbitrary predicted adjacent areas including at least one of the searched plurality of predicted adjacent areas 404a to 404c and 414a to 414c. Then, the prediction adjacent region acquisition unit 232 extracts the pixel signal of the derived prediction adjacent region, and the weighter 234 and the adder 235 use, for example, an average of the extracted pixel signals using a predetermined synthesis method. As a result, the comparison signal for the adjacent pixel signal is generated for each combination. Then, the comparison / selector 236 selects a combination having a high correlation between the comparison signal generated by the weighting unit 234 and the like and the adjacent pixel signal acquired by the target adjacent region acquisition unit 233. The prediction region acquisition unit 204 generates one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and the weighter 205 and the adder 206 determine the candidate prediction signal in advance. A prediction signal is generated by processing using a synthesis method. The subtractor 105 subtracts the prediction signal generated in this way from the target pixel signal acquired via the block divider 102 to generate a residual signal, and the converter 106, the quantizer 107, and the entropy coding The unit 111 encodes the generated residual signal.

  As a result, a combination of candidate prediction signals suitable for smoothing can be selected without increasing the amount of information using a target adjacent region composed of already-reproduced adjacent pixel signals adjacent to the target block. A prediction signal in consideration of noise characteristics can be generated.

  Further, the comparator / selector 236 in the image predictive encoding device 100 of the present embodiment selects a combination having a small SAD value that is a sum of absolute values of differences between the comparison signal and the adjacent pixel signal, thereby further smoothing. A suitable combination of candidate prediction signals can be selected.

  In addition, the weighter 234 and the adder 235 in the image predictive encoding device 100 according to the present embodiment generate a comparison signal by performing weighted averaging of the pixel signals in the prediction adjacent region belonging to the combination set in the combination setting unit 231. Thus, it is possible to generate an appropriate comparison signal for selecting a combination of candidate prediction signals more suitable for smoothing.

  In addition, the combination of prediction adjacent regions set by the combination setting unit 231 in the image predictive encoding device 100 of the present embodiment includes 2 n predicted adjacent regions in descending order of correlation with the target adjacent region. As a result, it can be performed only by addition and shift operation, and a simple configuration can be achieved. Here, the value of n is preferably an integer of 0 or more.

  Further, in the image predictive decoding device 300 of the present embodiment, the template matching unit 201 has a correlation with the target adjacent region 403 including the already reproduced adjacent pixel signal adjacent to the target region (target block) 402 including the target pixel signal. A plurality of high predicted adjacent areas 404a to 404c and 414a to 414c are searched from the search areas 401 and 417 made of the already reproduced images. The combination setting unit 231 in the candidate prediction signal combination determiner 203 derives two or more combinations of arbitrary predicted adjacent areas including at least one of the searched plurality of predicted adjacent areas 404a to 404c and 414a to 414c. Then, the prediction adjacent region acquisition unit 232 extracts the pixel signal of the derived prediction adjacent region, and the weighter 234 and the adder 235 use, for example, an average of the extracted pixel signals using a predetermined synthesis method. As a result, the comparison signal for the adjacent pixel signal is generated for each combination. Then, the comparison / selector 236 selects a combination having a high correlation between the comparison signal generated by the weighting unit 234 and the like and the adjacent pixel signal acquired by the target adjacent region acquisition unit 233. The prediction region acquisition unit 204 generates one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and the weighter 205 and the adder 206 determine the candidate prediction signal in advance. A prediction signal is generated by processing using a synthesis method.

  Then, the entropy decoder 302, the inverse quantizer 303, and the inverse transformer 304 restore the difference signal from the compressed data input through the input terminal 301, and the adder 305 includes the predicted signal generated as described above. The restored difference signal is added to generate a reproduced image signal.

  As a result, a combination of candidate prediction signals suitable for smoothing can be selected without increasing the amount of information using a target adjacent region composed of already-reproduced adjacent pixel signals adjacent to the target block. A prediction signal in consideration of noise characteristics can be generated.

  Further, the comparator / selector 236 in the image predictive decoding device 300 of the present embodiment is more suitable for smoothing by selecting a combination having a small SAD value that is the sum of absolute values of differences between the comparison signal and the adjacent pixel signal. A combination of candidate prediction signals can be selected.

  In addition, the weighting unit 234 and the adder 235 in the image predictive decoding device 300 according to the present embodiment generate a comparison signal by performing weighted averaging of the pixel signals of the prediction adjacent regions belonging to the combination set by the combination setting unit 231. Thus, it is possible to generate an appropriate comparison signal for selecting a combination of candidate prediction signals more suitable for smoothing.

  In addition, the combination of prediction adjacent regions set by the combination setting unit 231 in the image predictive decoding device 300 of the present embodiment includes 2 n predicted adjacent regions in descending order of correlation with the target adjacent region. Therefore, it can be performed only by addition and shift operation, and a simple configuration in mounting can be achieved. Here, the value of n is preferably an integer of 0 or more.

  Further, in the image predictive coding program P100 of the present embodiment, the template matching module P201 correlates with the target adjacent region 403 including the already reproduced adjacent pixel signal adjacent to the target region (target block) 402 including the target pixel signal. A plurality of predicted adjacent regions 404a to 404c and 414a to 414c having a high value are searched from the search regions 401 and 417 made of the already reproduced images. The candidate prediction signal combination determination module P202 derives two or more combinations of arbitrary predicted adjacent areas including at least one of the searched plurality of predicted adjacent areas 404a to 404c and 414a to 414c. Then, the candidate prediction signal combination determination module P202 extracts the pixel signals of the derived prediction adjacent regions, and processes the extracted pixel signals by, for example, averaging using a predetermined synthesis method. The comparison signal for the adjacent pixel signal is generated for each combination. Then, the predicted signal synthesis module P203 selects a combination having a high correlation between the generated comparison signal and the acquired adjacent pixel signal. The candidate prediction signal combination determination module P202 generates one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and processes the candidate prediction signals using a predetermined synthesis method. To generate a prediction signal. The subtraction module P105 subtracts the prediction signal generated in this way from the target pixel signal acquired via the block division module P102 to generate a residual signal, and transform module P106, quantization module P107, and entropy coding Module P111 encodes the generated residual signal.

  Further, in the image predictive decoding program P300 of the present embodiment, the template matching module P201 has a correlation with the target adjacent region 403 composed of the already reproduced adjacent pixel signal adjacent to the target region (target block) 402 composed of the target pixel signal. A plurality of high predicted adjacent areas 404a to 404c and 414a to 414c are searched from the search areas 401 and 417 made of the already reproduced images. The predicted signal synthesis module P203 derives two or more combinations of arbitrary predicted adjacent areas including at least one of the searched plurality of predicted adjacent areas 404a to 404c and 414a to 414c. Then, the prediction signal synthesis module P203 extracts the pixel signal of the derived prediction adjacent region, and processes the extracted pixel signal by, for example, averaging using a predetermined synthesis method, for each combination. The comparison signals for the adjacent pixel signals are respectively generated. Then, the predicted signal synthesis module P203 selects a combination having a high correlation between the generated comparison signal and the acquired adjacent pixel signal. The prediction signal synthesis module P203 generates one or more candidate prediction signals of the target pixel signal based on the prediction adjacent regions belonging to the selected combination, and the prediction signal synthesis module P203 combines the candidate prediction signals in advance. Is used to generate a prediction signal.

Then, the entropy decoding module P302, the inverse quantization module P303, and the inverse transform module P304 restore the difference signal from the input compressed data, and the addition module P305 outputs the prediction signal generated as described above and the restored difference signal. Addition to generate a reproduction image signal.

<Second Embodiment>

  Next, regarding the candidate prediction signal combination determiner 203 (see FIG. 2) and the candidate prediction signal combination determination method (see FIG. 6), another candidate prediction signal to be smoothed can be given diversity. An embodiment is shown. These embodiments are the same as the prediction signal generator 103 of the first embodiment, except for the candidate prediction signal combination determiner 203 (see FIG. 2) and the candidate prediction signal combination determination method (see FIG. 6). Since it is a structure, description is omitted. The invention according to the present embodiment is recorded in an image predictive encoding device, an image predictive decoding device, an image predictive encoding method, an image predictive decoding method, an image predictive decoding method, and a recording medium. It can be applied to the hardware configuration of a computer for executing the program. Also, the functions or methods indicated by the same numbers in FIGS. 2 and 6 are the same operations and processes as described above, and hence the description thereof will be omitted. In addition, the present embodiment can be applied to the image predictive encoding device 100 or the image predictive decoding device 300 described above.

