JP5415495B2 - Image predictive encoding device, image predictive encoding method, image predictive encoding program, image predictive decoding device, image predictive decoding method, and image predictive decoding program - Google Patents

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, and in particular, a predictive code using a texture synthesis method. 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, MPEG1-4 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. MPEG4 and H.264 In H.264, in order to further increase the coding efficiency, adjacent pre-reproduced image signals (reconstructed compressed image data) in the same screen as the target block are used for predictive coding in the screen. After generating the prediction signal, the difference signal obtained by subtracting the prediction signal from the signal of the target block is encoded. For predictive coding between screens, refer to adjacent previously reproduced image signals in a screen different from the target block, perform motion correction, generate a prediction signal, and subtract it from the signal of the target block Encode the signal.

  Specifically, H.C. 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. 8 shows ITUH. 2 is a schematic diagram for explaining an intra-screen prediction method used for H.264. In FIG. 8A, a target block 702 is a block to be encoded, and a pixel group 701 composed of pixels A to M adjacent to the boundary of the target block 702 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 the pixel group 701 that is an adjacent pixel directly above the target block 702 downward. In FIG. 8B, the already reproduced pixels (I to L) on the left of the target block 704 are extended to the right to generate a prediction signal. A specific method for generating a prediction signal is described in Patent Document 1, for example. Thus, the difference between each of the nine prediction signals generated by the method shown in FIGS. 8A to 8I 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.

US Pat. No. 6,765,964

  However, when the intra prediction signal is generated by the extrapolation prediction signal generation method used in the prior art, there is a problem that the prediction accuracy for pixels far away from the boundary of the target block is lowered. Further, the method of extending the pixel value in the geometric direction shown in FIG. 8 has a problem that an image signal having a complicated texture signal (picture) cannot be predicted efficiently. Further, as shown in FIG. 8, adjacent pixel groups (for example, 701, 703, 705...) Are decoded pixel values, and include distortion due to encoding (for example, quantization noise). When the prediction signal is generated in this manner, the prediction signal also includes coding distortion.

  Therefore, in order to solve the above-described problem, the present invention improves the prediction accuracy for pixels far from the boundary of the target block, and at the same time efficiently generates a prediction signal for an image having a complex texture signal (picture). An object is to provide a predictive image 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 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 to generate compressed data, wherein the predicted signal generating means is a previously reproduced pixel adjacent to the target pixel signal. A predicted adjacent region having a high correlation with respect to the adjacent region consisting of the signal is searched from a search region that is a predetermined reproduced image, and a plurality of image regions determined based on the searched predicted adjacent region are obtained. It generated as the number of texture signal, and generates a prediction signal by processing using a predetermined texture synthesis method a plurality of texture signals generated.

According to the present invention, the input image is divided into a plurality of regions, a prediction signal for the target pixel signal of the target region that is the processing target among the plurality of divided regions is generated, and the generated prediction signal and the above-described A residual signal with the target pixel signal can be generated, and the generated residual signal can be encoded to generate compressed data. Here, in the present invention, a texture signal is generated based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and the generated texture signal is predicted by processing using a predetermined texture synthesis method. A signal can be generated. As a result, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target region, and to efficiently generate the prediction signal even when the texture signal is complicated. Can do. Furthermore, using the property that the target area and the adjacent area adjacent to the target area are highly correlated, the area having the least error relative to the adjacent area is detected as the predicted adjacent area from the already reproduced image area. By making the pixel group that is adjacent to the region and corresponding to the target region as the texture signal of the target region, even if the texture is complex, it is possible to generate a prediction signal with the same pattern as the adjacent region, and from the boundary of the target region A texture signal with the same pattern can be generated even if they are separated.

  Further, the prediction signal generation means in the image predictive coding apparatus of the present invention further generates an extrapolated texture signal by forming pixels by repeating already reproduced pixel values adjacent to the target pixel signal, It is preferable to generate the prediction signal by combining the texture signal and the extrapolation texture signal using a predetermined combining method.

  According to the present invention, the extrapolated texture signal is generated by repeating the already reproduced pixel value adjacent to the target pixel signal to form a pixel, and the texture signal and the extrapolated texture signal are used in a predetermined synthesis method. A prediction signal can be generated by combining. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target region, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  The encoding means in the image predictive encoding apparatus of the present invention encodes related information indicating the texture synthesis method, and the encoded related information together with the encoded signal encoded by the encoding means. It is preferable to include a sending means for sending the.

  According to the present invention, related information indicating the texture synthesis method can be encoded, and the encoded related information can be transmitted together with the encoded signal. As a result, the image texture synthesis method can be notified to the reception side, and the reception side can generate a prediction signal using the informed texture synthesis method, so that prediction for pixels far away from the boundary of the target region can be performed. A decrease in accuracy can be prevented, and a prediction signal can be efficiently generated even when the texture signal is complicated.

  Further, the prediction signal generation means in the image predictive coding apparatus of the present invention synthesizes the plurality of texture signals for the target pixel signal by performing weighted average processing using a predetermined weighting coefficient, It is preferable to generate a prediction signal.

  According to the present invention, a prediction signal can be generated by combining a plurality of texture signals for a target pixel signal by performing a weighted average process using a predetermined weighting coefficient. By synthesizing (averaging) the texture signals thus generated, there is an effect that a prediction signal with a small statistical error can be generated.

  Furthermore, the problem of uncertainty when there is no large difference between the image areas already reproduced based on adjacent areas is, for example, synthesis of texture signals generated so as to have high correlation by template matching ( By averaging), there is an effect that a prediction signal with a small statistical error can be generated. Further, according to the present invention, by performing weighted average processing on a plurality of texture signals, there is an effect of suppressing coding distortion included in each texture, and a prediction signal with few errors can be generated.

  Further, the prediction signal generation means in the image predictive coding apparatus of the present invention is configured to make an adjacency composed of already reproduced pixel signals adjacent to the target pixel signal from the first search area within the same screen as the target pixel signal. Based on the region, at least one texture signal having a high correlation with the adjacent region is generated, and the already-reproduced pixel adjacent to the target pixel signal from the second search region that is a different screen from the first search region Generate at least one texture signal highly correlated with the adjacent region based on the adjacent region consisting of the signals, and generate a prediction signal by synthesizing each generated texture signal using a predetermined texture synthesis method It is preferable to do.

  According to the present invention, the first search area within the same screen as the target pixel signal has a high correlation with the adjacent area based on the adjacent area including the already reproduced pixel signal adjacent to the target pixel signal. Generate at least one texture signal, and based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a second search area that is a different screen from the first search area, the adjacent area Can generate a prediction signal by generating at least one texture signal having a high correlation with each other and synthesizing the generated texture signals using a predetermined texture synthesis method. As a result, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target region, and to efficiently generate the prediction signal even when the texture signal is complicated. Can do.

  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 of the plurality of regions divided by the region dividing unit. A prediction signal generation unit that generates a prediction signal for the signal, a residual signal generation unit that generates a residual signal between the prediction signal generated by the prediction signal generation unit and the target pixel signal, and the residual signal generation unit Encoding means for encoding the residual signal generated in step (b) to generate a compressed signal, wherein the prediction signal generation means forms a pixel by repeating the already reproduced pixel values adjacent to the target pixel signal. Generating an extrapolated texture signal and synthesizing the extrapolated texture signal using a predetermined synthesis method to generate a prediction signal.

  According to the present invention, the input image is divided into a plurality of regions, a prediction signal for the target pixel signal of the target region that is the processing target among the plurality of divided regions is generated, and the generated prediction signal and the above-described A residual signal with the target pixel signal is generated, and the generated residual signal can be encoded to generate a compressed signal. Further, an extrapolated texture signal is generated by repeating the previously reproduced pixel values adjacent to the target pixel signal to form a pixel, and the extrapolated texture signal is synthesized by using a predetermined synthesis method. A signal can be generated. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target region, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

The image predictive decoding apparatus according to the present invention also includes a residual signal restoring unit that extracts a residual signal related to a target region to be processed from compressed data and restores the residual signal to a reproduced residual signal, and a target pixel of the target region. A prediction signal generation means for generating a prediction signal for the signal, and a prediction signal generated by the prediction signal generation means and a reproduction residual signal restored by the residual signal restoration means, Image prediction means for restoring a pixel signal, wherein the prediction signal generation means is a predetermined prediction adjacent area having a high correlation with an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal. searched from the search area a reproduced image, a plurality of image areas defined on the basis of the prediction adjacent regions the search generates a plurality of texture signals, a plurality of generated text And generates a prediction signal by processing using a predetermined synthesis method char signal.

According to the present invention, a residual signal related to a target region to be processed is extracted from the compressed data, restored to a reproduction residual signal, and a prediction signal for the target pixel signal of the target region is generated and generated. By adding the prediction signal and the reproduction residual signal restored by the residual signal restoration means, the pixel signal of the target area can be restored, and the restored pixel signal can be stored as a reference image. Then, a prediction signal is generated by generating a texture signal based on a stored adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and processing the generated texture signal using a predetermined synthesis method Can be generated. Furthermore, using the property that the target area and the adjacent area adjacent to the target area are highly correlated, the area having the least error relative to the adjacent area is detected as the predicted adjacent area from the already reproduced image area. By making the pixel group that is adjacent to the region and corresponding to the target region as the texture signal of the target region, even if the texture is complex, it is possible to generate a prediction signal with the same pattern as the adjacent region, and from the boundary of the target region A texture signal with the same pattern can be generated even if they are separated.

  Further, the prediction signal generation means in the image predictive decoding apparatus of the present invention further generates an extrapolated texture signal by repeating the already reproduced pixel values adjacent to the target pixel signal to form a pixel, and the texture It is preferable to generate the prediction signal by combining the signal and the extrapolation texture signal using a predetermined combining method.

  According to the present invention, the extrapolated texture signal is generated by repeating the already reproduced pixel value adjacent to the target pixel signal to form a pixel, and the texture signal and the extrapolated texture signal are used in a predetermined synthesis method. A prediction signal can be generated by combining. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target region, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  The compressed data in the image predictive decoding device of the present invention may include related information indicating a texture synthesis method, and the predicted signal generation unit may form a texture signal of the target region using the related information. preferable.

