JP2011205693A - Image decoding apparatus and image encoding apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To precisely search for a locally decoded image when a predictive image to a processing object block to be encoded or decoded is generated by searching for the locally decoded image, and to reduce the amount of processing by decreasing the number of candidate blocks.SOLUTION: An image encoding apparatus and decoding apparatus include in-screen prediction parts 16 and 25. The in-screen prediction parts 16 and 25 include a derivation means (illustrated by 31) which derives a feature vector of the processing object block based on a pixel value in an area nearby, a search range determining means (illustrated by 33) which calculates a similarity degree by comparing the feature vector of the processing object block with a feature vector of a decoded block, regards a position of the decoded block corresponding to the feature vector having the highest similarity degree as a center and determines an area within a prescribed distance from the center as a search range, a prediction area determining means (illustrated by 34) which selects an area similar to the processing object block from among the search ranges as a prediction area, and a prediction image generating means (illustrated by 35) which generates a prediction image to the processing object block based on an image of the prediction area.

Description

  The present invention relates to an image decoding apparatus and an image encoding apparatus, and more specifically, decodes encoded data generated by dividing an image into a plurality of blocks and encoding a difference image between an input image and a predicted image for each block. The present invention relates to an image decoding apparatus that reproduces an image and an image encoding apparatus that generates encoded data thereof.

  In recent years, with the increase in the amount of image data, the importance of image encoding techniques that reduce the amount of image data by encoding images using the spatial correlation of images has increased.

  In particular, a block-based image encoding method for encoding an image by dividing an image into a plurality of small regions (blocks) is a moving image encoding method H.264. H.264 / AVC (Non-Patent Document 1) is widely used.

  The block-based image encoding method employed in Non-Patent Document 1 performs encoding using spatial correlation within a frame in a moving image. Therefore, the image coding method used in moving image coding is also called an intra-screen prediction method.

  In the intra prediction method described in Non-Patent Document 1, a locally decoded image of an adjacent block is used to generate a predicted image of a block to be encoded. However, in that case, only the correlation between the adjacent block and the encoding target block can be used for generating the predicted image. That is, it is not possible to use the correlation between a region located at a location away from the encoding target block and the encoding target block. Accordingly, there is a problem that the difference between the generated predicted image and the original image is large, that is, the accuracy of the predicted image is low.

  Therefore, an intra-screen prediction method has been proposed in which a predicted image is generated based on a locally decoded image in a region located at a location away from the encoding target block. In such a method, it is necessary to specify position information regarding a region used for generating a predicted image at the time of decoding. An example of a method for specifying position information is a method for encoding position information.

  Non-Patent Document 2 and Patent Document 1 propose a method that uses a search called template matching to specify the position information. In this method, the position information can be specified without encoding by using template matching. Therefore, the amount of encoded data can be reduced as compared with the case where position information is encoded.

  Here, the predicted image generation processing using template matching described in Non-Patent Document 2 and Patent Document 1 will be described in more detail. For the sake of simplicity, the following description will be made assuming that the image is divided into square blocks as shown in FIG. 1 and processed in the raster scan order from the upper left. Further, in the case of explaining processing in units of blocks, assuming that the kth block is a processing target, the block is referred to as a processing target block Bk. In addition, when the processing target block Bk is processed, an area where the encoding or decoding process is completed and the local decoded image exists is referred to as a decoded area.

  Hereinafter, the procedure of the predicted image generation process using template matching will be described with reference to the flowchart of FIG. First, the minimum evaluation value Dmin is set to the maximum value (step S101). Next, candidate areas to be processed are sequentially selected from among a plurality of candidate areas corresponding to the processing target block Bk (step S102). In addition, as candidate areas, it is assumed that there are N areas satisfying the two conditions of “the size and shape of the candidate area match those of the block Bk” and “the candidate area is located in the decoded area” (N number) ) Is used. Then, the processes in steps S103 to S105 up to the end of the loop (step S106) are performed on the selected candidate area (candidate area Rn).