  FIG. 18 is a block diagram illustrating a configuration of the prediction signal generator 103a according to the second embodiment. The prediction signal generator 103a includes a template matching unit 201, a coordinate information memory 202, a candidate prediction signal combination determination unit 203b, a prediction region acquisition unit 204, a weighting unit 205, and an adder 206. The configuration of the candidate prediction signal combination determiner 203b is different from that of the prediction signal generator 103 in the first embodiment. That is, the candidate predictive signal combination determiner 203 in FIG. 2 calculates M SAD values (absolute value error sums between the target adjacent area and each of the M predicted adjacent areas) input via L201b. The combination of predictive adjacent regions is set using the candidate predictive signal combination determiner 203b in FIG. 18, and is input via M 202 SAD values input via L201b and L202c. The combination of predicted adjacent areas is set using the coordinate information of the predicted adjacent areas.

  More specifically, the configuration of the combination setter 237 in FIG. 18 is different from the configuration of the combination setter 231 in FIG. 2, and the candidate predictive signal combination determiner 203b in FIG. In addition to the configuration of the prediction signal combination determiner 203, it includes a setting unit 230 for the number of predicted adjacent regions.

  The predictive adjacent region number setting unit 230 is a part that determines candidates for the number of predictive adjacent regions to be smoothed from the number M of predictive adjacent regions searched by the template matching unit 201. That is, the predictive adjacent region number setting unit 230 inputs the number M of predictive adjacent regions searched by the template matching unit 201 via the line L201b. Then, the predictive adjacent region number setting unit 230 calculates one or more multipliers N smaller than the M value, and sequentially outputs the values to the combination setting unit 237 via the line L230 while changing the value of N. For example, when M is 6, N = 1, 2, 4 are output, and when M is 30, N = 1, 2, 4, 8, 16 are output.

  The combination setter 237 uses the N value input via the line L230, the M SAD values input via the line L201b, and the coordinate information of the M predicted adjacent regions input via the line L202c. , N selected adjacent regions. Then, the selection information (information that can identify the selected prediction adjacent region) is used as combination information of the prediction adjacent region, and a plurality of combinations of N prediction adjacent regions are sequentially output to the prediction adjacent region acquisition unit 232 via the line L237. To do.

  In this way, in the combination predictor 203b for candidate prediction signals, the combination setter 237 generates combination information of prediction adjacent regions for different N values, and selects one combination information according to the procedure described in FIG.

  FIG. 19 is a block diagram showing a detailed configuration of the combination setting unit 237. The combination setting unit 237 includes a prediction adjacent region determination unit 281 to be evaluated, a counter 282, a reference image evaluation unit 283, and a motion vector evaluation unit 284. The combination setter 237 uses the coordinate information of the M predicted adjacent areas searched by the template matching unit 201 for the input value N, selects N predicted adjacent areas, and selects the selected information as the predicted adjacent Output as area combination information. In this embodiment, coordinate information, that is, a reference image identification number (reference image number), and spatial difference coordinates in the screen between the target adjacent region and the predicted adjacent region (hereinafter referred to as a motion vector). Is used to select a combination of predicted adjacent areas composed of N predicted adjacent areas, but the data used for the selection is not limited to the reference image number and the motion vector. This coordinate information may be information relating to the generation of each candidate prediction signal and uniquely determining the prediction adjacent region, and may be information for accessing the prediction adjacent region. For example, when one prediction method is determined from a plurality of predetermined prediction methods using pixel signals in the prediction adjacent region, the prediction method may be included in the data used for selection.

  Before selecting N predicted adjacent areas from the M predicted adjacent areas searched by the template matching unit 201, the predictive adjacent area determining unit 281 to be evaluated determines the number of predicted adjacent areas to be selected. This is a part limited to R pieces. This R value is a numerical value determined by counting the number of SADs that are equal to or greater than a predetermined threshold among the M predicted adjacent areas as will be described later. This determination of the R value has an effect of not selecting a predicted adjacent region having a large SAD value with respect to the target adjacent region when the N value is very small with respect to the M value.

  The determination target predictive region determiner 281 includes M SAD values corresponding to the M predictive adjacent regions searched by the template matching unit 201 via the line L201b (the target adjacent region and the M predictive adjacent regions). (Difference data between regions) is input, and simultaneously the number N of predicted adjacent regions to be selected via the line L230 is input. The predictive adjacent region determiner 281 to be evaluated compares a predetermined threshold value with respect to the inputted N value and M SAD values, and counts the number of SAD values smaller than the threshold value. Then, the number of SAD values smaller than the threshold value counted in this way is output to the counter 282 via the line L281 as the number R of predicted adjacent regions to be evaluated. However, when the R value is smaller than or equal to the N value, it is not necessary to select a prediction adjacent region, and thus the predictive adjacent region determination unit 281 to be evaluated has N (R) prediction adjacent regions. Is output to the predicted adjacent region acquisition unit 232 via the line L237. Note that, as described above, the determination of the R value has an effect of not selecting a predicted adjacent region having a large difference from the target adjacent region when the N value is too small with respect to the M value. Therefore, as long as this condition is satisfied, means other than the threshold may be used to determine the R value. For example, the 2 × N value and the M value may be compared, and the smaller value may be output as the R value. If the R value determined by the threshold processing is larger than the 2 × N value, the R value is set to 2 ×. You may make it restrict | limit to N value. The threshold value may be encoded instead of being determined in advance.

  When the number R of predicted adjacent regions to be evaluated input via the line L281 is input to the counter 282, the counter 282 updates the counter value P from 1 to R in increments of 1, while the lines L282a and L282b are updated. Via the reference image evaluator 283 and the motion vector evaluator 284.

The reference image evaluator 283 is a part that selects N reference image numbers from R reference image numbers in consideration of the diversity. First, when the P value is input via the line L282a, the reference image evaluator 283 sets the reference image number for accessing the predicted adjacent region where the SAD value between the target adjacent region is the Pth smallest. Obtained from the coordinate information memory 202 via L202c. After obtaining the R reference image numbers, the reference image evaluator 283 determines the number of predicted adjacent regions Ref [i] (i = 1, 2,..., F; i) included in each reference image in the search region. Is the reference image number, and F is the number of reference images to be searched by the template matching unit 201. Next, the reference image evaluator 283 performs the following operation with the reference image number i to which the predicted adjacent region having the smallest SAD value between the target adjacent region belongs as an initial value.
1) Select reference image number i.
2) Decrease the number of predicted adjacent regions Ref [i] by one. Reference image numbers for which the number of predicted adjacent regions Ref [i] is 0 are excluded from selection targets.
3) Increase the reference image number i by 1. When i is F, i is reset to 1.
This operation is repeated cyclically until N reference images are selected. The selected N reference image numbers are output to the motion vector evaluator 284 via the line L283.

  For example, in the conceptual diagram illustrating the relationship between the reference image 450 and the number of predicted adjacent areas Ref [i] illustrated in FIG. 26, a predicted adjacent area is formed for each reference image. For the reference image number 1, it is shown that three predicted adjacent regions 404i are formed, and in this case, Ref [1] = 3. By repeating the above-described processing 1) to processing 3), reference image numbers up to the top N Ref [i] values are acquired.

  Note that the reference image selection method is not limited to the method of counting up the identification number, as long as the diversity of predicted adjacent regions can be ensured. For example, the reference image may be selected in consideration of the ratio Ref [i] / R of Ref [i] corresponding to the reference image number i. In addition, the position of the reference image with respect to the encoding target image (past, future, present) may be taken into consideration and selected at a predetermined ratio.