  According to the present invention, the compressed data includes related information indicating the texture synthesis method, and a texture signal of the target region can be formed using the related information. As a result, the receiving side can know the texture synthesis method, and can generate a prediction signal using this known texture synthesis method, thus preventing a decrease in prediction accuracy for pixels far away from the boundary of the target region. Thus, even when the texture signal is complicated, the prediction signal can be generated efficiently.

  Further, the prediction signal generation means in the image predictive decoding apparatus of the present invention combines the plurality of texture signals with respect to the target pixel signal by performing weighted average processing using a predetermined weighting coefficient, thereby predicting Preferably, a signal is generated.

  According to the present invention, a prediction signal can be generated by combining a plurality of texture signals for a target pixel signal by performing a weighted average process using a predetermined weighting coefficient. As a result, there is an effect that a prediction signal with a small error can be generated statistically by synthesizing (averaging) the texture signals generated by template matching so that the correlation becomes high.

  Furthermore, the problem of uncertainty when there is no large difference between the image area already reproduced based on the adjacent area is, for example, synthesis of texture signals generated so as to have a high correlation by template matching ( By averaging), there is an effect that a prediction signal with a small statistical error can be generated. In addition, according to the present invention, by performing weighted averaging processing on a plurality of texture signals, there is an effect of suppressing coding distortion included in each texture, and a prediction signal with few errors can be generated.

  In addition, the prediction signal generation means in the image predictive decoding device of the present invention is configured such that the adjacent region including the already reproduced pixel signal adjacent to the target pixel signal from the first search region within the same screen as the target pixel signal. And at least one texture signal having a high correlation with the adjacent region, and a pixel signal that has already been reproduced and is adjacent to the target pixel signal from a second search region that is a different screen from the first search region. It is preferable that a prediction signal is generated by generating at least one texture signal having a high correlation with the adjacent region based on the adjacent region and synthesizing the generated texture signals.

  According to the present invention, the first search area within the same screen as the target pixel signal has a high correlation with the adjacent area based on the adjacent area including the already reproduced pixel signal adjacent to the target pixel signal. Generate at least one texture signal, and based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a second search area that is a different screen from the first search area, the adjacent area Can generate a prediction signal by generating at least one texture signal having a high correlation with each other and synthesizing the generated texture signals using a predetermined texture synthesis method. As a result, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target region, and to efficiently generate the prediction signal even when the texture signal is complicated. Can do.

  In addition, the image predictive decoding apparatus of the present invention includes a residual signal restoration unit that extracts a residual signal related to the target region from compressed data and restores the residual signal to a reproduced residual signal, and predicts the target pixel signal of the target region. A prediction signal generation unit that generates a signal; and an image restoration unit that restores the pixel signal of the target region by adding the prediction signal and the reproduction residual signal, and the prediction signal generation unit includes: An extrapolated texture signal is generated by forming a pixel by repeating previously reproduced pixel values adjacent to the target pixel signal, and a predicted signal is synthesized by synthesizing the extrapolated texture signal using a predetermined synthesis method. Generate.

  According to the present invention, a residual signal related to the target region is extracted from the compressed data and restored to a reproduction residual signal, a prediction signal for the target pixel signal of the target region is generated, and the prediction signal and the reproduction residual are generated. By adding the signal, the pixel signal of the target region can be restored. Then, an extrapolated texture signal is generated by repeating the already reproduced pixel values adjacent to the target pixel signal to form a pixel, and the extrapolated texture signal is synthesized by using a predetermined synthesis method. Can be generated. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target region, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  Note that the present invention can be described as the invention of the image predictive encoding device and the image predictive decoding device as described above. In addition, the present invention can be described as follows. It can also be described as an invention of an image decoding program. This is substantially the same invention only in different categories and the like, and has the same operations and effects.

The image predictive coding method according to the present invention includes a region dividing step of dividing an input image into a plurality of regions, and a target pixel signal of a target region that is a processing target among the plurality of regions divided by the region dividing step. Generated by a prediction signal generation step for generating a prediction 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 encoded step of generating compressed data by encoding the residual signal, wherein the predicted signal generating step correlates with an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal. A search adjacent region having a high image quality is searched from a search region that is a previously reproduced image, and a plurality of image regions determined based on the searched predicted adjacent region are searched. The generated as a plurality of texture signals, and generates a prediction signal by processing using a predetermined texture synthesis method a plurality of texture signals generated.

  The image predictive coding method according to 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 of encoding the residual signal generated by the step of generating a compressed signal, and the prediction signal generation step forms pixels by repeating already reproduced pixel values adjacent to the target pixel signal To generate an extrapolation texture signal, and generate the prediction signal by synthesizing the extrapolation texture signal using a predetermined synthesis method. Characterized in that it.

The image predictive decoding method according to the present invention also includes a residual signal restoration step for extracting a residual signal relating to a target region to be processed from compressed data and restoring the residual signal to a reproduced residual signal, and a target pixel of the target region. A prediction signal generation step for generating a prediction signal for the signal, and a prediction signal generated by the prediction signal generation step and a reproduction residual signal restored by the residual signal restoration step, An image restoration step for restoring a pixel signal, wherein the prediction signal generation step is a predetermined prediction adjacent region having a high correlation with an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal. Search from the search area, which is a playback image, and generate multiple image areas as multiple texture signals determined based on the searched predicted adjacent areas , And generates a prediction signal by processing using a predetermined synthesis method a plurality of texture signals generated.

  The image predictive decoding method of the present invention also includes a residual signal restoration step for extracting a residual signal related to the target region from compressed data and restoring it to a reproduced residual signal, and a prediction for the target pixel signal of the target region. A prediction signal generation step of generating a signal, and an image restoration step of restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal, and the prediction signal generation step includes: An extrapolated texture signal is generated by forming a pixel by repeating previously reproduced pixel values adjacent to the target pixel signal, and a predicted signal is synthesized by synthesizing the extrapolated texture signal using a predetermined synthesis method. It is characterized by generating.

The image predictive coding program according to the present invention includes a region dividing module that divides 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 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 generates compressed data by encoding the residual signal generated by the prediction signal generation module, wherein the prediction signal generation module applies to an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal. Thus, a predicted adjacent area having a high correlation is searched from a search area that is a predetermined reproduced image, and the predicted adjacent area is searched based on the searched predicted adjacent area. A plurality of image areas defined by have generated as a plurality of texture signals, thereby executing a computer to generate a prediction signal by processing using a predetermined texture synthesis method a plurality of texture signals generated It is characterized by.

The image predictive coding program according to the present invention includes a region dividing module that divides 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 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 step (b) to generate a compressed signal, and the prediction signal generation module forms a pixel by repeating the already reproduced pixel values adjacent to the target pixel signal To generate an extrapolation texture signal and synthesize the extrapolation texture signal using a predetermined synthesis method. Characterized in that to be executed by a computer to generate a prediction signal by the.

The image predictive decoding program according to the present invention includes a residual signal restoration module that extracts a residual signal related to a target region to be processed from compressed data and restores the residual signal to a reproduced residual signal, and a target pixel of the target region. A prediction signal generation module that generates a prediction signal for the signal; and by adding the prediction signal generated by the prediction signal generation module and the reproduction residual signal restored by the residual signal restoration module, An image restoration module that restores a pixel signal, and the prediction signal generation module includes an already-established prediction adjacent region that has a high correlation with an adjacent region that is composed of already reproduced pixel signals adjacent to the target pixel signal. A search area that is a reproduced image is searched, and a plurality of image areas determined based on the searched predicted adjacent area are converted into a plurality of texts. It generated as over signal, characterized in that to be executed by a computer to generate a prediction signal by processing using a predetermined synthesis method a plurality of texture signals generated.

In addition, the image predictive decoding method of the present invention includes a residual signal restoration module that extracts a residual signal related to the target region from compressed data and restores the residual signal to a reproduced residual signal, and predicts the target pixel signal of the target region. A prediction signal generation module that generates a signal, and an image restoration module that restores the pixel signal of the target region by adding the prediction signal and the reproduction residual signal, and the prediction signal generation module includes: An extrapolated texture signal is generated by forming a pixel by repeating previously reproduced pixel values adjacent to the target pixel signal, and a predicted signal is synthesized by synthesizing the extrapolated texture signal using a predetermined synthesis method. It is characterized by causing a computer to execute the generation.

  According to the present invention, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target region, and to efficiently generate the prediction signal even when the texture signal is complicated. Can be generated.

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. 2 is a block diagram showing a texture synthesizer 201. FIG. It is a schematic diagram for demonstrating the matching process in a texture production | generation method. It is a flowchart which shows the texture generation method in the texture generator 307. It is a figure which shows an example of the shape of object adjacent area | region. It is a figure which shows an example of the shape of another object adjacent area | region. It is a schematic diagram which shows the production | generation of a texture signal from the pixel formed by repeating the value of the adjacent already reproduced pixel as it is. 3 is a flowchart illustrating an image predictive encoding method in the image predictive encoding device 100. It is a flowchart which shows the production | generation method of the synthetic | combination texture signal in a modification. It is a block diagram which shows the image prediction decoding apparatus 900. FIG. It is a block diagram which shows the texture synthesizer 908. 5 is a flowchart illustrating an image predictive decoding method of the image predictive decoding device 900. 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 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 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.

  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.

  FIG. 1 is a block diagram showing an image predictive encoding device 100 according to an embodiment of the present invention. The image predictive coding apparatus 100 includes an input terminal 101, a block divider 102, a prediction signal generator 103, a frame memory 104, a subtractor 105, a transformer 106, a quantizer 107, an inverse quantizer 108, and an inverse transformer. 109, an adder 110, an entropy encoder 111, and an output terminal 112. 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 is a moving image signal input from the input terminal 101 and divides an image to be encoded 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 generates the prediction signal generated by the prediction signal generator 103 input via the line L103 from the target region obtained by the division by the block divider 102 input via the line L102. Is a part for generating a residual signal. The subtractor 105 outputs the residual signal obtained by subtraction to the converter 106 via 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 encodes the transform coefficient quantized by the quantizer 107 and outputs it along with information on the prediction method via the L111. The output terminal 112 outputs 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 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, and stores it in the frame memory 104. In this embodiment, converter 106 and inverse converter 109 are used. However, other conversion processes may be used instead of these converters, and converter 106 and inverse converter 109 are not necessarily required. . As described above, in order to perform intra-screen prediction or inter-screen prediction for the subsequent target region, the compressed pixel signal of the target region is inversely processed and restored, and stored in the frame memory 104.