  In step S103, the evaluation value Ds representing the similarity between the processing target block Bk and the candidate region Rn is calculated by the following method. A region having a predetermined shape associated with each block / candidate region is defined as a template for the block / candidate region. In the following, for simplicity, a key-type area including a pixel group outside the block adjacent to the upper side and the left side of the block / candidate area and a pixel located at the upper left of the block / candidate area as shown in FIG. It will be used as.

  As an amount representing the correlation between the pixel group on the template for the block Bk and the pixel group on the template for the candidate region Rn, an evaluation value Ds is obtained by the following equation.

Here, lBk (io, jo) and lRn (io, jo) are respectively io pixels in the horizontal direction and jo pixels in the vertical direction from the upper left pixel in the block / candidate area in the processing target block Bk or candidate area Rn. It represents the pixel value of a pixel located at a distant place. T is a set of relative positions with respect to the upper left pixel in the block / candidate area of each pixel on the template. For example, when the template has a key shape shown in FIG. 22, the set T is expressed by the following equation.
T = {(-1, -1), (0, -1), (1, -1), (2, -1), (3, -1), (-1,0), (-1, 1), (-1,2), (-1,3)}

  In step S104, if the evaluation value Ds has a value equal to or smaller than the minimum evaluation value Dmin set in step S101 or step S105, the process proceeds to step S105. Otherwise, the loop is terminated, the next candidate area is selected, and the process proceeds to step S103. In step S105, the value of the minimum evaluation value Dmin is set as the evaluation value Ds. Then, a candidate area Rn is set as a predicted image candidate. Thereafter, the loop is terminated, the next candidate area is selected, and the process proceeds to step S103. When N loops are completed, the predicted image candidate set in step S105 is output as the predicted image of the processing target block Bk (step S107).

  A predicted image for the processing target block Bk can be generated by the procedure described above. At this time, since the position of the area used as the predicted image is specified by template matching, it is not necessary to encode the position information of the area as side information. That is, by using template matching, a highly accurate predicted image can be generated based on a local decoded image in a region located away from the encoding target block without encoding extra side information. Efficiency can be improved.

  Patent Document 2 proposes a technique for searching for a predicted region position by block matching. FIG. 23 is a block diagram showing a configuration of an image encoding device in the prior art (Patent Document 2). An image encoding device 100 illustrated in FIG. 23 includes an orthogonal transform unit 101 that performs orthogonal transform on an input difference image, a quantization unit 102 that performs quantization of transform coefficients obtained as a result of the orthogonal transform, and a quantization unit. An inverse quantization unit 103 that performs inverse quantization corresponding to the inverse operation of 102, an inverse orthogonal transform unit 104 that performs inverse orthogonal transform corresponding to the inverse operation of the orthogonal transform unit 101, and a frame memory 105 that temporarily stores an image. . Furthermore, the image coding apparatus 100 includes an intra-screen prediction unit 106 that generates a prediction image for a processing target block, a prediction region search unit 107 that determines a prediction region position using an input image and a decoded image, and input data. Is provided with a variable length coding unit 108 that performs variable length coding.

  Next, the schematic operation of the image coding apparatus 100 will be described with reference to the flowchart shown in FIG. First, blocks constituting an image are selected in accordance with the encoding processing order (step S111), and the processing of steps S112 to S116 is performed up to the end of the loop (step S117) with the selected block as the encoding target block Bk.

  In step S112, the prediction region search unit 107 selects a region having a high correlation with the region corresponding to the processing target block Bk on the input image from the decoded image regions recorded in the frame memory 105, and predicts the prediction region position. Ask for. The correlation between the processing target block Bk and a certain area can be estimated by the square sum of errors between corresponding pixels in all the pixels in the area.

  In step S113, the obtained prediction region position is input to the intra prediction unit 106, and the intra prediction unit 106 generates a prediction image based on the region on the decoded image indicated by the prediction region position. In step S114, the prediction residual image, which is the difference between the block to be encoded and the prediction image generated in step S113, is input in the order of the orthogonal transform unit 101 and the quantization unit 102, and subjected to orthogonal transform and quantization processing. A transform coefficient level image is obtained.