When N reference image numbers are input from the reference image evaluator 283, the motion vector evaluator 284 selects N motion vectors from the R motion vectors in consideration of the diversity of the motion vectors. It is a part to do. The R motion vectors correspond to the R predicted adjacent regions described above. When the P value is input via the line L282b, the motion vector evaluator 284 receives the reference image number and the motion vector information for accessing the predicted adjacent region where the SAD value between the target adjacent region is the Pth smallest. Obtained via line L202c. After acquiring the R motion vectors, the motion vector evaluator 284 confirms the motion vector SMV [i] corresponding to the predicted adjacent region having the smallest SAD value with respect to the target adjacent region for each reference image. . Next, the motion vector evaluator 284 associates the motion vector with the N reference image numbers input via the line L283. In particular,
4) Classify R motion vectors by reference image number, and calculate the number of motion vectors NumMV [i] to be selected.
5) Next, SMV [i] with reference image number i with NumMV [i] greater than 1 is selected.
6) Then, for each reference image number i, “NumMV [i] −1” motion vectors are selected in descending order of the sum of absolute value differences from SMV [i].
Information on the N predicted adjacent areas corresponding to the set of the motion vector and the reference image number selected in this way is output to the predicted adjacent area acquisition unit 232 via the line L237 as prediction adjacent area combination information.

The processing concept of the processing 4) to 6) regarding the motion vector is shown in FIG. According to FIG. 27, each reference image includes a plurality of predicted adjacent areas, and a motion vector also exists corresponding thereto. For example, in the example of FIG. 27, it is assumed that motion vectors MV1 and MV2 exist in the reference image number 1. In this case, NumMV [1] = 2. For example, when the motion vector MV1 is smaller than MV2, the motion vector MV1 is selected as SMV [1].
Further, since the number of motion vectors NumMV [3] = 0, in the process 2), the reference image number 3 is not selected, and other reference image numbers 1, 2, 4, and 5 are selected.

  Note that the motion vector selection method is not limited to the method described above because it is sufficient to ensure diversity. For example, by updating SMV [i] with the motion vector of the selected reference image i, a motion vector having a large absolute value difference sum from the motion vector selected immediately before may be selected.

  Further, as shown in FIG. 19, by using coordinate information (reference image number and motion vector) for accessing each prediction adjacent region, it is possible to select a prediction adjacent region having diversity in the spatio-temporal direction. It becomes. In the above description, the combination setting unit 237 selects N predicted adjacent areas from the R predicted adjacent areas, but it is not necessary to select all N using the coordinate information. For example, for N / 2, a method of selecting a predicted adjacent area having a small SAD value between the target adjacent areas is also conceivable, and in that case, an effect of stabilizing the prediction performance can be expected. In addition, if the sum of absolute difference of motion vectors from SMV [i] is limited and larger than a predetermined threshold, it is considered that means for removing from the selection target may stabilize the prediction performance. The configuration in the combination setting device 237 is not limited to FIG. For example, when the R value is the M value, or when the R value is preset with respect to the N value, the predictive adjacent region determiner 281 to be evaluated is not necessarily required. In this case, the prediction adjacent region can be selected only by the coordinate information, that is, the motion parameter.

  As described above, the combination setting unit 237 capable of selecting a combination of predicted adjacent regions in consideration of the diversity illustrated in FIG. 19 is used by using the relationship between the target adjacent region and the predicted adjacent region illustrated in FIG. Candidates to be smoothed in consideration of space-time diversity by using together with candidate predictor combination determiner 203b (predictive signal generator 103a) that can select the number of candidate predictive signals to be smoothed It becomes possible to perform the selection operation of the number of prediction signals. The configuration in the candidate prediction signal combination determiner 203b is not limited to the configuration shown in FIG. For example, the setting unit 230 for the predicted number of adjacent regions may be included in the combination setting unit 237.

  Next, a candidate prediction signal combination determination method, which is the operation of the prediction signal generator 103a shown in FIGS. 18 and 19, will be described. FIG. 20 is a flowchart showing the operation of the prediction signal generator 103a using the combination setter 237, and shows a candidate prediction signal combination determination method for the purpose of giving diversity to the candidate prediction signals to be smoothed. It is a thing. In FIG. 20, unlike FIG. 6, the process (S602, S609) for setting the number N of candidate prediction signals is performed by the setting unit 230 for the number of predicted adjacent areas.

  First, the combination setting unit 237 of the combination determining unit 203 sets the number N of candidate prediction signals to 1 (S602). Next, the target adjacent area acquisition unit 233 acquires the target adjacent area (template signal) for the target block from the frame memory 104 (S603).

  On the other hand, the combination setting unit 237 uses the coordinate information (reference image number and motion vector) for accessing the predicted adjacent area and the difference SAD value between the target adjacent area and the predicted adjacent area, so that N predicted adjacent areas are used. Is selected (S604a). The subsequent processing is the same as in FIG.

  That is, the prediction adjacent region acquisition unit 232, the weighter 234, and the adder 235 acquire pixel signals of N prediction adjacent regions, and generate a comparison signal by averaging (may be weighted average) (S605). The comparator / selector 236 calculates a SAD value that is a difference value between the generated comparison signal and the adjacent pixel signal (S606). At the same time, the comparator / selector 236 compares the calculated SAD value with the minimum SAD value so far (S607), and if the SAD value is determined to be the minimum value, the process proceeds to S608, and if not, , The process proceeds to S609. In S607, if the calculated SAD value and the minimum SAD value so far are the same, the process proceeds to S609, but may proceed to S608.

  When the calculated SAD value is the minimum SAD value so far, the combination of coordinate information acquired in S604 is stored in the comparator / selector 236 (S608).

  Here, the combination setter 237 updates the N value by a factor of 2 (S609). Then, the magnitudes of the updated N value and the M value are compared (S610). If the updated N value is smaller than the M value, the process returns to S604, and if the updated N value is larger than the M value, the process is stored in S608. The combination of coordinate information thus determined is determined as a combination of candidate prediction signals, and the candidate prediction signal combination selection process is terminated (S611).

  Next, detailed processing of the above-described processing S604a will be described. FIG. 21 is a flowchart showing the operation of the combination setter 237 when N predicted adjacent areas are selected using the coordinate information of the predicted adjacent areas. In this process, for the input value N, N predictions are made using the coordinate information of the M predicted adjacent areas searched in the process (see FIG. 5) for searching for candidate prediction signals by the template matching unit 201. An adjacent region is selected.

  Note that the combination setter 237 in this embodiment selects a combination of N predicted adjacent regions using coordinate information, that is, a reference image number and a motion vector, and the data used for the selection is a reference image number and a reference image number. It is not limited to a motion vector. Any information that is related to the generation of each candidate prediction signal and that is uniquely determined by using the predicted adjacent area may be included in the data used for selection, and may be information for accessing the predicted adjacent area. For example, when one prediction method is determined from a plurality of predetermined prediction methods using signals in the prediction adjacent region, the prediction method may be included in the data used for selection.

  First, in the predictive adjacent region determination unit 281 to be evaluated, before selecting N predicted adjacent regions from the M predicted adjacent regions searched by the template matching unit 201, the prediction adjacent region to be selected is selected. The number is limited by setting the number to R (S802). This setting of the R value has an effect of preventing the prediction adjacent region having a large SAD value between the target adjacent region from being selected when the N value is very small with respect to the M value. Therefore, as long as this condition is satisfied, a method other than the threshold may be used as the R value determination method. For example, 2 × N and M may be compared and the smaller one may be output as R. If R calculated by the threshold processing is greater than 2 × N, R is limited to 2 × N. It may be. The threshold value may be encoded instead of being determined in advance.

  Next, the predictive adjacent region determiner 281 to be evaluated compares the R value with the N value. If the R value is smaller than or equal to the N value, the prediction adjacent region need not be selected. The adjacent area selection process is terminated (S803: NO). If the R value is greater than the N value, the counter 282 initializes the count value P (S804). The reference image evaluator 283 and the motion vector evaluator 284 store the reference image number and the motion vector for accessing the predicted adjacent region having the SAD value between the target adjacent region and the Pth smallest memory for the coordinate information. Sequentially from 202 (S805). The P value and the R value are compared (S806). If the P value is smaller than the R value (S806: YES), 1 is added to the count value P (S807), and the process of S805 is repeated.