  Next, details of the prediction signal generator 103 will be described. The prediction signal generator 103 generates a prediction signal for a target region (hereinafter referred to as a target block) that is a target of encoding processing. In the embodiment according to the present invention, two kinds of prediction methods are used. That is, the prediction signal generator 103 generates a prediction signal using at least one of the inter-screen prediction method and the intra-screen prediction method.

  The prediction signal generator 103 using the inter-screen prediction method obtains motion information that gives a prediction signal with the smallest error with respect to the target block from the reference image using a reproduced image that has been encoded and restored in the past as a reference image. . This process is called motion detection. Further, according to circumstances, the target block may be subdivided, and the inter-screen prediction method may be determined for the subdivided small area. In this case, the prediction signal generator 103 determines the most efficient division method and each piece of motion information for the entire target block from among various division methods. In the embodiment according to the present invention, the prediction signal generator 103 receives the target block via the line L102 and the reference image via the L104 in order to generate the above-described prediction signal. Moreover, the prediction signal generator 103 uses, as a reference image, a plurality of images that have been encoded and restored in the past. For details, see MPEG-2 or 4, ITUH. It is the same as any of the H.264 methods.

  The motion information and the small area dividing method determined in this way are sent to the entropy encoder 111 via the line L112, encoded, and sent out from the output terminal 112. The prediction signal generator 103 acquires a reference image from the frame memory 104 based on the small region dividing method and the motion information corresponding to each small region, and generates an inter-screen prediction signal. The inter-screen prediction signal generated in this way is sent to the subtractor 105 via the line L103.

  The prediction signal generator 103 using the intra prediction method generates an intra prediction signal using the already reproduced pixel values spatially adjacent to the target block. Specifically, the prediction signal generator 103 acquires the already reproduced pixel signal in the same screen from the frame memory 104, determines an in-screen prediction method for generating a prediction signal by a predetermined method, and determines An intra-screen prediction signal is generated based on the intra-screen prediction method. On the other hand, the information on the prediction method is sent to the entropy encoder 111 via the line L112, encoded, and sent from the output terminal 112. The inter-screen prediction signal generated in this way is sent to the subtractor 105. Detailed processing of the prediction signal generator 103 will be described later.

  Of the inter-screen prediction signal and the intra-screen prediction signal obtained as described above, the signal having the smallest error is selected and sent to the subtractor 105. However, since there is no past image for the first image, all target blocks are processed by intra prediction. Note that the intra-screen prediction signal generation method described below can also be applied to encoding / decoding of still images such as photographs.

  The subtractor 105 subtracts the prediction signal (via the line L103) from the signal of the target block (via the line L102) to generate a residual signal. This residual signal is subjected to discrete cosine transform by a transformer 106, and each transform coefficient is quantized by a quantizer 107. Finally, the transform coefficient quantized by the entropy encoder 111 is encoded together with information relating to the generation method of the prediction signal, and is transmitted from the output terminal 112.

  Next, the processing of the prediction signal generator 103 according to the present invention will be described focusing on the case of the intra prediction method. The same applies to the inter-screen prediction method. FIG. 2 is a block diagram showing a prediction signal generator 103 used in the image predictive coding apparatus according to the embodiment of the present invention. The prediction signal generator 103 includes a texture synthesizer 201 and a texture signal determiner 202. It consists of

  The texture synthesizer 201 receives an image signal (reproduced image signal) that has already been reproduced in the past processing from the frame memory 104 of FIG. The texture synthesizer 201 receives a reproduced image signal in the same screen as the target block when the intra prediction method is used, and a screen different from the target block when the inter prediction method is used. A reproduced image signal in (frame or field) is input. The texture synthesizer 201 generates a plurality of texture signals by a method as described below, and outputs the generated plurality of texture signals to the texture signal determiner 202.

  The texture signal determiner 202 inputs the image signal of the target block via the line L102 at the same time as the texture signal is output from the texture synthesizer 201. Then, the texture signal determiner 202 compares each of the plurality of input texture signals with the image signal of the target block, and outputs the texture signal with the smallest error as a prediction signal of the target block via the line L103. In the present embodiment, the texture signal determiner 202 obtains an error value by using the sum of absolute difference values of pixel values between the image signal and the texture signal of the target block, and gives a texture signal that gives the smallest error value. Is determined as a prediction signal.

  Next, the texture synthesizer 201 will be described with reference to FIG. FIG. 3 is a block diagram showing the texture synthesizer 201 used in the image predictive coding apparatus 100 according to the embodiment of the present invention. The texture synthesizer 201 includes a synthesizer 301, an adder 302, weighters 303 to 306, texture generators 307 to 308, and a first extrapolator 309 to an Nth extrapolator 310. Here, the texture generator 307 includes a first texture acquisition unit 311 and a first matching unit 312, and the texture generation unit 308 includes an Mth texture acquisition unit 313 and an Mth matching unit 314. In FIG. 3, M texture generators and N signal extrapolators are used. In this embodiment, M = 5 and N = 9, but other numerical values may be used. For example, M = 0, N = 9, or M = 5, N = 0.

  First, the operation of the texture generator 307 will be described. The first matching unit 312 accesses the reproduced image signal in the frame memory 104 via the line L104 and performs matching processing. Here, this matching process will be described. FIG. 4 is a schematic diagram for explaining a matching process in the texture generation method according to the embodiment of the present invention. Here, a case where a prediction signal for the target block 402 is generated will be described.

  A “target adjacent region” (also referred to as a template) and a “search region” are set for a certain target block by a predetermined method. In this embodiment, 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. As the “target adjacent region”, the already reproduced pixel group (reverse L character region) 406 adjacent to the target block 402 is used.

  Note that the search area 401 may not be in contact with the target block 402 at all. Further, the target adjacent region 406 may be in contact with the target block 402 with at least one pixel. Further, in the search area 401, the sum of absolute error values (SAD) between corresponding pixels is obtained between the pixel groups having the same shape as the target adjacent area 406, and the area that gives the smallest SAD is searched. Is assumed to be a “predicted adjacent region”.

  In FIG. 4, the predicted adjacent area 408 is determined as the “predicted adjacent area”. Then, a region 407 in contact with the predicted adjacent region 408 is determined as a texture signal for the target block 402. Here, the positional relationship between the predicted adjacent region 408 and the region 407 indicating the texture signal is the same as the target block 402 and the target adjacent region 406, but this need not be the case.

  When the texture signal is determined in this way, the first matching unit 312 receives the coordinate information related to the coordinates of the predicted adjacent region 408 or the coordinate information related to the coordinates of the region 407 that is the texture signal via the L312 as the first texture acquisition unit. 311 is sent. The first texture acquisition unit 311 acquires the pixel signal of the region 407 as a texture signal from the frame memory 104 (see FIG. 1) via the line L314 based on the coordinate information, and outputs it to the combiner 301 via the line L307. To do.

  By such processing, a region 407 close to the adjacent region 406 of the target block 402 (with a small error) is searched from the search region 401 and generated as a texture signal.

  FIG. 5 is a flowchart illustrating a texture generation method in the texture generator 307 according to an embodiment of the present invention. A target adjacent region adjacent to the target block is acquired by the first matching unit 312 (S411). Next, for the target adjacent area acquired by the first matching unit 312, the first texture acquisition unit 311 obtains the absolute value sum SAD value between the target adjacent area and the pixel group having the same shape in the search area (S 412). ). The first texture acquisition unit 311 compares the SAD value with the minimum value of the SAD values so far (S413). If the first texture acquisition unit 311 determines that the SAD value is smaller than the minimum value, the process proceeds to S414. If it is determined that this is not the case, the process proceeds to S415. Note that the initial value of the minimum value is determined in advance.

  The SAD value obtained in S412 is set by the first texture acquisition unit 311 as a minimum value, and the first texture acquisition unit 311 determines that the pixel signal of the region adjacent to the prediction target region is the texture signal of the target block. (S414). Thereafter, the first matching unit 312 confirms whether or not the entire search area 401 has been searched (S415). If it is determined that the search has not been completed, the process returns to S412 and the first texture acquisition unit 311 performs the target adjacent area. An absolute value sum SAD between 406 and another pixel group (having the same shape) in the search area 401 is obtained. When all the search is completed, the processing for one target block (here, the target block 402) is completed. Then, the same processing as that of the next target blocks 403 to 405 is sequentially performed.

  With the above processing, the texture generator 307 searches for the predicted adjacent area 408 from the search area 401 using the target adjacent area 406 adjacent to the target block 402, and the area 407 adjacent to the predicted adjacent area 408 is a texture signal. Can be determined and obtained.

  In FIG. 5, the processing using the search area 401 in the same screen as the target block (which may be either a frame or a field) has been described. However, the search area is in a different screen from the target block and has already been reproduced. May be used. In this case, the predicted adjacent area and the texture signal adjacent to the predicted adjacent area are acquired from a screen different from the target block. Furthermore, as a search area, an already reproduced image signal on both the same screen and a different screen as the target block may be used. In this case, the texture signal to be synthesized is a texture signal searched from the same screen as the target block and a texture signal searched from a different screen.

  Next, the operation of the texture generator 308 will be described. The operation of the texture generator 308 is almost the same as that of the texture generator 307, but the texture generator 308 searches the target adjacent area 406 shown in FIG. 4 using an area having another shape. FIG. 6 is a diagram illustrating an example of the shape of the different target adjacent regions. As shown in FIG. 6, the shape includes a rectangular area 501 formed on the left side of the target block 502, a rectangular area 503 formed on the upper side of the target block 504, and an area formed on the upper left side of the target block 504. A region 507 formed in a triangular shape from the upper side to the left side of the target block 508 is conceivable, but is not limited thereto.