  In step S115, the transform coefficient level image is input in the order of the inverse quantization unit 103 and the inverse orthogonal transform unit 104, and after the inverse quantization process and the inverse orthogonal transform are performed, the prediction image generated in step S113 and By adding together, a locally decoded image of the encoding target block Bk is obtained and recorded in the frame memory 105. In step S116, the transform coefficient level image and the prediction region position are also input to the variable length encoding unit 108, subjected to variable length encoding, and then output as encoded data. Then, the process ends when K loops corresponding to all the encoding target blocks are completed. Through the procedure described above, the image encoding device 100 can encode an input image and output encoded data corresponding to the input image.

  FIG. 25 is a block diagram showing a configuration of an image decoding device according to the prior art (Patent Document 2). An image decoding apparatus 110 illustrated in FIG. 25 includes an inverse quantization unit 112 and an inverse orthogonal transform unit having the same functions as the inverse quantization unit 103, the inverse orthogonal transform unit 104, the frame memory 105, and the intra-screen prediction unit 106 in FIG. 113, a frame memory 114, and an intra-screen prediction unit 115, and further includes a variable-length decoding unit 111 that performs a decoding process on variable-length-encoded input data.

  Next, the schematic operation of the image decoding apparatus 110 will be described with reference to the flowchart shown in FIG. First, decoding target blocks are sequentially selected according to the encoding processing order (step S121), and the processing of steps S122 to S125 is performed up to the end of the loop (step S126) using the selected block as the decoding target block Bk.

  In step S122, the variable length decoding unit 111 decodes the encoded data corresponding to Bk for the decoding target block, and reproduces the transform coefficient level image and the prediction region position. In step S123, the prediction region position information is input to the intra prediction unit 115, and a prediction image is generated based on the region on the decoded image indicated by the prediction region position. In step S124, the transform coefficient level image is input in the order of the inverse quantization unit 112 and the inverse orthogonal transform unit 113, subjected to inverse quantization processing and inverse orthogonal transform, and then added to the prediction image generated in step S123. As a result, a decoded image of the decoding target block Bk is generated. In step S125, the generated decoded image is recorded in the frame memory 105 as a local decoded image and is output as a decoded image. Then, the process ends when the K loops corresponding to the decoding target blocks are completed. Through the procedure described above, the encoded data generated by the image encoding device 100 can be input, and the decoded image corresponding to the encoded data can be output by the image decoding device 110.

  As described above, in the image encoding device described in Patent Document 2, a region having a high correlation with the processing target block is searched from the decoded image region by block matching, and the position of the region (predicted region position) is determined. It is encoded as the displacement of the position of the prediction region with respect to the position of the processing target block. Then, the image decoding device generates a prediction image of the processing target block based on the decoded prediction region position. As described above, the image encoding / decoding device described in Patent Document 2 can generate a predicted image based on a region located spatially away from the processing target block.

JP 2007-036551 A JP 2005-159947 A

ITU-T Recommendation H.264: "Advanced Video Coding for generic audiovisual services" (2003) T.-K. Tan et al. "Intra Prediction by Template Matching," International Conference of Image Processing 2006, Oct. 2006

  However, when performing predicted image generation processing using template matching as described in Non-Patent Document 2 and Patent Document 1, it is necessary to calculate evaluation values between a large number of candidate regions and processing target blocks. Therefore, there has been a problem that the amount of processing during encoding and decoding is large.

  As a method for reducing the template matching processing amount, a method of reducing the number of candidate regions is conceivable. For example, as indicated by hatching in FIG. 27, by setting a region on the decoded image within a certain distance from the processing target block Bk as a region where the candidate region (candidate block) Rn can exist, The number of candidate blocks can be reduced. In other words, it is possible to reduce the processing amount by defining candidate regions as regions that satisfy the following three conditions (I) to (III) in the above-described predicted image generation processing procedure using template matching. it can.