When the P value reaches the R value (S805: NO), the reference image evaluator 283 selects N reference image numbers from the R reference image numbers in consideration of the diversity (S808). . Specifically, first, for the obtained R reference image numbers, the number Ref [i] (i = 1, 2,..., F; i) of predicted adjacent regions included in each reference image in the search region. Is the reference image number, and F is the number of reference images to be searched by the template matching unit 201. Next, the reference image evaluator 283 performs the following processing with the reference image number i to which the predicted adjacent region having the smallest SAD value from the target adjacent region belongs as an initial value.
1) Select reference image number i.
2) Decrease the number of predicted adjacent areas Ref [i] by one. Reference image numbers for which Ref [i] is 0 are excluded from selection targets.
3) Increase the reference image number i by 1. When i is F, i is reset to 1.
This process is repeated cyclically until N reference images are selected. Note that the selection method of the reference image is not limited to the method of counting up the identification number, as long as diversity can be ensured. For example, the reference image may be selected in consideration of the ratio Ref [i] / R of the number of predicted adjacent areas Ref [i] corresponding to the reference image number i. In addition, the position of the reference image with respect to the encoding target image (past, future, present) may be taken into consideration and selected at a predetermined ratio.

Next, the motion vector evaluator 284 selects N motion vectors from the R motion vectors in consideration of the diversity (S809). Specifically, first, the motion vector evaluator 284 confirms the motion vector SMV [i] having the smallest SAD value with respect to the target adjacent region for each reference image. Next, the motion vector evaluator 284 associates the motion vector with the N reference image numbers selected in S808. In particular,
4) Classify R motion vectors by reference image number, and calculate the number of motion vectors NumMV [i] to be selected.
5) Select SMV [i] with reference image number i with NumMV [i] greater than 1.
6) Then, for each reference image number i, “NumMV [i] −1” motion vectors are selected in descending order of the sum of absolute value differences from SMV [i]. Note that the motion vector selection method is not limited to the method described above because it is sufficient to ensure diversity. For example, by updating SMV [i] with the motion vector of the selected reference image i, a motion vector having a large absolute value difference sum from the motion vector selected immediately before may be selected.

  In this way, in addition to the difference data between the target adjacent area and the plurality of predicted adjacent areas, by using the motion parameter for accessing each predicted adjacent area, the predicted adjacent areas having diversity in the spatio-temporal direction can be obtained. It becomes possible to select. In the above description, R to N predicted adjacent regions are selected, but it is not necessary to select all N using motion parameters. For example, for N / 2, the method of selecting a predicted adjacent region having a small SAD value between the target adjacent regions can also be expected to stabilize the prediction performance. In addition, if the sum of absolute difference of motion vectors from SMV [i] is limited and is larger than a predetermined threshold, it is considered that the method of removing from the selection target also has the effect of stabilizing the prediction performance. Note that if the R value is the M value, or if the R value is set in advance for the N value, the process S803 is unnecessary, and in this case, the prediction adjacent region selection process is performed using coordinate information, That is, it can be implemented with only motion parameters.

  Further, the “process for selecting a combination of predicted adjacent regions in consideration of diversity” shown in FIG. 21 is replaced with “candidate prediction signal smoothed using the relationship between the target adjacent region and the predicted adjacent region” shown in FIG. By using this together with the “number selection process”, it is possible to perform the selection process of the number of candidate prediction signals to be smoothed in consideration of the diversity of time and space.

  The image predictive encoding method and the image predictive decoding method according to the present embodiment can be provided by being stored in a recording medium as a program. Examples of the recording medium include a recording medium such as a floppy disk (registered trademark), CD-ROM, DVD, or ROM, or a semiconductor memory. The specific module configuration is as shown in FIG. 11, FIG. 12, and FIG. Here, the function of the prediction signal generation module 103 corresponds to the prediction signal generator 103a. Further, the combination setting corresponding to the combination setting unit 237 is as shown in FIG. FIG. 28 is a block diagram showing the module configuration of the combination setting module P237. This combination setting module P237 includes a prediction adjacent region determination module to be evaluated, P281, a counter module P282, a reference image evaluation module P283, and a motion vector evaluation module P284. Each of these modules has functions equivalent to those of the predictive adjacent region determining unit 281, the counter 282, the reference image evaluating unit 283, and the motion vector evaluating unit 284 in the combination setting unit 237 shown in FIG. 19.

  The image predictive encoding program P100 or the image predictive decoding program P300 thus configured is stored in the recording medium 10 and executed by the computer shown in FIGS. 16 and 17 described above.

  Next, operational effects of the image predictive encoding device 100 and the image predictive decoding device 300 in the second embodiment will be described.

  The image predictive coding device 100 according to the second embodiment performs a block divider 102 that divides an input image into a plurality of regions, and a target pixel signal of a target region that is a processing target among the divided regions. A prediction signal generator 103 that generates a prediction signal, a subtractor 106 that functions as a residual signal generation unit that generates a residual signal between the generated prediction signal and a target pixel signal, and a code of the generated residual signal A converter 106, a quantizer 107, and an entropy encoder 111.

  Then, in the prediction signal generator 103a, the template matching unit 201 has a plurality of prediction adjacent regions 404a-414c having high correlation with the target adjacent region 403 composed of the already reproduced adjacent pixel signals adjacent to the target region 402 composed of the target pixel signal. Are searched from the search areas 401 and 417 made up of already reproduced images. Then, the combination setting unit 237 generates two or more combinations of predicted adjacent areas including different numbers of predicted adjacent areas using at least coordinate information indicating the positions of the predicted adjacent areas.

  The prediction adjacent region acquisition unit 232, the weighting unit 234, and the adder 235 each generate a comparison signal for the adjacent pixel signal by processing the pixel signal of the prediction adjacent region belonging to the combination using a predetermined synthesis method. . The comparison / selector 236 selects a combination of predicted adjacent regions having a high correlation between the comparison signal and the adjacent pixel signal, and outputs the combination to the predicted region acquisition unit 204.

  The prediction region acquisition unit 204 generates one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and the weighter 205 and the adder 206 determine the candidate prediction signal in advance. A prediction signal is generated by processing using a synthesis method.

  Thereby, it becomes possible to give diversity to the candidate prediction signal to be smoothed, and it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, encoded data can be generated efficiently.

  In addition, the image predictive decoding device 300 according to the second embodiment includes an entropy decoder 302 that decodes encoded data of a residual signal related to a target region that is a processing target in compressed data, and a signal obtained by decoding. An inverse quantizer 303 and an inverse transformer 304 for restoring the reproduction residual signal from the prediction signal generator 308, a prediction signal generator 308 for generating a prediction signal for the target pixel signal in the target region, and the generated prediction signal and the restored reproduction residual An adder 305 that generates a reproduced image signal by adding the difference signal is provided.

  Then, the prediction signal generator 103a generates a plurality of prediction adjacent regions 404a to 414c having a high correlation with the target adjacent region 403 including the already reproduced adjacent pixel signal adjacent to the target region 401 including the target pixel signal. Search is performed from search areas 401 and 417 consisting of. Then, the combination setting unit 237 generates two or more combinations of predicted adjacent areas including different numbers of predicted adjacent areas using at least coordinate information indicating the positions of the predicted adjacent areas.

  The prediction adjacent region acquisition unit 232, the weighting unit 234, and the adder 235 each generate a comparison signal for the adjacent pixel signal by processing the pixel signal of the prediction adjacent region belonging to the combination using a predetermined synthesis method. . The comparison / selector 236 selects a combination of predicted adjacent regions having a high correlation between the comparison signal and the adjacent pixel signal, and outputs the combination to the predicted region acquisition unit 204.

  The prediction region acquisition unit 204 generates one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and the weighter 205 and the adder 206 determine the candidate prediction signal in advance. A prediction signal is generated by processing using a synthesis method.

Thereby, it becomes possible to give diversity to the candidate prediction signal to be smoothed, and it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, the encoded data encoded efficiently can be decoded.

<Third Embodiment>
As described above, in FIG. 18 to FIG. 21, “a process of selecting a combination of predicted adjacent areas in consideration of diversity” is performed using “a candidate predicted signal to be smoothed using the relationship between the target adjacent area and the predicted adjacent area”. It was used in conjunction with the “number selection process”. However, “a process of selecting a combination of predicted adjacent areas in consideration of diversity” is not combined with “a process of selecting the number of candidate prediction signals to be smoothed using the relationship between the target adjacent area and the predicted adjacent area”. Effective in processing. The embodiment will be described below.