  Then, the texture generator 308 performs the processing shown in FIG. 5 on the target adjacent region having the shape shown in FIG. 6D, and generates a corresponding texture signal. In addition, other texture generators (not shown) generate texture signals using target adjacent regions having shapes shown in FIGS. 6A, 6B, and 6C, respectively.

  In this manner, the texture generators 307 and 308 and other texture generators (not shown) each generate M (M = 5) texture signals and output them to the combiner 301. In FIG. 6, the case where the target block is composed of 2 × 2 pixels has been described. However, as shown in FIG. 7, the texture signal is generated by the same process even when the target block is 4 × 4 pixels or 8 × 8 pixels. When the target block is 4 × 4 pixels, a region 601 for the target block 602, a region 603 for the target block 604, a region 605 for the target block 606, and regions 607a and 607b for the target block 608 can be considered. . Similarly, for 8 × 8 pixels, a region 609 for the target block 610, a region 611 for the target block 612, a region 613 for the target block 614, and regions 615a and 615b for the target block 616 are considered. .

  Next, the operation of the first extrapolator 309 will be described. The signal extrapolator 309 generates a texture signal by extrapolating a so-called signal using pixels formed by repeating the values of already reproduced pixels adjacent to the target block (for example, the target block 402) as they are. FIG. 8 is a schematic diagram showing the generation of a texture signal using this method. FIG. 8A extrapolates a signal by repeating the values of already reproduced pixels 701 (A to D) above the target block 702 in the direction of the arrow. FIG. 8B extrapolates the signal by repeating the previously reproduced pixels 703 (I to L) in the direction of the arrow. FIG. 8C shows the average value of the already reproduced pixels 705 (A to D and I to L), which is the pixel value of the target block 706.

  As shown in FIGS. 8D to 8I, the signal is formed by extrapolating the signal by repeating the already reproduced pixels (707 to 717) in the direction of the arrow. The N (N = 9) first extrapolator 309 to N extrapolator 310 shown in FIG. 3 generate N texture signals by these methods, and the generated texture signals are sent to the combiner 301. Output.

  Next, the configuration shown in FIG. 3 will be described. The synthesizer 301 synthesizes a plurality of (M + N) texture signals by performing predetermined arithmetic processing on the plurality of input texture signals, and outputs the synthesized signal via the line L202. In the present embodiment, a plurality of K synthesized texture signals (or texture signals) are generated from a plurality of input texture signals.

  That is, the synthesizer 301 averages the texture signals from the texture generators 307 and 308 to generate a first synthesized texture signal. In this case, the weight values of the weighters 303 and 304 are (1 / M), and the weight values of the weighters 305 and 306 are 0. The synthesizer 301 generates a second synthesized texture signal using the output of the first extrapolator 309. In this case, the weight value of the weighter 305 is 1, and the weight values of the other weighters are 0. Furthermore, the synthesizer 301 generates a third synthesized texture signal using the output of the Nth extrapolator 310. In this case, the weight value of the weighter 306 is 1, and the weight values of the other weighters are 0.

  Further, the synthesizer 301 generates and outputs a texture signal (extrapolated texture signal) generated using an output from another signal extrapolator (not shown) as a synthesized texture signal. Note that the combiner 301 may generate a plurality of combined texture signals using weight values other than the above-described weight values. For example, when the weight value of the weighter 303 is 1 and the weight value of the other weighter is 0, or when the weight value of the weighter 304 is 1 and the weight value of the other weighters is 0, signal extrapolation It is possible to use a case in which the weight values of the weighting devices 305 and 306 connected to the device are each 1 / N and the weight values of the other weighting devices are 0. As another combination, two texture signals input to the combiner 301 may be combined and weighted and averaged to generate a combined texture signal. In this case, the texture signal from the texture generator and the texture signal from the signal extrapolator may be combined and combined.

  As described above, the synthesizer 301 may generate a synthesized texture signal (that is, a prediction signal) using at least one texture signal. For example, the texture generators 307 to 308 and the first extrapolators 309 to Nth. What is necessary is just to make it a synthetic | combination texture signal using the texture signal (or extrapolation texture signal) output from any one texture generator or extrapolation device among the extrapolation devices 310.

  In this way, a plurality (K pieces) of synthesized texture signals (or texture signals) are sent to the texture signal determiner 202 of FIG. 2 via the line L202, and the texture closest to the target block (that is, the error is small). A signal is determined and output as a prediction signal via line L103.

  FIG. 9 is a flowchart showing an image predictive encoding method in the image predictive encoding apparatus 100 according to the embodiment of the present invention. First, the texture generators 308 to 309 and the first extrapolator 309 to the Nth extrapolator 310 generate (M + N) texture signals for the target block (step 802). Next, the combiner 301 combines the M texture signals to generate at least two texture signals (M in this embodiment), and L combined texture signals (in this embodiment). L = 1) is generated (step 803). Although at least two texture signals are generated here, it is sufficient that at least one texture signal (or extrapolated texture signal) is generated.

  One texture signal or synthesized texture signal closest to the target block is selected from K (= N + L) of N texture signals and L synthesized texture signals, and the selected texture signal or synthesized texture signal is selected. Is determined to be a prediction signal (step 804). The residual signal indicating the difference between the prediction signal determined in this way and the signal of the target block is encoded by the converter 106, the quantizer 107, and the entropy encoder 111 (S805). The encoded residual signal and the prediction signal generation related information are output via the output terminal 112 (S806).

  Here, the prediction signal generation related information includes at least one of information on which texture generation method should be used to generate the prediction signal and information on the weight value of the weighting unit used. Based on these pieces of information, a prediction signal is generated on the receiving side. After these processes (or in parallel), the residual signal encoded to predictively encode the subsequent target block is decoded by the inverse quantizer 108 and the inverse transformer 109. 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 (S807). If all the target areas have not been processed, the process returns to S802, the process for the next target block is performed, and if completed, the process ends (S808).

  Next, a modified example of the process shown in FIG. 4 will be described. FIG. 10 is a flowchart showing a method of generating a composite texture signal in the modified example. First, the target matching area is acquired in the target block by the first matching unit 312 (S901). Next, for the target adjacent region acquired by the first matching unit 312, the first texture acquisition unit 311 obtains the absolute value sum SAD between the target adjacent region and the pixel group having the same shape formed in the search region ( S902).

  If the first texture acquisition unit 311 compares the SAD with a predetermined threshold value and determines that the threshold value is smaller than the threshold value, the process proceeds to S904. If not, the process proceeds to S905 (S903). If it is determined that the value is smaller than the threshold value, a texture signal (for example, 407 in FIG. 4) in contact with the predicted adjacent region that gives an SAD smaller than the threshold value is stored as a candidate texture signal (S904). Then, the first matching unit 312 confirms whether or not the entire search area has been searched (S905).

  If it is determined that the search has not been completed, the process returns to S902, and the absolute value sum SAD between the target adjacent area and another pixel group in the search area is obtained (S902). When all the search has been completed, the obtained one or more candidate texture signals are all averaged by the first texture acquisition unit 311 to generate a texture signal (S906).

  This texture signal is sent to the target block by the texture signal determiner 202 shown in FIG. 2 together with the other texture signals generated by the texture generator 308 and the first extrapolator 309 to the Nth extrapolator 310 described above. A close (small error) texture signal is selected, and the selected texture signal is determined to be a prediction signal.

  In this embodiment, it is necessary to use the same threshold on the reproduction side, and it is necessary to determine the threshold in advance on the transmission side and the reception side. As another embodiment, the threshold value may be transmitted together with the compressed data from the transmission side to the reception side, and the reception side may generate the synthesized texture signal by the method shown in FIG. 10 using the received threshold value. That is, the prediction signal generator 103 outputs information indicating a threshold value to the entropy encoder 111 via the line L112, and the entropy encoder 111 transmits the threshold information together with the compressed data from the output terminal 112 to the reception side. Thus, the threshold value may be transmitted to the receiving side.

  Furthermore, it is conceivable to set an upper limit for the number of candidate texture signals to be averaged in addition to the threshold value. In this case, when the number of candidate texture signals exceeds the upper limit, only the number of candidate texture signals equal to or less than the upper limit value is averaged. The upper limit value may be changed depending on the frame. In this case, the transmission side sends the upper limit value to the reception side for each frame. For example, as described above, the prediction signal generator 103 can output to the entropy encoder 111 via the line L112 and transmit it to the receiving side via the output terminal 112. Alternatively, the number of candidate texture signals to be averaged may be determined without using a threshold, and the number of candidate texture signals may be always synthesized. Information regarding this number may be sent for each region in the frame, for each frame, or for each sequence, or a value determined in advance on the transmitting side and the receiving side may be used.

  If the SAD value obtained by the template matching as described above is smaller than a certain threshold value, it means that there is no great difference between these SAD values. Such a problem of the uncertainty of the optimal solution has an effect that a prediction signal with a small statistical error can be generated by combining (averaging) a plurality of similar texture signals generated by template matching. In addition, since the image predictive coding apparatus according to the present embodiment can uniquely obtain these texture signals on the reproduction side (image predictive decoding apparatus on the receiving side), (which texture signal is used for synthesis) Since it is not necessary to send auxiliary information to the reproduction side, there is an effect that encoding can be performed without handling such auxiliary information.

  In the weighted average of texture signals according to the present invention, weighting averaging is performed after determining weights used when averaging candidate texture signals in a relationship (for example, inversely proportional relationship) that can be expressed by a predetermined function in the SAD value. Also good. Furthermore, these weight values may be learned and updated in units of blocks or frames. Specifically, it is updated using a sequential update formula that matches the SAD value.

  In addition, about the process in FIG. 10, you may change to the following process as a modification. That is, the image predictive coding apparatus 100 acquires position information in the same frame of the candidate texture signal determined in step 904, stores it in a buffer unit (not shown), and stores the stored position information in entropy. The data is encoded and transmitted to the image predictive decoding device on the receiving side together with the difference signal.