(I) The size and shape of the candidate area matches that of the block Bk.
(II) The candidate area is located within the decoded area.
(III) The position VRn of the candidate region Rn and the position VBk of the block Bk satisfy the following conditional expression.
‖VRn−VBk‖ ≦ LS1
Here, LS1 is a predetermined constant that determines the size of the search range by template matching. For example, a value such as LS1 = 16 [pixel] is set.

  However, when the candidate area is limited by the above-described method, the position of the candidate area is limited to the vicinity of the processing target block. For this reason, if there is a region having a feature similar to that of the processing target block at a position away from the processing target block by the distance LS1, that region cannot be used for derivation of the predicted image. Therefore, as shown in FIG. 27, when the range where the candidate block exists is limited, the accuracy of the predicted image generated is lower than when the range is not limited.

  On the other hand, in the image encoding device described in Patent Document 2, since the prediction region position is encoded based on the position of the processing target block, the processing is performed when the distance between the processing target block position and the prediction region position is large. There is a problem that the amount of code when the displacement from the position of the target block to the predicted region position is encoded becomes large.

  The present invention has been made in view of the actual situation as described above, and can accurately search when generating a predicted image for a block to be encoded or decoded from a locally decoded image, and An object of the present invention is to provide an image decoding device and an image encoding device capable of reducing the processing amount by reducing the number of candidate blocks.

  In order to solve the above-described problems, the first technical means of the present invention adds the difference image decoded from the encoded data to the predicted image, thereby obtaining each block in the image divided into a plurality of blocks. In an image decoding apparatus for reproducing an image, feature vector deriving means for deriving a feature vector of the processing target block based on a pixel value in a neighborhood region of the processing target block, and a feature vector corresponding to the processing target block and decoding A similarity is calculated by comparing the feature vector corresponding to the completed block, and a region within a predetermined distance from the center is searched with the position of the decoded block corresponding to the feature vector having the highest similarity as the center. Search range determining means for determining a range, and prediction for selecting a region similar to the processing target block from the search range as a prediction region A range determining means is obtained by comprising: a prediction image generation unit that generates a predicted image for the current block based on the image of the prediction region.

  According to a second technical means, in the first technical means, the image decoding apparatus, apart from the predicted image generating means, one or a plurality of different predicted image generating means for generating a predicted image for the processing target block; A prediction mode decoding means for decoding the prediction mode, selecting a prediction image generation means to be used according to a prediction mode decoded from the encoded data, and generating a prediction image by the selected prediction image generation means It is characterized by doing.

  A third technical means is the second technical means, wherein the prediction mode decoding means derives an estimated prediction mode for estimating a prediction mode of the processing target block, and the prediction mode of the processing target block and the estimated prediction mode Is read from the encoded data. If they match, the estimated prediction mode is decoded as the prediction mode of the processing target block. If they do not match, information indicating the prediction mode is encoded data. And the read prediction mode is decoded as the prediction mode of the processing target block.

  According to a fourth technical means, in an image encoding apparatus that encodes a difference image between an input image and a predicted image for each block in an image divided into a plurality of blocks, pixel values in a neighborhood region of the processing target block are obtained. Based on the feature vector deriving means for deriving the feature vector of the processing target block, the feature vector corresponding to the processing target block is compared with the feature vector corresponding to the decoded block, and the similarity is calculated. Centering on the position of the decoded block corresponding to the feature vector having the highest degree, search range determining means for determining a region within a predetermined distance from the center as a search range, and the processing target block from the search range Prediction region determining means for selecting a region similar to the prediction region, and prediction for the processing target block based on the image of the prediction region It is obtained by comprising: a prediction image generation means for generating an image.