  In the present embodiment, after setting the number FN of candidate prediction signals to be smoothed, the combination of candidate prediction signals is determined using “processing for selecting a combination of prediction adjacent regions in consideration of diversity”. be able to. That is, the operation of the combination setter 237 shown in FIG. 19 and the process of step S604a shown in FIG. 21 are performed only on the determined FN value, and the combination setter 237 determines from the M predicted adjacent regions. FN predicted adjacent regions are selected. Then, weighted averaging processing of FN candidate prediction signals is performed based on the combination of the selected FN prediction adjacent regions.

  FIG. 22 and FIG. 23 are embodiments in the case of selecting the number of candidate prediction signals to be smoothed using the relationship between the target adjacent region and the predicted adjacent region. Hereinafter, the difference between the prediction signal generator 103a in the second embodiment and the prediction signal generator 103b in the present embodiment will be described. Of course, the prediction signal generator 103b of the present embodiment can be applied to the image predictive coding device 100 and the image predictive decoding device 300 described in the first embodiment.

  FIG. 22 is a block diagram illustrating a configuration of the prediction signal generator 103b including the candidate prediction signal combination setting unit 237 according to the third embodiment. Unlike the candidate prediction signal combination determiner 203b in FIG. 18, the candidate prediction signal combination determiner 203c in FIG. 22 receives a plurality of different N values (N predicted adjacent regions) from the predictor adjacent region number setting unit 230. ) Is directly input to the prediction adjacent region acquisition unit 232 without passing through the combination setting unit 237.

  Then, as described with reference to FIG. 2, the predictive adjacent region number setting unit 230 changes the N value (that is, changes the multiplier N of 2 smaller than the M value), and supplies the predictive adjacent region acquisition unit 232 with N The predicted adjacent area acquisition unit 232, the weighting unit 234, and the adder 235 output N values using a plurality of N values (N predicted adjacent areas), and N predicted adjacent areas having a small SAD value with respect to the target adjacent area And a plurality of comparison signals are generated for each combination of N predicted adjacent regions. The comparator / selector 236 compares the plurality of comparison signals having different N values with the target adjacent region, determines the N value with the smallest difference as the FN value, and outputs this to the combination setting unit 237. The comparator / selector 236 performs the above-described processing while changing the N value until it exceeds the M value, and updates the minimum SAD value.

  The combination setter 237 uses the determined FN values, the M SAD values input via the line L201b, and the coordinate information of the predicted adjacent regions input via the line L202c, from the M predicted adjacent regions to FN. Select the predicted adjacent regions. That is, the combination setter 237 counts the number of SAD values that are smaller than the threshold value determined in association with the FN value, and selects FN predicted adjacent regions based on the counted value. This detailed process is as shown in FIG.

  Then, the combination setting unit 237 outputs the combination of the prediction adjacent regions as the combination information of the candidate prediction signal to the prediction region acquisition unit 204 via the line L203. The operation in the combination setter 237 is as described above with reference to FIG.

  Then, the prediction region acquisition unit 204 acquires the coordinate information of candidate prediction signals belonging to this combination via the line L202b according to the combination information of candidate prediction signals input via the line L203. Then, the prediction region acquisition unit 204 acquires a candidate prediction signal corresponding to the acquired coordinate information from the frame memory 104 via the line L104 and outputs it to the weighting unit 205 as needed. The weighting unit 205 multiplies each candidate prediction signal input via the line L204 by a weighting coefficient, and outputs the result to the adder 206 via the line L205. The adder 206 sequentially adds the weighted candidate prediction signals and outputs the result as a prediction signal to the subtracter 105 in FIG. 1 via the line L103.

  FIG. 23 is a flowchart illustrating processing when a prediction adjacent region is selected by the prediction signal generator 103b according to the third embodiment. In this method, similarly to the process shown in FIG. 6, the predictive adjacent region number setting unit 230 sets the number N of candidate prediction signals to 1 (S602), and the target adjacent region acquiring unit 233 sets the target for the target block. An adjacent region (template signal) is acquired from the frame memory 104 (S603), and N predicted adjacent regions belonging to the combination set by the predictive adjacent region number setting unit 230 are Obtained via L104 (S604).

  After that, in the prediction adjacent region acquisition unit 232, the weighter 234, and the adder 235, a comparison signal is generated by averaging (may be weighted average) the pixel signals of the N prediction adjacent regions (S605), and the comparison / selector In S236, an SAD value that is a difference value between the generated comparison signal and the adjacent pixel signal is calculated (S606). At the same time, the comparator / selector 236 compares the calculated SAD value with the minimum SAD value so far (S607). If the SAD value is determined to be the minimum value, the process proceeds to S608a, and if not, , The process proceeds to S609. In S607, when the calculated SAD value is the same as the minimum SAD value so far, the process proceeds to S609, but may proceed to S608a.

  When the SAD value calculated in S606 by the comparator / selector 236 is determined to be smaller than the minimum value of SAD calculated so far, the minimum value is updated to the newly calculated SAD value, and Is determined as the number FN of candidate prediction signals (S608a).

  Then, the N value is updated twice (S609). If the updated N value is smaller than the M value, the process returns to S604. If the updated N value is larger than the M value, the combination of coordinate information determined in S608a is selected. It is determined as a combination of candidate prediction signals. Then, the combination setting unit 237 uses the coordinate information (reference image number and motion vector) for accessing the predicted adjacent area and the difference SAD value between the target adjacent area and the predicted adjacent area, so that M predicted adjacent areas are used. FN prediction adjacent regions are selected from the region, and the reproduction signal of the FN prediction adjacent regions is set as a candidate prediction signal for the target region (S610a). Since the process of S610a is the same as S604a (see FIG. 21), the description is omitted.

  The image predictive encoding method and the image predictive decoding method according to the present embodiment can be provided by being stored in a recording medium as a program. Examples of the recording medium include a recording medium such as a floppy disk (registered trademark), CD-ROM, DVD, or ROM, or a semiconductor memory. The specific module configuration is as shown in FIG. 11, FIG. 12, FIG. 13, and FIG. Here, the function of the prediction signal generation module 103 corresponds to the above-described prediction signal generator 103b. The image predictive encoding program P100 or the image predictive decoding program P300 thus configured is stored in the recording medium 10 and executed by the computer shown in FIGS. 16 and 17 described above.

  Next, operational effects of the image predictive coding device 100 and the image predictive decoding device 300 according to the third embodiment will be described.

  The image predictive coding apparatus 100 according to the third embodiment performs a block divider 102 that divides an input image into a plurality of regions, and a target pixel signal of a target region that is a processing target among the divided regions. A prediction signal generator 103 that generates a prediction signal, a subtractor 106 that functions as a residual signal generation unit that generates a residual signal between the generated prediction signal and a target pixel signal, and a code of the generated residual signal A converter 106, a quantizer 107, and an entropy encoder 111.

  Then, the template matching unit 201 in the prediction signal generator 103 selects a plurality of prediction adjacent regions 404a to 414c having a high correlation with the target adjacent region 403 composed of the already reproduced adjacent pixel signals adjacent to the target region 402 composed of the target pixel signal. The search is performed from search areas 401 and 411 made up of already reproduced images.

  On the other hand, the predictor adjacent region number setting unit 230 sets the number N of predictive adjacent regions to be selected.

  Then, the combination setting unit 237 selects the number of predicted adjacent areas set from the search area using at least coordinate information indicating the position of the predicted adjacent area. The prediction region acquisition unit 204 generates a set number of candidate prediction signals of the target pixel signal based on the selected prediction adjacent region, and the weighting unit 205 and the adder 206 use a synthesis method in which candidate prediction signals are determined in advance. The prediction signal is generated by processing.

  In the present embodiment, a comparison signal is generated by the prediction adjacent region acquisition unit 232, the weighting unit 234, and the adder 235 based on the number N of prediction adjacent regions set by the prediction adjacent region 230. Here, the value of the N value is appropriately changed and input to the predicted adjacent region acquisition unit 232, and a comparison signal is generated according to the N value. Based on the comparison signal, the comparison / selector 236 selects a comparison signal that minimizes the difference SAD value from the pixel signal in the target adjacent region, and N predicted adjacent regions from which the selected comparison signal is based. N value for specifying is output to the combination setter 237.

  Thereby, it becomes possible to give diversity to the candidate prediction signal to be smoothed, and it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, encoded data can be generated efficiently.