  In the image predictive decoding device on the receiving side, the position information received from the image predictive encoding device 100 is entropy-decoded, and based on this position information, a plurality of candidate texture signals are acquired from the frame memory, and they are weighted averaged Then, a texture signal is generated. That is, in this modified example, in the image predictive decoding apparatus, instead of the processing in steps 901 to 905, processing for acquiring a plurality of candidate texture signals from the frame memory based on the received position information is performed.

  As another modification, the image predictive coding device 100 does not acquire the target adjacent region (template) in S901 of FIG. 10, and in S902, calculates the difference value SAD from the target block and the pixel group in the search region. The position information (for example, coordinate information on the screen) of the candidate texture signal is acquired at the same time as the candidate texture signal is determined, and this is transmitted together with the difference signal to the image predictive decoding apparatus on the receiving side. The image predictive decoding apparatus on the receiving side performs entropy decoding on the received position information, acquires a plurality of candidate texture signals from the frame memory based on the position information, and performs weighted averaging on the acquired candidate texture signals. To generate a texture signal.

  In this way, the image predictive coding apparatus 100 acquires the position information of the candidate texture signal and transmits it to the image predictive decoding apparatus on the receiving side, and the image predictive decoding apparatus on the receiving side uses this position information to generate the candidate texture signal. Can be acquired, and the processing load of the image predictive decoding apparatus on the receiving side can be reduced.

  Next, an image predictive decoding method according to the present invention will be described. FIG. 11 is a block diagram illustrating an image predictive decoding apparatus 900 according to an embodiment of the present invention. The image predictive decoding apparatus 900 includes an input terminal 901, a data analyzer 902, an inverse quantizer 903, an inverse transformer 904, an adder 905, an output terminal 906, a frame memory 907, and a texture synthesizer 908. The inverse quantizer 903 and the inverse transformer 904 function as a decoding unit, but other units may be used as the decoding unit. Further, the inverse converter 904 may be omitted. Each configuration will be described below.

  The input terminal 901 inputs compressed data that has been compression-encoded by the above-described image predictive encoding method. The compressed data includes a residual signal encoded by predicting a target block obtained by dividing an image into a plurality of blocks, and information related to generation of a prediction signal. As the prediction signal generation related information related to the generation of the prediction signal, which one of the texture generators 307 and 308, the first extrapolator 309, and the Nth extrapolator 310 shown in FIG. ) Is included. Moreover, you may include the information regarding the value of the weight used for the weighters 303-306 arrange | positioned at the combiner 301. FIG. Furthermore, when a synthesized texture signal is generated using the template matching method shown in FIG. 10, the threshold value and the upper limit value of the number of texture signals used for synthesis are included in the information related to the generation of the prediction signal.

  The data analyzer 902 extracts the residual signal of the target block, the prediction signal generation related information related to the generation of the prediction signal, and the quantization parameter by analyzing the compressed data input at the input terminal 901.

  The inverse quantizer 903 inversely quantizes the residual signal of the target block based on the quantization parameter (via the line L902). The inverse transformer 904 performs inverse discrete cosine transform on the inversely quantized data.

  The texture synthesizer 908 inputs information related to the generation of the prediction signal from the data analyzer 902 via the line L902b. The texture synthesizer 908 acquires a reference image from the frame memory 907 and generates a prediction signal based on prediction signal generation related information related to generation of a prediction signal (in a method as described later). The texture synthesizer 908 outputs the generated prediction signal to the adder 905 via the line L908. Note that the information for generating the prediction signal may be determined in advance without inputting the information for generating the prediction signal. For example, the threshold values shown in FIG. 10 and the weight values in the weighters 303 to 306 may be stored in advance.

  The adder 905 adds the prediction signal generated by the texture synthesizer 908 to the residual signal restored by the inverse quantizer 903 and the inverse transformer 904, and the reproduced image which is the target block via the line L905. The data is output to the output terminal 906 and the frame memory 907. The output terminal 906 outputs to the outside (for example, a display).

  The frame memory 907 stores the playback image output from the adder 905 as a reference playback image for the next decoding process as a reference image.

  Next, the texture synthesizer 908 will be described. The details of the texture synthesizer 908 are shown in FIG. 12, and basically have the same configuration as the configuration shown in FIG. 3 or a function corresponding thereto. A texture signal is generated using any one of the texture generators 1007 to 1008 and the first extrapolator 1009 to the Nth extrapolator 1010 according to the prediction signal generation related information related to the generation of the prediction signal input via the line L902b. Is controlled.

  For example, when the prediction signal generation related information indicating that the texture generators 1007 to 1008 are used is received for the target block to be reproduced, only the texture generators 1007 to 1008 are operated, and the other first outside The insertion device 1009 to the Nth extrapolation device 1010 do not operate. For example, when prediction signal generation related information indicating that a texture signal is generated using the first extrapolator 1009 is received, the other Nth extrapolator 1010 and texture generators 1007 to 1008 do not operate. . Note that the prediction signal generation related information may be generated by other combinations. For example, there may be prediction signal generation related information indicating that only the texture generator 1007 and the first extrapolator 1009 are used.

  As described above, the synthesizer 1001 may generate a synthesized texture signal (that is, a prediction signal) using at least one texture signal. For example, based on the prediction signal generation related information received via the line 902b. A texture signal (or extrapolation texture signal) output from any one of the texture generators or extrapolators among the texture generators 1007 to 1008 and the first extrapolator 1009 to the Nth extrapolator 1010. It may be used as a synthesized texture signal.

  The texture generators 1007 and 1008 generate texture signals by the same processing as the texture generators 307 and 308 shown in FIG. Details are as described above. The first extrapolator 1009 and the Nth extrapolator 1010 generate texture signals in the same manner as the first extrapolator 309 and the Nth extrapolator 310 described in FIG.

  In this way, a texture signal is generated by the texture generator 1007 or 1008, the first extrapolator 1009 or the Nth extrapolator 1010 specified by the prediction signal generation related information (via the line L92b) related to the generation of the prediction signal. And sent to the synthesizer 1001. In the present embodiment, weighters 1003 to 1006 perform weighting processing on the sent texture signal with a predetermined weight value, and are sent to adder 1002. The adder 1002 adds the weighted texture signals and outputs the added texture signal as a prediction signal via the line L908. When the weight values used for the weighters 1003 to 1006 are not determined in advance, the weight values added to the prediction signal generation related information related to the generation of the prediction signal can be applied to the weighters 1003 to 1006. It is.

  In FIG. 12, when a texture signal is generated only by a texture generator, there is no need for a signal extrapolator. In this case, the prediction signal generation related information related to the generation of the prediction signal does not require instruction information indicating which one of the texture generator, the first extrapolator, and the Nth extrapolator is used.

  As described above, when the position information of the candidate texture signal obtained from the target adjacent region or target block is received from the image predictive coding apparatus 100, the image predictive decoding apparatus 900 entropy-decodes the received position information. On the basis of this position information, a plurality of candidate texture signals are acquired from the frame memory 907, and the acquired candidate texture signals are weighted and averaged to generate a texture signal. Thereby, the generation process of a candidate texture signal can be reduced.

  Next, an image predictive decoding method in the image predictive decoding apparatus 900 according to the embodiment of the present invention will be described with reference to FIG. First, the compressed data is input via the input terminal 901 (S1102). Then, in the data analyzer 902, entropy decoding is performed on the compressed data, and quantized transform coefficients, quantization parameters, and prediction signal generation related information are extracted (S1103). Based on the extracted prediction signal generation related information, the texture synthesizer 908 generates a plurality of texture signals, and generates a prediction signal by combining the generated plurality of texture signals (S1104).

  On the other hand, the quantized transform coefficient is inversely quantized using the quantization parameter in the inverse quantizer 903, and inverse transform is performed in the inverse transformer 904 to generate a reproduction difference signal (S1105, S1106). . Then, the generated prediction signal and the reproduction difference signal are added to generate a reproduction signal (S1107). The generated reproduction signal is stored in the frame memory to reproduce the next target block (S1108). If there is next compressed data, this process is repeated again (S1103), and all data is processed to the end (S1109). If necessary, the process may return to S1102 to capture compressed data.

  The image predictive encoding method and the image predictive decoding method according to the present invention 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. 14 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. 15, the prediction signal generation module P103 includes a synthesis module P301, a texture generation module P311 and an extrapolation module P309.

  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.

  FIG. 16 is a block diagram showing modules of a program that can execute the image predictive decoding method. The image predictive decoding program P900 includes a data analysis module P902, an inverse quantization module P903, an inverse transform module P904, an addition module P905, a storage module P907, and a texture synthesis module P908.

  The functions realized by executing the modules are the same as those of the components of the image predictive decoding apparatus 900 described above. That is, the function of each module of the image predictive decoding program P900 is the same as the functions of the data analyzer 902, the inverse quantizer 903, the inverse transformer 904, the adder 905, the frame memory 907, and the texture synthesizer 908.

  The image predictive encoding program P100 or the image predictive decoding program P900 configured as described above is stored in the recording medium 10 and executed by a computer described later.

  FIG. 17 is a diagram illustrating a hardware configuration of a computer for executing a program recorded in a recording medium, and FIG. 18 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. 17, the 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 programs stored therein, 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 programs. And. When the recording medium 10 is inserted into the reading device 12, the computer 30 can access the image predictive encoding / decoding program stored in the recording medium 10 from the reading device 12, and by the image encoding / decoding program, It becomes possible to operate as an image encoding device or an image decoding device according to the present invention.

  As shown in FIG. 18, the image predictive encoding program or the image decoding program may be provided via a network as a computer data signal 40 superimposed on a carrier wave. In this case, the computer 30 can store the image predictive encoding program or the image decoding program received by the communication device 24 in the memory 16 and execute the image predictive encoding program or the image predictive decoding program.

  As described above, according to the image predictive encoding device, the image predictive decoding device, the image predictive encoding method, the image predictive decoding method, the image predictive encoding program, and the image predictive decoding program according to the present invention, the extrapolation used in the prior art is used. When the intra-screen prediction signal is generated by this method, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target block and to efficiently predict an image signal having a complicated pattern.