  According to a fifth technical means, in the fourth technical means, the image encoding device comprises one or a plurality of different predicted image generating means for generating a predicted image for the processing target block separately from the predicted image generating means. A predicted image determination unit that selects one predicted image generation unit by comparing a predicted image generated by each predicted image generation unit included in the image encoding device with an input image, and the selected one predicted image A prediction mode encoding unit that encodes a prediction mode indicating a generation unit, wherein the prediction image generated by the selected one prediction image generation unit is a prediction image in the processing target block. It is what.

  A sixth technical means is the fifth technical means, wherein the prediction mode encoding means derives an estimated prediction mode for estimating a prediction mode of the processing target block, and the prediction mode of the processing target block and the estimated prediction Information on whether or not the modes match is encoded, and when the prediction mode does not match the estimated prediction mode, information on a prediction mode indicating the selected one prediction image generation unit is encoded. It is what.

  According to the present invention, in the image decoding device or the image encoding device, when the predicted image for the block to be decoded or encoded is searched for and generated from the locally decoded image, the search can be performed with high accuracy and the candidate can be searched. It is possible to reduce the processing amount by reducing the number of blocks.

In the image coding apparatus or image decoding apparatus which concerns on this invention, it is a figure for demonstrating an example of the process target block used as an encoding object or a decoding object, and an example of the process order of the process object block. It is a block diagram which shows the example of 1 structure of the image coding apparatus which concerns on the 1st Embodiment of this invention. It is a flowchart for demonstrating schematically the operation example in the image coding apparatus of FIG. It is a block diagram which shows the example of 1 structure of the image decoding apparatus which concerns on the 1st Embodiment of this invention. FIG. 5 is a flowchart for schematically explaining an operation example in the image decoding apparatus of FIG. 4. It is a block diagram which shows the example of 1 structure of the prediction part in a screen in FIG.2 and FIG.4. It is a flowchart for demonstrating an example of the estimated image generation operation | movement in the prediction part in a screen of FIG. It is a conceptual diagram which shows an example of a candidate area | region at the time of performing template matching in the prediction area | region determination part in FIG. It is a block diagram which shows the example of 1 structure of the image coding apparatus which concerns on the 2nd Embodiment of this invention. FIG. 10 is a flowchart for schematically explaining an operation example in the image encoding device of FIG. 9. It is a block diagram which shows the example of 1 structure of the image decoding apparatus which concerns on the 2nd Embodiment of this invention. FIG. 12 is a flowchart for schematically explaining an operation example in the image decoding apparatus in FIG. 11. It is a block diagram which shows the example of 1 structure of the prediction part in a screen in FIG.9 and FIG.11. It is a flowchart for demonstrating an example of the search center determination operation | movement in the prediction part in a screen of FIG. It is a block diagram which shows the example of 1 structure of the image coding apparatus which concerns on the 3rd Embodiment of this invention. FIG. 16 is a flowchart for schematically explaining an operation example in the image encoding device of FIG. 15. It is a block diagram which shows the example of 1 structure of the image decoding apparatus which concerns on the 3rd Embodiment of this invention. FIG. 18 is a flowchart for schematically explaining an operation example in the image decoding apparatus in FIG. 17. It is a block diagram which shows one structural example of the prediction part in a screen in FIG.15 and FIG.17. It is a flowchart for demonstrating an example of the estimated image generation operation | movement in the prediction part in a screen of FIG. It is a flowchart for demonstrating the predicted image generation operation | movement by template matching of a prior art. It is a figure which shows the example of the template shape utilized in the operation | movement of FIG. It is a block diagram which shows the structure of the image coding apparatus by a prior art. It is a flowchart for demonstrating operation | movement in the image coding apparatus of FIG. It is a block diagram showing the structure of the image decoding apparatus by a prior art. It is a flowchart for demonstrating operation | movement in the image decoding apparatus of FIG. It is a conceptual diagram for demonstrating the method to restrict | limit a candidate area | region with respect to the operation | movement of FIG.