  The image predictive decoding device 300 according to the third embodiment also includes an entropy decoder 302 that decodes encoded data of a residual signal related to a target region that is a processing target in compressed data, and a signal obtained by decoding. An inverse quantizer 303 and an inverse transformer 304 for restoring the reproduction residual signal from the prediction signal generator 308, a prediction signal generator 308 for generating a prediction signal for the target pixel signal in the target region, and the generated prediction signal and the restored reproduction residual An adder 305 that generates a reproduced image signal by adding the difference signal is provided.

  Then, the template matching unit 201 in the prediction signal generator 103 selects a plurality of prediction adjacent regions 404a to 414c having a high correlation with the target adjacent region 403 composed of the already reproduced adjacent pixel signals adjacent to the target region 402 composed of the target pixel signal. The search is performed from search areas 401 and 411 made up of already reproduced images.

  On the other hand, the predictor adjacent region number setting unit 230 sets the number N of predictive adjacent regions to be selected.

  Then, the combination setting unit 237 selects the number of predicted adjacent areas set from the search area using at least coordinate information indicating the position of the predicted adjacent area. The prediction region acquisition unit 204 generates a set number of candidate prediction signals of the target pixel signal based on the selected prediction adjacent region, and the weighting unit 205 and the adder 206 use a synthesis method in which candidate prediction signals are determined in advance. The prediction signal is generated by processing.

  In the third embodiment, a comparison signal is generated by the prediction adjacent region acquisition unit 232, the weighting unit 234, and the adder 235 based on the number N of prediction adjacent regions set by the prediction adjacent region 230. . Here, the value of the N value is appropriately changed and input to the predicted adjacent region acquisition unit 232, and a comparison signal is generated according to the N value. Based on the comparison signal, the comparison / selector 236 selects a comparison signal that minimizes the difference SAD value from the pixel signal in the target adjacent region, and N predicted adjacent regions from which the selected comparison signal is based. N value for specifying is output to the combination setter 237.

Thereby, it becomes possible to give diversity to the candidate prediction signal to be smoothed, and it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, the encoded data encoded efficiently using this prediction signal can be decoded.

<Fourth Embodiment>
FIG. 24 is a block diagram illustrating a configuration of a candidate prediction signal combination determiner according to the fourth embodiment. In the fourth embodiment, the number of candidate prediction signals to be smoothed is determined by threshold processing using the SAD value between the target adjacent region and the predicted adjacent region. In the prediction signal generator 103c in the fourth embodiment, the candidate prediction signal combination determiner 203d is different from the other prediction signal generators 103, 103a, or 103b. Hereinafter, the candidate prediction signal combination determiner 203d as a difference will be described.

  The candidate prediction signal combination determiner 203d in FIG. 24 includes a candidate prediction signal number determiner 238 and a combination setter 237d.

  The candidate prediction signal number determiner 238 inputs M SAD values between the target adjacent region and each of the M predicted adjacent regions via the line L201b, and calculates the number of SAD values smaller than a predetermined threshold. calculate. Then, the number of SAD values smaller than the threshold is output to the combination setting unit 237d as the number FN of candidate prediction signals to be smoothed.

  The combination setter 237d has the configuration shown in FIG. 19, and uses the coordinate information of the predicted adjacent areas to select FN predicted adjacent areas from the M predicted adjacent areas, and the combination of predicted adjacent areas Is output to the prediction region acquisition unit 204 via L203 as combination information of candidate prediction signals. More specifically, when it is determined that the FN value is smaller than the M value, the combination setting unit 237d uses the coordinate information of the predicted adjacent area to FN predicted adjacent areas. Are selected, and the reproduction signals of the FN prediction areas adjacent to them are used as candidate prediction signals for the target area. When the FN value is the same as the M value, the selection process is not performed and a combination of FN predicted adjacent regions is output.

  Next, the operation of the candidate prediction signal combination determiner 203d will be described. FIG. 25 is a flowchart showing the operation of the candidate prediction signal combination determiner 203d.

  In the candidate prediction signal determiner 238, the SAD value between the target adjacent region and each of the M predicted adjacent regions is compared with a predetermined threshold value, and the number of SAD values smaller than the threshold value is smoothed. The number FN of candidate prediction signals is determined (S615). Then, the FN value and the M value are compared in the combination setter 237 (S616). When the combination setting unit 237 determines that the FN value is smaller than the M value, FN predicted adjacent areas are selected from the M predicted adjacent areas using the coordinate information of the predicted adjacent areas, and these The reproduction signal of the FN prediction areas adjacent to is used as a candidate prediction signal of the target area (S617).

  The process S617 here is the same process as the process S610a in FIG. If the M value and the F value are the same (S616: NO), the selection process is not performed, and the combination of M prediction adjacent regions is output to the prediction region acquisition unit 204 as it is, and a candidate prediction signal determiner is determined. The process at 238 ends.

  As described above with reference to FIG. 18 to FIG. 25, “the operation and processing for generating a combination of prediction adjacent regions that are diversified in the spatio-temporal direction using the coordinate information and motion parameters” is the specified number of candidate predictions. The present invention can be applied to all apparatuses and methods for selecting a combination of signals or predicted adjacent regions.

  Furthermore, in the above, N reference frames and N motion vectors are individually evaluated and selected, but the present invention is not limited to this evaluation method. The P-th predicted adjacent signal is represented by λ * (“P-1” th selected reference image number−evaluation target reference image number) + (“P−1” th selected motion vector−evaluation target. You may make it evaluate simultaneously using evaluation functions ((lambda) is a conversion coefficient) like the sum of absolute difference of a motion vector.

  Furthermore, since the output data of the combination setter 231 in FIG. 2 and the output data of the candidate predictor combination determiner shown in FIG. 18, FIG. 22, and FIG. The combination setting unit 231 can be configured to include the functions shown in FIGS. 18, 22, and 24. That is, the combination setting unit 231 in FIG. 2 and the combination setting unit 237 (237d) in FIGS. 18, 22 and 24 generate combination information of two or more predicted adjacent regions to be output, and the candidates in FIG. The prediction signal combination determiner 203 may be configured to select one.

  The image predictive encoding method and the image predictive decoding method according to the present embodiment can be provided by being stored in a recording medium as a program. Examples of the recording medium include a recording medium such as a floppy disk (registered trademark), CD-ROM, DVD, or ROM, or a semiconductor memory. The specific module configuration is as shown in FIG. 11, FIG. 12, FIG. 13, and FIG. Here, the function of the prediction signal generation module 103 corresponds to the prediction signal generator 103c. The image predictive encoding program P100 or the image predictive decoding program P300 thus configured is stored in the recording medium 10 and executed by the computer shown in FIGS. 16 and 17 described above.

  Next, operational effects of the image predictive coding device 100 and the image predictive decoding device 300 according to the fourth embodiment will be described. The image predictive coding device 100 according to the fourth embodiment includes a block divider 102 that divides an input image into a plurality of regions, and a target pixel signal in a target region that is a processing target among the divided regions. A prediction signal generator 103 that generates a prediction signal, a subtractor 106 that functions as a residual signal generation unit that generates a residual signal between the generated prediction signal and a target pixel signal, and a code of the generated residual signal A converter 106, a quantizer 107, and an entropy encoder 111.

  Then, in the predicted signal generator 103c, the template matching unit 201 has a plurality of predicted adjacent areas 404a-414c that have a high correlation with the target adjacent area 403 composed of the already reproduced adjacent pixel signals adjacent to the target area 402 composed of the target pixel signal. Are searched from search areas 401 and 417 made up of already reproduced images. Then, the combination setting unit 237 selects one or more predicted adjacent areas from the plurality of searched predicted adjacent areas using at least coordinate information indicating the position of the predicted adjacent area, for example, the reference image number and the motion vector.

  The prediction region acquisition unit 204 generates one or more candidate prediction signals of the target pixel signal based on the selected prediction adjacent region, and the weighter 205 and the adder 206 use a synthesis method in which the candidate prediction signals are determined in advance. The prediction signal is generated by processing.

  In the fourth embodiment, the number N of predicted adjacent areas is set by the determiner 238 for the number of candidate prediction signals, and the combination setter 237 selects the set number of predicted adjacent areas. .