  Next, the operational effects of the image predictive coding device 100 of the present embodiment will be described. In the image predictive coding device 100 according to the present embodiment, a block divider 102 divides an input image input from an input terminal 101 into a plurality of regions, and a target region that is a processing target of the plurality of divided regions. The prediction signal generator 103 generates a prediction signal for the target pixel signal. Then, the subtractor 105 generates a residual signal between the generated prediction signal and the target pixel signal, and the converter 106, the quantizer 107, and the entropy encoder 111 encode the residual signal and compress the compressed data. Can be generated.

  Here, the prediction signal generator 103 generates a texture signal based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal, and processes the generated texture signal using a predetermined texture synthesis method. Thus, a prediction signal can be generated. As a result, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target block, and to efficiently generate the prediction signal even when the texture signal is complicated. Can do.

  Further, the prediction signal generator 103 generates a plurality of texture signals having a high correlation with the adjacent area based on the adjacent area composed of the already reproduced pixel signals adjacent to the target pixel signal, and the generated plurality of texture signals. A prediction signal can be generated by processing using a predetermined synthesis method. This makes it possible to take advantage of the high correlation between the target block and the “target adjacent region” that touches the target block, prevents a decrease in prediction accuracy for pixels far from the boundary of the target block, and makes the texture signal complex. Even in this case, the prediction signal can be generated efficiently.

  Further, the prediction signal generator 103 generates and generates a plurality of texture signals having high correlation with each of the plurality of adjacent regions based on a plurality of adjacent regions composed of already reproduced pixel signals adjacent to the target pixel signal. The prediction signal can be generated by processing the plurality of texture signals using a predetermined synthesis method. This makes it possible to take advantage of the high correlation between the target block and the “target adjacent region” that touches the target block, prevents a decrease in prediction accuracy for pixels far from the boundary of the target block, and makes the texture signal complex. Even in this case, the prediction signal can be generated efficiently.

  Further, the prediction signal generator 103 generates an extrapolated texture signal by repeating the already reproduced pixel values adjacent to the target pixel signal in the first extrapolator 309 or the Nth extrapolator 310 to form pixels. The prediction signal can be generated by combining the texture signal and the extrapolation texture signal using a predetermined combining method. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target block, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  Further, the prediction signal generator 103 searches for a prediction adjacent region having a high correlation with respect to an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal from a search region that is a predetermined reproduced image, and performs the search. An image area determined based on the predicted adjacent area can be used as a texture signal. As a result, by utilizing the property that the target block and the “target adjacent region” adjacent to the target block are highly correlated, the region having the least error relative to the target adjacent region is detected as the predicted adjacent region from the already reproduced image region. By using the pixel group corresponding to the target block at the same time as adjacent to the adjacent area as the texture signal of the target block, a prediction signal having the same pattern as the target adjacent area can be generated even when the texture is complex. A texture signal with the same pattern can be generated even if it is far from the boundary.

  The entropy encoder 111 encodes related information indicating a texture synthesis method (for example, information indicating a weight value and which texture generator is used) and is encoded together with the encoded signal. Related information can be transmitted via the output terminal 112. As a result, the image texture synthesis method can be notified to the reception side, and the reception side can generate a prediction signal by using the informed texture synthesis method, so that prediction for pixels far from the boundary of the target block can be performed. A decrease in accuracy can be prevented, and a prediction signal can be efficiently generated even when the texture signal is complicated.

  The texture synthesizer 201 also synthesizes the texture signals generated by the texture generators 307 and 308, the first extrapolator 309, and the Nth extrapolator 310 with respect to the plurality of texture signals for the target pixel signal. (Weighters 303 to 306) can generate a prediction signal by performing a weighted average process using a predetermined weighting coefficient. Thereby, there is an effect that a prediction signal with a small statistical error can be generated by synthesizing (averaging) the texture signals generated by template matching.

  Furthermore, the problem of uncertainty in the case where there is no large difference between the difference values obtained by matching (template matching) with the already reproduced image area based on the “target adjacent area” is the problem of the texture signal generated by template matching. By combining (averaging), it is possible to generate a prediction signal with a small statistical error. Further, according to the present invention, by synthesizing a plurality of texture signals, there is an effect of suppressing coding distortion included in each texture, and a prediction signal with few errors can be generated.

  In addition, for example, the texture generator 307 may determine whether the adjacent region based on the adjacent region including the already reproduced pixel signal adjacent to the target pixel signal from the first search region within the same screen as the target pixel signal. At least one texture signal having a high correlation is generated. For example, the texture generator 308 generates from the already reproduced pixel signal adjacent to the target pixel signal from the second search area which is a screen different from the first search area. Generating a prediction signal by generating at least one texture signal having a high correlation with the adjacent area and combining the generated texture signals using a predetermined texture synthesis method. Can do. As a result, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target block, and to efficiently generate the prediction signal even when the texture signal is complicated. Can do.

  Further, the block divider 102 divides the input image input via the input terminal 101 into a plurality of regions, and predicts a prediction signal for the target pixel signal in the target region that is the processing target among the divided regions. The signal generator 103 generates. Then, a subtractor 105 generates a residual signal between the generated prediction signal and the target pixel signal, and a converter 106, a quantizer 107, and an entropy encoder 111 generate the generated residual signal. It can be encoded to produce a compressed signal. Further, the prediction signal generator 103 generates at least one extrapolated texture signal from the first extrapolator 309 and the Nth extrapolator by repeating the previously reproduced pixel values adjacent to the target pixel signal to form pixels. 310, and the synthesizer 301 synthesizes the extrapolation texture signal using a predetermined synthesis method, thereby generating a prediction signal. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target block, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  Next, operational effects of the image predictive decoding apparatus 900 will be described. In the image predictive decoding apparatus 900, the data analyzer 902 extracts a residual signal related to the target region to be processed from the compressed data input from the input terminal 901, and reproduces it in the inverse quantizer 903 and the inverse transformer 904. Restore to residual signal. Then, the texture synthesizer 908 generates a prediction signal for the target pixel signal in the target region, and adds the prediction signal generated by the adder 905 and the reproduction residual signal restored by the residual signal restoration means. Thus, the pixel signal of the target area can be restored, and the restored pixel signal can be stored in the frame memory 907 as a reference image. Then, a texture signal is generated based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal stored in the frame memory 907, and the generated texture signal is processed using a predetermined synthesis method. Thus, a prediction signal can be generated.

  Further, the texture synthesizer 908 generates a plurality of texture signals having high correlation with the adjacent region based on the adjacent region composed of the already reproduced pixel signals adjacent to the target pixel signal, and the generated plurality of texture signals in advance. A prediction signal can be generated by processing using a predetermined synthesis method. This makes it possible to take advantage of the high correlation between the target block and the “target adjacent region” that touches the target block, prevents a decrease in prediction accuracy for pixels far from the boundary of the target block, and makes the texture signal complex. Even in this case, the prediction signal can be generated efficiently.

  Further, the texture synthesizer 908 generates and generates a plurality of texture signals having high correlation with each of the plurality of adjacent regions based on a plurality of adjacent regions composed of already reproduced pixel signals adjacent to the target pixel signal. The prediction signal can be generated by processing the plurality of texture signals using a predetermined synthesis method. This makes it possible to take advantage of the high correlation between the target block and the “target adjacent region” that touches the target block, prevents a decrease in prediction accuracy for pixels far from the boundary of the target block, and makes the texture signal complex. Even in this case, the prediction signal can be generated efficiently.

  Further, the first extrapolator 1009 or the Nth extrapolator 1010 of the texture synthesizer 908 generates an extrapolated texture signal by repeating the already reproduced pixel values adjacent to the target pixel signal to form a pixel, The prediction signal can be generated by synthesizing the texture signal and the extrapolation texture signal using a predetermined synthesis method. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target block, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  Further, the texture synthesizer 908 searches for a predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a search area that is a previously reproduced image, and performs the search. An image area determined based on the predicted adjacent area can be used as a texture signal. As a result, by utilizing the property that the target block and the “target adjacent region” adjacent to the target block are highly correlated, the region having the least error relative to the target adjacent region is detected as the predicted adjacent region from the already reproduced image region. By using the pixel group corresponding to the target block at the same time as adjacent to the adjacent area as the texture signal of the target block, a prediction signal having the same pattern as the target adjacent area can be generated even when the texture is complex. A texture signal with the same pattern can be generated even if it is far from the boundary.

  In addition, the compressed data input from the input terminal 901 includes related information indicating a texture synthesis method, and the texture synthesizer 908 can form a texture signal of the target region using this related information. . As a result, the receiving side can know the texture synthesis method, and can generate a prediction signal using the informed texture synthesis method, thus preventing a decrease in prediction accuracy for pixels far from the boundary of the target block. Thus, even when the texture signal is complicated, the prediction signal can be generated efficiently.

  The texture synthesizer 908 can generate a prediction signal by synthesizing a plurality of texture signals for the target pixel signal by performing a weighted average process using a predetermined weighting coefficient. Thereby, there is an effect that a prediction signal with a small statistical error can be generated by synthesizing (averaging) the texture signals generated by template matching.

  Furthermore, the problem of uncertainty in the case where there is no large difference between the difference values obtained by matching (template matching) with the already reproduced image area based on the “target adjacent area” is the problem of the texture signal generated by template matching. By combining (averaging), it is possible to generate a prediction signal with a small statistical error. Further, according to the present invention, by synthesizing a plurality of texture signals, there is an effect of suppressing coding distortion included in each texture, and a prediction signal with few errors can be generated.

  Further, in the texture synthesizer 908, for example, the texture generator 1007 is based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal from the first search region in the same screen as the target pixel signal. Then, at least one texture signal having a high correlation with the adjacent area is generated, and for example, the texture generator 1008 is adjacent to the target pixel signal from a second search area that is a different screen from the first search area. Based on the adjacent area composed of the already reproduced pixel signals to be generated, at least one texture signal highly correlated with the adjacent area is generated, and the synthesizer 1001 uses a predetermined texture synthesis method for each of the generated texture signals. The prediction signal can be generated by synthesizing them. As a result, since the prediction signal is generated using the image texture synthesis method, it is possible to prevent a decrease in prediction accuracy for pixels far from the boundary of the target block, and to efficiently generate the prediction signal even when the texture signal is complicated. Can do.