(First embodiment)
With reference to FIG. 1, an image encoding device and an image decoding device according to the first embodiment of the present invention will be described based on FIGS. FIG. 1 is a diagram for explaining an example of a processing target block to be encoded or decoded and an example of a processing order of the processing target block in the image encoding device or the image decoding device according to the present invention. . In the following description of the present invention, as in the description of the prior art, the image is divided into square blocks as shown in FIG. 1, and the blocks are encoded in the raster scan order from the upper left on the image. It is assumed that processing or decryption processing is executed. In addition, the size of the block divided at that time is assumed to be 4 × 4 pixels for simplicity.

  FIG. 2 is a block diagram showing a configuration example of the image coding apparatus according to the first embodiment of the present invention, and FIG. 3 is a diagram for schematically explaining an operation example in the image coding apparatus of FIG. FIG.

  The image encoding device 1 illustrated in FIG. 2 is a device that encodes a difference image between an input image and a predicted image for each block in an image divided into a plurality of blocks. The image encoding device 1 includes an orthogonal transform unit 11 that performs orthogonal transform on an input difference image, a quantization unit 12 that quantizes transform coefficients obtained as a result of the orthogonal transform, and an inverse of the quantization unit 12. An inverse quantization unit 13 that performs inverse quantization corresponding to the operation, an inverse orthogonal transform unit 14 that performs inverse orthogonal transform corresponding to the inverse operation of the orthogonal transform unit 11, a frame memory 15 that temporarily stores an image, and a processing target block An intra-screen prediction unit 16 that generates a predicted image and a variable-length encoding unit 17 that performs variable-length encoding on input data are provided.

  Next, a schematic operation example of the image encoding device 1 will be described with reference to FIG. First, blocks constituting an image are selected in accordance with the encoding processing order (step S1), and the processing of steps S2 to S5 is performed up to the end of the loop (step S6) with the selected block as the encoding target block Bk.

  In step S2, the intra-screen prediction unit 16 generates a predicted image that estimates an image of an area corresponding to the position of the encoding target block Bk on the input image. The predicted image generation process in the intra-screen prediction unit 16 is a characteristic process in the present invention, and will be described in detail later. In step S3, the prediction residual image, which is the difference between the encoding target block Bk and the prediction image generated in step S2, is input in the order of the orthogonal transform unit 11 and the quantization unit 12, and subjected to orthogonal transform and quantization processing. A transform coefficient level image is obtained.

  In step S4, the transform coefficient level image is input in the order of the inverse quantization unit 13 and the inverse orthogonal transform unit 14, and after the inverse quantization process and the inverse orthogonal transform are performed, the prediction image generated in step S2 and By adding together, a locally decoded image of the encoding target block Bk is obtained and recorded in the frame memory 15. In step S5, the transform coefficient level image is also input to the variable length encoding unit 17, subjected to variable length encoding, and then output as encoded data. Then, the process ends when K loops corresponding to all the encoding target blocks are completed. Through the procedure described above, the image encoding device 1 can encode an input image and output encoded data corresponding to the input image.

  FIG. 4 is a block diagram showing a configuration example of the image decoding apparatus according to the first embodiment of the present invention, and FIG. 5 is a flowchart for schematically explaining an operation example in the image decoding apparatus of FIG. It is.

  The image decoding apparatus 2 illustrated in FIG. 4 is an apparatus that reproduces an image of each block in an image divided into a plurality of blocks by adding a difference image decoded from encoded data to a predicted image. The image decoding apparatus 2 includes an inverse quantization unit 22, an inverse orthogonal transform unit 23, a frame memory having the same functions as the inverse quantization unit 13, the inverse orthogonal transform unit 14, the frame memory 15, and the intra-screen prediction unit 16 in FIG. 24, an intra-screen prediction unit 25, and a variable-length decoding unit 21 that performs decoding processing on variable-length-encoded input data.

  Next, a schematic operation example of the image decoding device 2 will be described with reference to FIG. First, the decoding target block Bk is sequentially selected in accordance with the encoding processing order (step S11), and the processing of steps S12 to S15 until the end of the loop (step S16) is performed with the selected block as the decoding target block Bk.