  Thereby, it becomes possible to give diversity to the candidate prediction signal to be smoothed, and it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, encoded data can be generated efficiently.

  Further, the image predictive decoding device 300 according to the fourth embodiment includes an entropy decoder 302 that decodes encoded data of a residual signal related to a target region that is a processing target in compressed data, and a signal obtained by decoding. An inverse quantizer 303 and an inverse transformer 304 for restoring the reproduction residual signal from the prediction signal generator 308, a prediction signal generator 308 for generating a prediction signal for the target pixel signal in the target region, and the generated prediction signal and the restored reproduction residual An adder 305 that generates a reproduced image signal by adding the difference signal is provided.

  Then, in the prediction signal generator 103, the template matching unit 201 has a plurality of prediction adjacent regions 404a-414c having a high correlation with the target adjacent region 403 composed of the already reproduced adjacent pixel signals adjacent to the target region 402 composed of the target pixel signal. Are searched from search areas 401 and 417 made up of already reproduced images. Then, the combination setting unit 237 selects one or more predicted adjacent areas from the plurality of searched predicted adjacent areas using at least coordinate information indicating the position of the predicted adjacent area, for example, the reference image number and the motion vector.

  The prediction region acquisition unit 204 generates one or more candidate prediction signals of the target pixel signal based on the selected prediction adjacent region, and the weighter 205 and the adder 206 use a synthesis method in which the candidate prediction signals are determined in advance. The prediction signal is generated by processing.

  Thereby, it becomes possible to give diversity to the candidate prediction signal to be smoothed, and it is possible to efficiently generate a prediction signal in consideration of local noise characteristics. Therefore, the encoded data encoded efficiently using this prediction signal can be decoded.

It is a block diagram which shows the image predictive coding apparatus by embodiment of this invention. It is a block diagram which shows the prediction signal generator 103 used for an image predictive coding apparatus. FIG. 10 is a schematic diagram for explaining template matching processing and processing for searching for a prediction adjacent region and a prediction region candidate in the template matching unit 201. FIG. 11 is a second schematic diagram for explaining template matching processing and processing for searching for a prediction adjacent region and a prediction region candidate in the template matching unit 201. 10 is a flowchart for explaining a template matching in template matching unit 201 and a method of searching for a prediction adjacent region and a prediction region candidate. It is a flowchart for demonstrating the combination determination method of the candidate prediction signal in the combination determination unit 203 of a candidate prediction signal. It is a flowchart for demonstrating the method of synthesize | combining a candidate prediction signal and producing | generating a prediction signal. 3 is a flowchart illustrating an image predictive encoding method in the image predictive encoding device 100. It is a block diagram which shows the image prediction decoding apparatus. It is a flowchart which shows the image prediction decoding method of the image prediction decoding apparatus 300. It is a block diagram which shows the module of the program which can perform the image predictive coding method. It is a block diagram which shows the module of the prediction signal generation module P103. It is a block diagram which shows the module of the program which can perform the image prediction decoding method. It is a schematic diagram for demonstrating the prediction method in a screen. It is a schematic diagram for demonstrating a block matching process. It is a figure which shows the hardware constitutions of the computer for performing the program recorded on the recording medium. It is a perspective view of a computer for executing a program stored in a recording medium. It is a block diagram which shows the structure of the prediction signal generator 103a of 2nd Embodiment. 4 is a block diagram showing a detailed configuration of a combination setter 237. FIG. It is a flowchart which shows operation | movement of the prediction signal generator 103a using the combination setting device 237. It is a flowchart which shows operation | movement of the combination setting device 237 when selecting N prediction adjacent area | regions using the coordinate information of a prediction adjacent area | region. It is a block diagram which shows the structure of the prediction signal generator 103b containing the candidate prediction signal combination setter 237 which is 3rd Embodiment. It is a flowchart which shows a process when selecting the prediction adjacent area | region by the prediction signal generator 103b which is 3rd Embodiment. It is a block diagram which shows the structure of the combination determination device of the candidate prediction signal which is 4th Embodiment. It is a flowchart which shows operation | movement of the combination determination part 203d of a candidate prediction signal. It is a conceptual diagram which shows the processing concept of the evaluator 283 of a reference image. It is a conceptual diagram which shows the processing concept of the evaluator 284 of a motion vector. It is a block diagram which shows the module structure of the combination setting module P237.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Image predictive coding apparatus, 101 ... Input terminal, 102 ... Block divider, 103 ... Prediction signal generator, 104 ... Frame memory, 105 ... Subtractor, 106 ... Converter, 107 ... Quantizer, 108 ... Inverse Quantizer, 109 ... Inverse transformer, 110 ... Adder, 111 ... Entropy encoder, 112 ... Output terminal, 201 ... Template matching unit, 202 ... Memory for coordinate information, 203 ... Combination determiner of candidate prediction signals, 204 ... Predictive region acquisition unit, 205 ... Weighter, 206 ... Adder, 211 ... Target adjacent region acquisition unit, 212 ... Predictive adjacent region acquisition unit, 213 ... Comparator, 214 ... Switch, 230 ... Prediction adjacent region setting device , 231, 237, 237d ... combination setter, 232 ... predicted adjacent area acquirer, 233 target adjacent area acquirer, 234 ... weighted 235 ... adder, 236 ... comparison / selector, 281 ... determiner of prediction adjacent region to be evaluated, 238 ... determiner of candidate prediction signal number, 282 ... counter, 283 ... evaluator of reference image, 284 DESCRIPTION OF SYMBOLS ... Motion vector evaluator, 300 ... Image prediction decoding device, 301 ... Input terminal, 302 ... Entropy decoder, 303 ... Inverse quantizer, 304 ... Inverse transformer, 305 ... Adder, 306 ... Output terminal, 307 ... Frame memory, 308 ... Predictive signal generator, P100 ... Predictive image encoding program, P102 ... Block division module, P103 ... Predictive signal generation module, P104 ... Storage module, P105 ... Subtraction module, P106 ... Conversion module, P108 ... Inverse quantum Module, P109 ... inverse transform module, P110 ... addition module, P111 ... entry P201 encoding module, P201 ... template matching module, P202 ... determination module, P203 ... prediction signal synthesis module, P300 ... image prediction decoding program, P302 ... entropy decoding module, P303 ... inverse quantization module, P304 ... inverse transformation module, P305 ... addition module, P307 ... storage module, P308 ... prediction signal generation module.

Claims (18)