  Further, in the image predictive decoding apparatus 900, the data analyzer 902 extracts the residual signal related to the target region from the compressed data, and the inverse quantizer 903 and the inverse transformer 904 restore the reproduced residual signal, A texture synthesizer 908 generates a prediction signal for the target pixel signal in the target region. Then, the adder 905 adds the prediction signal and the reproduction residual signal, so that the pixel signal in the target region can be restored. In the texture synthesizer 908, the first extrapolator 1009 and the Nth extrapolator 1010 repeat at least one pixel value already reproduced adjacent to the target pixel signal to form a pixel, thereby at least one extrapolation texture signal. , And the prediction signal can be generated by combining the extrapolated texture signal using a predetermined combining method. Thereby, it is possible to prevent a decrease in prediction accuracy for pixels far away from the boundary of the target block, and it is possible to efficiently generate a prediction signal even when the texture signal is complicated.

  In addition, according to this embodiment, it can also be grasped as the following image predictive encoding method and image predictive decoding method.

  That is, in the image predictive coding method, the prediction signal generation step generates a plurality of texture signals having a high correlation with the adjacent region based on the adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal. Preferably, the prediction signal is generated by processing the plurality of generated texture signals using a predetermined synthesis method.

  In the image predictive coding method, the prediction signal generation step includes a plurality of high correlations with each of the plurality of adjacent regions based on a plurality of adjacent regions composed of already reproduced pixel signals adjacent to the target pixel signal. It is preferable to generate the prediction signal by generating the texture signal and processing the plurality of generated texture signals using a predetermined synthesis method.

  Further, in the image predictive coding method, the prediction signal generation step further generates an extrapolation texture signal by repeating the already reproduced pixel values adjacent to the target pixel signal to form a pixel, and the texture signal It is preferable to generate the prediction signal by combining the extrapolation texture signal using a predetermined combining method.

  Further, in the image predictive coding method, the prediction signal generation step is a pre-reproduced image in which a predictive adjacent region having a high correlation with an adjacent region composed of pre-reproduced pixel signals adjacent to the target pixel signal is determined in advance. It is preferable to search from the search area and use an image area defined based on the searched predicted adjacent area as a texture signal.

  Further, in the image predictive decoding method, the prediction signal generation step generates a plurality of texture signals having high correlation with the adjacent region based on the adjacent region composed of the already reproduced pixel signals adjacent to the target pixel signal. It is preferable to generate the prediction signal by processing the plurality of generated texture signals using a predetermined synthesis method.

  Further, in the image predictive decoding method, the prediction signal generation step includes a plurality of high correlations with each of the plurality of adjacent regions based on a plurality of adjacent regions composed of already reproduced pixel signals adjacent to the target pixel signal. It is preferable to generate a prediction signal by generating a texture signal and processing the generated plurality of texture signals using a predetermined synthesis method.

  Further, in the image predictive decoding method, the prediction signal generation step further generates an extrapolated texture signal by repeating the already reproduced pixel values adjacent to the target pixel signal to form pixels, and the texture signal and It is preferable to generate the prediction signal by synthesizing the extrapolation texture signal using a predetermined synthesis method.

  Further, in the image predictive decoding method, the prediction signal generation step is a search that is a pre-reproduced image in which a prediction adjacent region having a high correlation with respect to an adjacent region composed of pre-reproduced pixel signals adjacent to the target pixel signal is predetermined It is preferable to search from an area and use an image area defined based on the searched predicted adjacent area as a texture signal.

Next, the features of the present invention will be listed.
[Invention 1]
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 to generate compressed data;
The prediction signal generation unit generates a texture signal based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and processes the generated texture signal using a predetermined texture synthesis method. An image predictive coding apparatus characterized by generating a prediction signal.
[Invention 2]
The prediction signal generation unit generates a plurality of texture signals having a high correlation with the adjacent region based on an adjacent region composed of already-reproduced pixel signals adjacent to the target pixel signal, and the generated plurality of texture signals are stored in advance. The image predictive coding apparatus according to claim 1, wherein the predictive signal is generated by processing using a predetermined synthesis method.
[Invention 3]
The prediction signal generation unit generates and generates a plurality of texture signals having high correlation with each of the plurality of adjacent regions based on a plurality of adjacent regions composed of already reproduced pixel signals adjacent to the target pixel signal. The image predictive coding apparatus according to claim 1, wherein a prediction signal is generated by processing a plurality of texture signals using a predetermined synthesis method.
[Invention 4]
The prediction signal generation unit further generates an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal,
4. The image predictive coding apparatus according to claim 1, wherein a prediction signal is generated by combining the texture signal and the extrapolation texture signal using a predetermined combining method. 5.
[Invention 5]
The predicted signal generation means searches for a predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a search area that is a predetermined reproduced image, and performs the search. The image predictive coding apparatus according to any one of claims 1 to 4, wherein an image region determined based on the predicted adjacent region is used as a texture signal.
[Invention 6]
The encoding means encodes related information indicating the texture synthesis method,
The image predictive coding apparatus according to any one of claims 1 to 5, further comprising sending means for sending the related information encoded together with the encoded signal encoded by the encoding means. .
[Invention 7]
The prediction signal generation means generates a prediction signal by synthesizing a plurality of texture signals for the target pixel signal by performing a weighted average process using a predetermined weighting coefficient. The image predictive coding device according to any one of 2 to 6.
[Invention 8]
The predicted signal generation means correlates with the adjacent area based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a first search area within the same screen as the target pixel signal. Generate at least one high texture signal,
At least one texture signal having a high correlation with the adjacent area is based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a second search area that is a different screen from the first search area. Generate
The image predictive coding apparatus according to any one of claims 2 to 7, wherein the prediction signal is generated by combining the generated texture signals using a predetermined texture combining method.
[Invention 9]
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 to generate a compressed signal;
The prediction signal generating means includes
Generating at least one extrapolated texture signal by repeating the previously reproduced pixel values adjacent to the target pixel signal to form a pixel;
An image predictive coding apparatus that generates a prediction signal by combining the at least one extrapolated texture signal using a predetermined combining method.
[Invention 10]
A residual signal restoring means for extracting a residual signal related to a target region to be processed from the compressed data and restoring it to a reproduced residual signal;
Prediction signal generating means for generating a prediction signal for the target pixel signal of the target region;
Image restoration means for restoring the pixel signal of the target area by adding the prediction signal generated by the prediction signal generation means and the reproduction residual signal restored by the residual signal restoration means;
With
The prediction signal generating means generates a texture signal based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and predicts by processing the generated texture signal using a predetermined texture synthesis method. An image predictive decoding apparatus characterized by generating a signal.
[Invention 11]
The prediction signal generation unit generates a plurality of texture signals having a high correlation with the adjacent region based on an adjacent region composed of already-reproduced pixel signals adjacent to the target pixel signal, and the generated plurality of texture signals are stored in advance. The image predictive decoding apparatus according to claim 10, wherein a prediction signal is generated by processing using a predetermined synthesis method.
[Invention 12]
The prediction signal generation unit generates and generates a plurality of texture signals having high correlation with each of the plurality of adjacent regions based on a plurality of adjacent regions composed of already reproduced pixel signals adjacent to the target pixel signal. The image predictive decoding apparatus according to claim 10, wherein a prediction signal is generated by processing a plurality of texture signals using a predetermined synthesis method.
[Invention 13]
The prediction signal generation unit further generates an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal,
13. The image predictive decoding apparatus according to claim 10, wherein the prediction signal is generated by combining the texture signal and the extrapolation texture signal using a predetermined combining method.
[Invention 14]
The predicted signal generation means searches for a predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a search area that is a predetermined reproduced image, and performs the search. The image predictive decoding apparatus according to any one of claims 10 to 13, wherein an image area determined based on the predicted adjacent area is used as a texture signal.
[Invention 15]
The compressed data includes related information indicating a texture synthesis method;
The image predictive decoding apparatus according to any one of claims 10 to 14, wherein the prediction signal generation unit forms a texture signal of the target region using the related information.
[Invention 16]
The prediction signal generation means generates a prediction signal by synthesizing a plurality of texture signals for the target pixel signal by performing a weighted average process using a predetermined weighting coefficient. The image predictive decoding device according to any one of 11 to 15.
[Invention 17]
The predicted signal generation means correlates with the adjacent area based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a first search area within the same screen as the target pixel signal. Generate at least one high texture signal,
At least one texture signal having a high correlation with the adjacent area is based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a second search area that is a different screen from the first search area. Generate
17. The image predictive decoding apparatus according to any one of claims 11 to 16, wherein a prediction signal is generated by combining the generated texture signals.
[Invention 18]
A residual signal restoring means for extracting a residual signal related to the target region from the compressed data and restoring it to a reproduced residual signal;
Prediction signal generating means for generating a prediction signal for the target pixel signal of the target region;
Image restoration means for restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal;
With
The prediction signal generating means includes
Generating at least one extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
An image predictive decoding apparatus, wherein a prediction signal is generated by combining the at least one extrapolated texture signal using a predetermined combining method.
[Invention 19]
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;
Encoding the residual signal generated by the residual signal generating step to generate compressed data, and
The predicted signal generation step generates a texture signal based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and processes the generated texture signal using a predetermined texture synthesis method. An image predictive encoding method characterized by generating a prediction signal.
[Invention 20]
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;
Encoding the residual signal generated by the residual signal generating step to generate a compressed signal, and
The predicted signal generation step includes:
Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
An image predictive coding method characterized in that a prediction signal is generated by synthesizing the extrapolation texture signal using a predetermined synthesis method.
[Invention 21]
A residual signal restoration step for extracting a residual signal related to a target region to be processed from the compressed data and restoring it to a reproduction residual signal;
A prediction signal generation step of generating a prediction signal for the target pixel signal of the target region;
An image restoration step of restoring the pixel signal of the target region by adding the prediction signal generated by the prediction signal generation step and the reproduction residual signal restored by the residual signal restoration step;
With
The prediction signal generation step generates a texture signal based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and processes the generated texture signal using a predetermined synthesis method to generate a prediction signal The image predictive decoding method characterized by producing | generating.
[Invention 22]
A residual signal restoration step of extracting a residual signal related to the target region from the compressed data and restoring it to a reproduced residual signal;
A prediction signal generation step of generating a prediction signal for the target pixel signal of the target region;
An image restoration step of restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal;
With
The predicted signal generation step includes:
Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
An image predictive decoding method, wherein a prediction signal is generated by combining the extrapolated texture signal using a predetermined combining method.
[Invention 23]
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 to generate compressed data;
The prediction signal generation module generates a texture signal based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and processes the generated texture signal using a predetermined texture synthesis method. An image predictive coding program for generating a prediction signal.
[Invention 24]
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 to generate a compressed signal;
The prediction signal generation module includes:
Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
An image predictive coding program that generates a prediction signal by synthesizing the extrapolation texture signal using a predetermined synthesis method.
[Invention 25]
A residual signal restoration module that extracts a residual signal related to a target region to be processed from the compressed data and restores it to a reproduction residual signal;
A prediction signal generation module that generates a prediction signal for a target pixel signal in the target region;
An image restoration module for restoring the pixel signal of the target area by adding the prediction signal generated by the prediction signal generation module and the reproduction residual signal restored by the residual signal restoration module;
With
The prediction signal generation module generates a texture signal based on an adjacent region composed of already reproduced pixel signals adjacent to the target pixel signal, and processes the generated texture signal using a predetermined synthesis method to generate a prediction signal An image predictive decoding program characterized by generating
[Invention 26]
A residual signal restoration module that extracts a residual signal related to the target region from the compressed data and restores it to a reproduction residual signal;
A prediction signal generation module that generates a prediction signal for a target pixel signal in the target region;
An image restoration module for restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal;
With
The prediction signal generation module includes:
Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
An image predictive decoding program that generates a prediction signal by combining the extrapolation texture signal using a predetermined combining method.