  In step S12, the intra prediction unit 25 generates a prediction image corresponding to the decoding target block Bk. In step S13, the variable length decoding unit 21 generates a transform coefficient level image by decoding the encoded data corresponding to the decoding target block Bk. In step S14, the transform coefficient level image is input in the order of the inverse quantization unit 22 and the inverse orthogonal transform unit 23, subjected to inverse quantization processing and inverse orthogonal transform, and then added to the prediction image generated in step S12. Together, a decoded image of the decoding target block Bk is generated. In step S5, the generated decoded image is stored in the frame memory 24 as a local decoded image and is output as a decoded image. Then, the process is terminated when K loops corresponding to all the decoding target blocks are completed. With the procedure described above, the encoded data generated by the image encoding device 1 can be input, and the decoded image corresponding to the encoded data can be output by the image decoding device 2.

  The image encoding device 1 and the image decoding device 2, as exemplified by the in-screen prediction units 16 and 25, are both the following feature vector deriving unit, feature vector recording unit, search range determining unit, prediction region determining unit, and Predictive image generation means is provided. The feature vector deriving unit obtains a feature vector representing the property of the image near the processing target block based on the pixel value in the processing target block vicinity region. The feature vector recording unit records (stores) the feature vector obtained by the feature vector deriving unit. The search range determining means determines the search range by comparing the feature vector corresponding to the processing target block with the feature vector recorded in the feature vector recording unit. The prediction region determination means selects a region similar to the processing target block from the search range as the prediction region. The predicted image generation means generates a predicted image for the processing target block based on the image of the predicted region.

  Next, the in-screen prediction units 16 and 25 which are characteristic components in the image encoding device 1 shown in FIG. 2 and the image decoding device 2 shown in FIG. 4 will be described in detail with reference to FIGS. Explained. 6 is a block diagram showing an example of the configuration of the intra-screen prediction unit in FIGS. 2 and 4, and FIG. 7 is a flowchart for explaining an example of a predicted image generation operation in the intra-screen prediction unit of FIG. FIG. 8 is a conceptual diagram illustrating an example of a candidate area when template matching is executed by the prediction area determination unit in FIG. 6.

  The in-screen prediction units 16 and 25 illustrated in FIG. 6 are feature vectors (peripheral feature vectors) representing the properties of the peripheral region of the block based on the pixel values of the locally decoded image in a predetermined region around the specified block. And a peripheral feature vector calculation unit 31 that outputs an input peripheral feature vector in association with the position of the corresponding block. The peripheral feature vector calculation unit 31 is an example of a feature vector deriving unit, and the peripheral feature vector recording unit 32 is an example of a feature vector recording unit.

  In addition, the intra-screen prediction units 16 and 25 compare the peripheral feature vector for the processing target block with the peripheral feature vector recorded in the peripheral feature vector recording unit 32 to determine a search center for template matching. 33. The search center determination unit 33 is a part included in an example of the search range determination unit, and the search range determination unit determines a region within a predetermined distance from the search center determined (set) here as a search range ( Set as the search range).

  Furthermore, the intra-screen prediction units 16 and 25 perform prediction matching based on the input search center, thereby predicting a prediction region that outputs a position of a region (prediction region position) used for deriving a prediction image for the processing target block. Unit 34 and a predicted image generation unit 35 that generates a predicted image for the processing target block based on the input prediction region position and the decoded image. The prediction region determination unit 34 is an example of a prediction region determination unit. In this example, a region similar to the processing target block is selected as a prediction region from the search range by template matching. The predicted image generating unit 35 is an example of a predicted image generating unit.

  Next, a predicted image generation operation example in the in-screen prediction units 16 and 25 will be described with reference to FIGS. First, it is determined whether or not the processing target block Bk is positioned at the left end or the upper end of the screen (step S21). If it is positioned, the process proceeds to step S27. In other cases, steps S22 to S26 are sequentially executed.