Area dividing means for dividing the input image into a plurality of areas;
A prediction signal generation unit that generates a prediction signal for a target pixel signal of a target region that is a processing target among the plurality of regions divided by the region dividing unit;
Residual signal generation means for generating a residual signal between the prediction signal generated by the prediction signal generation means and the target pixel signal;
Encoding means for encoding the residual signal generated by the residual signal generating means,
The prediction signal generating means includes
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
Using at least coordinate information indicating the position of the predicted adjacent area, and selecting one or more predicted adjacent areas from the plurality of searched predicted adjacent areas;
One or more candidate prediction signals for the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. An image predictive encoding device. Area dividing means for dividing the input image into a plurality of areas;
A prediction signal generation unit that generates a prediction signal for a target pixel signal of a target region that is a processing target among the plurality of regions divided by the region dividing unit;
Residual signal generation means for generating a residual signal between the prediction signal generated by the prediction signal generation means and the target pixel signal;
Encoding means for encoding the residual signal generated by the residual signal generating means,
The prediction signal generating means includes
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
Set the number of predicted adjacent regions to select,
Using the coordinate information indicating at least the position of the predicted adjacent area, the set number of predicted adjacent areas is selected from the search area,
Generating the set number of candidate prediction signals of the target pixel signal based on the selected prediction adjacent region, and generating the prediction signal by processing the candidate prediction signals using a predetermined synthesis method; An image predictive encoding device as a feature. Area dividing means for dividing the input image into a plurality of areas;
A prediction signal generation unit that generates a prediction signal for a target pixel signal of a target region that is a processing target among the plurality of regions divided by the region dividing unit;
Residual signal generation means for generating a residual signal between the prediction signal generated by the prediction signal generation means and the target pixel signal;
Encoding means for encoding the residual signal generated by the residual signal generating means,
The prediction signal generating means includes
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
Generating at least two combinations of predicted adjacent areas including different numbers of predicted adjacent areas using coordinate information indicating at least the positions of the predicted adjacent areas;
By processing pixel signals of predicted adjacent regions belonging to the combination using a predetermined synthesis method, a comparison signal for each of the adjacent pixel signals is generated, and prediction with a high correlation between the comparison signal and the adjacent pixel signal is performed. Selecting a combination of adjacent regions from the generated combination of predicted adjacent regions;
Generate one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and generate the prediction signal by processing the candidate prediction signal using a predetermined synthesis method An image predictive coding apparatus characterized by:   The said prediction signal production | generation means selects a prediction adjacent area | region using the coordinate information containing the positional information of either time or space, or time and space, The Claim 1 characterized by the above-mentioned. Image predictive coding apparatus.   4. The image predictive coding apparatus according to claim 1, wherein the prediction signal generation unit performs processing using coordinate information including an identification number of a reference image to which a prediction adjacent region belongs. 5. .   The said prediction signal production | generation means selects a prediction adjacent area | region using the coordinate information containing the spatial position information on the reference image to which a prediction adjacent area | region belongs, The one of Claim 1 to 3 characterized by the above-mentioned. Image predictive coding apparatus.   2. The prediction signal generation unit selects a prediction adjacent area using coordinate information including a motion vector indicating a spatial positional relationship between the target area and the prediction adjacent area. 4. The image predictive coding apparatus according to claim 1. Data decoding means for decoding the encoded data of the residual signal relating to the target region to be processed in the compressed data;
Residual signal restoration means for restoring a reproduction residual signal from a signal obtained by decoding by the data decoding means;
Prediction signal generating means for generating a prediction signal for the target pixel signal of the target region;
Reproduction image signal generation means for generating a reproduction image signal by adding the prediction signal generated by the prediction signal generation means and the reproduction residual signal restored by the residual signal restoration means, and
The prediction signal generating means includes
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
One or more predictions from the searched plurality of predicted neighboring areas using coordinate information indicating the position of at least one of two or more arbitrary predicted prediction neighboring areas including at least one of the searched plurality of predicted neighboring areas Select the adjacent area,
One or more candidate prediction signals for the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. An image predictive decoding apparatus. Data decoding means for decoding the encoded data of the residual signal relating to the target area to be processed of the compressed data;
Residual signal restoration means for restoring a reproduction residual signal from a signal obtained by decoding by the data decoding means;
Prediction signal generating means for generating a prediction signal for the target pixel signal of the target region;
Reproduction image signal generation means for generating a reproduction image signal by adding the prediction signal generated by the prediction signal generation means and the reproduction residual signal restored by the residual signal restoration means, and
The prediction signal generating means includes
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area from the search area consisting of the already reproduced image with respect to the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target pixel signal,
Set the number of predicted adjacent regions to select,
Using the coordinate information indicating at least the position of the predicted adjacent area, select the set number of predicted adjacent areas from any position of the search area,
Based on the selected prediction adjacent region, generating the number of selected candidate prediction signals of the target pixel signal, and generating the prediction signal by processing the candidate prediction signals using a predetermined synthesis method An image predictive decoding apparatus characterized by the above. Data decoding means for decoding the encoded data of the residual signal relating to the target region to be processed in the compressed data;
Residual signal restoration means for restoring a reproduction residual signal from a signal obtained by decoding by the data decoding means;
Prediction signal generating means for generating a prediction signal for the target pixel signal of the target region;
Reproduction image signal generation means for generating a reproduction image signal by adding the prediction signal generated by the prediction signal generation means and the reproduction residual signal restored by the residual signal restoration means, and
The prediction signal generating means includes
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area from the search area consisting of the already reproduced image with respect to the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target pixel signal,
Set the number of predicted adjacent regions to select,
Generating at least two combinations of predicted adjacent areas including different numbers of predicted adjacent areas using at least coordinate information representing the positions of the predicted adjacent areas;
A combination of a signal having a high correlation between the comparison signal and the adjacent pixel signal by generating a comparison signal for the adjacent pixel signal by processing a pixel signal of a predicted adjacent region belonging to the combination using a predetermined synthesis method. Is selected from the generated combination of predicted adjacent regions,
Generate one or more candidate prediction signals of the target pixel signal based on the prediction adjacent region belonging to the selected combination, and generate the prediction signal by processing the candidate prediction signal using a predetermined synthesis method An image predictive decoding apparatus characterized by:   The said prediction signal production | generation means selects a prediction adjacent area | region using the coordinate information containing either the time or space, or the position of time and space, The Claim 1 characterized by the above-mentioned. Image predictive decoding apparatus.   The image according to any one of claims 8 to 10, wherein the prediction signal generation unit selects a prediction adjacent region using coordinate information including an identification number of a reference image to which the prediction adjacent region belongs. Predictive decoding device.   The said prediction signal production | generation means selects a prediction adjacent area | region using the coordinate information containing the spatial position on the reference image to which a prediction adjacent area | region belongs, The one of Claim 7 to 9 characterized by the above-mentioned. Image predictive decoding apparatus.   9. The prediction signal generation unit selects a prediction adjacent area using coordinate information including a motion vector indicating a spatial positional relationship between the target area and the prediction adjacent area. The image predictive decoding device according to any one of 1 to 10. A region dividing step for dividing the input image into a plurality of regions;
A prediction signal generating step for generating a prediction signal for a target pixel signal of a target region that is a processing target among the plurality of regions divided by the region dividing step;
A residual signal generation step for generating a residual signal between the prediction signal generated by the prediction signal generation step and the target pixel signal;
An encoding step for encoding the residual signal generated by the residual signal generation step,
The predicted signal generation step includes:
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
Using at least coordinate information indicating the position of the predicted adjacent area, and selecting one or more predicted adjacent areas from the plurality of searched predicted adjacent areas;
One or more candidate prediction signals for the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. An image predictive encoding method. A data decoding step for decoding encoded data of a residual signal related to a target region to be processed in the compressed data;
A residual signal restoration step for restoring a reproduction residual signal from the signal obtained by decoding in the data decoding step;
A prediction signal generation step of generating a prediction signal for the target pixel signal of the target region;
A reproduction image signal generation step of generating a reproduction image signal by adding the prediction signal generated in the prediction signal generation step and the reproduction residual signal restored in the residual signal restoration step, and
The predicted signal generation step includes:
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
One or more predictions from the searched plurality of predicted neighboring areas using coordinate information indicating the position of at least one of two or more arbitrary predicted prediction neighboring areas including at least one of the searched plurality of predicted neighboring areas Select the adjacent area,
One or more candidate prediction signals for the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. An image predictive decoding method. An area division module for dividing an input image into a plurality of areas;
A prediction signal generation module that generates a prediction signal for a target pixel signal of a target region that is a processing target among the plurality of regions divided by the region division module;
A residual signal generation module that generates a residual signal between the prediction signal generated by the prediction signal generation module and the target pixel signal;
An encoding module that encodes the residual signal generated by the residual signal generation module;
The prediction signal generation module includes:
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
Using at least coordinate information indicating the position of the predicted adjacent area, and selecting one or more predicted adjacent areas from the plurality of searched predicted adjacent areas;
One or more candidate prediction signals for the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. An image predictive encoding program. A data decoding module that decodes encoded data of a residual signal related to a target region to be processed in the compressed data;
A residual signal restoration module for restoring a reproduction residual signal from a signal obtained by decoding by the data decoding module;
A prediction signal generation module that generates a prediction signal for a target pixel signal in the target region;
A reproduction image signal generation module that generates a reproduction image signal by adding the prediction signal generated by the prediction signal generation module and the reproduction residual signal restored by the residual signal restoration module;
The prediction signal generation module includes:
Search for a plurality of predicted adjacent areas having a high correlation with the target adjacent area consisting of the already reproduced adjacent pixel signal adjacent to the target area consisting of the target pixel signal from the search area consisting of the already reproduced image,
One or more predictions from the searched plurality of predicted neighboring areas using coordinate information indicating the position of at least one of two or more arbitrary predicted prediction neighboring areas including at least one of the searched plurality of predicted neighboring areas Select the adjacent area,
One or more candidate prediction signals for the target pixel signal are generated based on the selected prediction adjacent region, and the prediction signal is generated by processing the candidate prediction signal using a predetermined synthesis method. An image predictive decoding program.

Images