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 ... Texture synthesizer, 202 ... Texture signal determiner, 301 ... Synthesizer, 302 ... Adder, 307 ... Texture generator, 308 ... Texture generator, 309 ... First extrapolator, 310 ... Nth extrapolator, 900 ... Image prediction decoding device, 901 ... Input terminal, 902 ... Data analyzer, 903 ... Inverse quantum 904: Inverter, 905 ... Adder, 906 ... Output terminal, 907 ... Frame memory, 908 ... Texture synthesizer, 1001 ... Synthesizer, 007 ... texture generator, 1008 ... texture generator, 1009 ... first outer interpolator, 1010 ... N-th extrapolator.

Claims (20)

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 to generate compressed data;
The prediction signal generating means includes
A predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal is searched from a search area that is a predetermined reproduced image, and is determined based on the searched predicted adjacent area. Generating a plurality of image regions as a plurality of texture signals,
An image predictive coding apparatus that generates a prediction signal by processing a plurality of generated texture signals using a predetermined texture synthesis method.   The prediction signal generation unit further generates an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal,
  The image predictive coding apparatus according to claim 1, wherein the prediction signal is generated by combining the texture signal and the extrapolation texture signal using a predetermined combining method.   The encoding means encodes related information indicating the texture synthesis method,
  The image predictive coding apparatus according to claim 1, further comprising: a sending unit that sends the encoded related information together with the compressed data encoded by the encoding unit.   The prediction signal generation means generates a prediction signal by synthesizing a plurality of texture signals for the target pixel signal by performing a weighted average process using a predetermined weighting coefficient. Item 4. The image predictive encoding device according to any one of Items 1 to 3.   The predicted signal generation means correlates with the adjacent area based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a first search area within the same screen as the target pixel signal. Generating at least one high texture signal, based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a second search area, which is a screen different from the first search area; 5. The prediction signal is generated by generating at least one texture signal having a high correlation with a region and combining the generated texture signals using a predetermined texture combining method. The image predictive coding apparatus according to any one of the above.   Area dividing means for dividing an input image into a plurality of areas; and a prediction signal generating means for generating a prediction signal for a target pixel signal of a target area to be processed among the plurality of areas divided by the area dividing means; 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 encoding the residual signal generated by the residual signal generating means. Encoding means for generating a compressed signal, the prediction signal generating means,
  Generating at least one extrapolated texture signal by repeating the previously reproduced pixel values adjacent to the target pixel signal to form a pixel;
  An image predictive coding apparatus that generates a prediction signal by combining the at least one extrapolated texture signal using a predetermined combining method.   A residual signal restoring means for extracting a residual signal related to a target region to be processed from the compressed data and restoring it to a reproduced residual signal;
  Prediction signal generating means for generating a prediction signal for the target pixel signal of the target region;
  Image restoration means for restoring the pixel signal of the target area by adding the prediction signal generated by the prediction signal generation means and the reproduction residual signal restored by the residual signal restoration means,
  The prediction signal generating means includes
  A predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal is searched from a search area that is a predetermined reproduced image, and is determined based on the searched predicted adjacent area. Generating a plurality of image regions as a plurality of texture signals,
  An image predictive decoding apparatus that generates a prediction signal by processing a plurality of generated texture signals using a predetermined texture synthesis method.   The prediction signal generation unit further generates an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal,
  The image predictive decoding apparatus according to claim 7, wherein the prediction signal is generated by combining the texture signal and the extrapolation texture signal using a predetermined combining method.   The compressed data includes related information indicating a texture synthesis method;
  9. The image predictive decoding apparatus according to claim 7, wherein the prediction signal generation unit forms a texture signal of the target region using the related information.   The prediction signal generation means generates a prediction signal by synthesizing a plurality of texture signals for the target pixel signal by performing a weighted average process using a predetermined weighting coefficient. Item 10. The image predictive decoding device according to any one of Items 7 to 9.   The predicted signal generation means correlates with the adjacent area based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a first search area within the same screen as the target pixel signal. Generating at least one high texture signal, based on an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal from a second search area, which is a screen different from the first search area; The image prediction according to any one of claims 7 to 10, wherein a prediction signal is generated by generating at least one texture signal highly correlated with a region and synthesizing each of the generated texture signals. Decoding device.   A residual signal restoring means for extracting a residual signal related to the target region from the compressed data and restoring it to a reproduced residual signal;
  Prediction signal generation means for generating a prediction signal for the target pixel signal in the target area; and image restoration means for restoring the pixel signal in the target area by adding the prediction signal and the reproduction residual signal. Prepared,
  The prediction signal generating means includes
  Generating at least one extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
  An image predictive decoding apparatus, wherein a prediction signal is generated by combining the at least one extrapolated texture signal using a predetermined combining method.   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;
  Encoding the residual signal generated by the residual signal generating step to generate compressed data, and
  The predicted signal generation step includes:
  A predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal is searched from a search area that is a predetermined reproduced image, and is determined based on the searched predicted adjacent area. Generating a plurality of image regions as a plurality of texture signals,
  An image predictive encoding method, wherein a prediction signal is generated by processing a plurality of generated texture signals using a predetermined texture synthesis method.   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;
  Encoding the residual signal generated by the residual signal generating step to generate a compressed signal, and
  The predicted signal generation step includes:
  Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
  An image predictive coding method characterized in that a prediction signal is generated by synthesizing the extrapolation texture signal using a predetermined synthesis method.   A residual signal restoration step for extracting a residual signal related to a target region to be processed from the compressed data and restoring it to a reproduction residual signal;
  A prediction signal generation step of generating a prediction signal for the target pixel signal of the target region;
  An image restoration step for restoring the pixel signal of the target area by adding the prediction signal generated by the prediction signal generation step and the reproduction residual signal restored by the residual signal restoration step,
  The predicted signal generation step includes:
  A predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal is searched from a search area that is a predetermined reproduced image, and is determined based on the searched predicted adjacent area. Generating a plurality of image regions as a plurality of texture signals,
  An image predictive decoding method, wherein a prediction signal is generated by processing a plurality of generated texture signals using a predetermined texture synthesis method.   A residual signal restoration step for extracting a residual signal related to the target region from the compressed data and restoring it to a reproduced residual signal;
  A prediction signal generation step of generating a prediction signal for the target pixel signal of the target region;
  An image restoration step of restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal;
  The predicted signal generation step includes:
  Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
  An image predictive decoding method, wherein a prediction signal is generated by combining the extrapolated texture signal using a predetermined combining method.   In an image predictive encoding program executed by a computer,
  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 to generate compressed data;
  The prediction signal generation module includes:
  A predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal is searched from a search area that is a predetermined reproduced image, and is determined based on the searched predicted adjacent area. Generating a plurality of image regions as a plurality of texture signals,
  An image predictive coding program that causes a computer to generate a prediction signal by processing a plurality of generated texture signals using a predetermined texture synthesis method.   In an image predictive encoding program executed by a computer,
  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 to generate a compressed signal;
  The prediction signal generation module includes:
  Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
  An image predictive coding program for causing a computer to generate a prediction signal by synthesizing the extrapolated texture signal using a predetermined synthesis method.   In an image predictive decoding program executed by a computer,
  A residual signal restoration module that extracts a residual signal related to a target region to be processed from the compressed data and restores it to a reproduction residual signal;
  A prediction signal generation module that generates a prediction signal for a target pixel signal in the target region;
  An image restoration module that restores the pixel signal of the target region 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:
  A predicted adjacent area having a high correlation with respect to an adjacent area composed of already reproduced pixel signals adjacent to the target pixel signal is searched from a search area that is a predetermined reproduced image, and is determined based on the searched predicted adjacent area. Generating a plurality of image regions as a plurality of texture signals,
  An image predictive decoding program for causing a computer to generate a prediction signal by processing a plurality of generated texture signals using a predetermined texture synthesis method.   In an image predictive decoding program executed by a computer,
  A residual signal restoration module that extracts a residual signal related to the target region from the compressed data and restores it to a reproduction residual signal;
  A prediction signal generation module that generates a prediction signal for a target pixel signal in the target region;
  An image restoration module for restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal;
  The prediction signal generation module includes:
  Generating an extrapolated texture signal by forming a pixel by repeating a previously reproduced pixel value adjacent to the target pixel signal;
  An image predictive decoding program for causing a computer to generate a prediction signal by combining the extrapolation texture signal using a predetermined combining method.

Images