  In step S22, the peripheral feature vector calculation unit 31 calculates a peripheral feature vector PBk for the processing target block Bk. At this time, it is assumed that the peripheral feature vector PBa for the block Ba is calculated by the following equation.

  Here, Q is a quantization step for quantizing each element of the peripheral feature vector PBa, and an arbitrary real number of 1 or more is set. Here, the meaning of the peripheral feature vector PBa derived by the above formula will be supplementarily described. FHa, m represents a transform coefficient when a pixel on the template that is in contact with the upper side of the block Ba is subjected to discrete cosine transform. That is, FHa, m is an amount representing the size of each frequency component in the horizontal direction of the area on the template that is in contact with the upper side of the block Ba. Similarly, FV, m is an amount representing the magnitude of each frequency component in the vertical direction of the area on the template that is in contact with the left side of the block Ba. Further, fHa, m, fVa, m are values obtained by quantizing FHa, m, FVa, m by the quantization step Q, respectively. Accordingly, the peripheral feature vector PBa is a vector including horizontal and vertical frequency components of the peripheral region of the block Ba.

  As described above, the peripheral feature vector calculation unit 31 uses the horizontal frequency component obtained by frequency conversion and quantization of the pixel value of the region adjacent to the upper end of the processing target block and the pixel value of the region adjacent to the left end of the processing target block as the frequency. A vector having the vertical frequency components obtained by the transformation and quantization as elements is a feature vector corresponding to the processing target block.

  In step S23, the search center determining unit 33 first selects a peripheral feature vector having properties similar to the peripheral feature vector PBk obtained in step S22 from the peripheral feature vectors recorded in the peripheral feature vector recording unit 32. To do. Specifically, for all the peripheral feature vectors recorded in the peripheral feature vector recording unit 32, the similarity S with the peripheral feature vector PBk is calculated, and the peripheral feature vector that maximizes the similarity S is selected. . Note that the similarity S between the two peripheral feature vectors PBa and PBb is calculated by the following equation.

  In step S23, subsequently, the search center determination unit 33 reads the position of the block associated with the selected peripheral feature vector from the peripheral feature vector recording unit 32, and uses the read block position as the search center Vc in template matching. To do.

  As described above, the search center determining unit 33 calculates the similarity by comparing the feature vector corresponding to the processing target block with the feature vector corresponding to the decoded block recorded in the peripheral feature vector recording unit 32. The position of the decoded block corresponding to the feature vector having the highest similarity is set at the center of the search range.

  In step S24, the prediction region determination unit 34 determines the prediction region position VR by performing template matching based on the search center position Vc obtained in step S23. The processing in step S24 is basically the same processing as the prediction region position determination processing (FIG. 21) in the intra prediction method using template matching described as the conventional technique. Add restrictions. That is, the candidate area is an area that satisfies the following three conditions (i) to (iii).

(I) The size and shape of the candidate area matches that of the block Bk.
(Ii) The candidate area is located within the decoded area.
(Iii) The position VRn of the candidate area and the position Vc of the search center obtained in step S23 satisfy the following conditional expression.
‖VRn−Vc‖ ≦ LS2

  Here, LS2 is a predetermined constant representing the width of the search range (the predetermined distance), and is determined and set by the search range determination means. For example, a value such as LS2 = 16 [pixel] is set. As shown in FIG. 8, the above formula means that the position of the candidate block B is limited within the distance LS from the search center position Vc.

  In step S25, the predicted image generation unit 35 outputs a 4 × 4 pixel region on the decoded image corresponding to the position indicated by the predicted region position VR obtained in step S24 as a predicted image. In step S26, the peripheral feature vector PBk is recorded in the peripheral feature vector recording unit 32 in association with the position of the processing target block, and the process ends.

  On the other hand, in step S27, an image composed of 4 × 4 pixels in which a predetermined value is set for each pixel value is output as a predicted image of the processing target block Bk, and the process ends. For example, an intermediate value in a range of possible pixel values can be used as the predetermined value.

  Through the above procedure, the in-screen prediction units 16 and 25 can generate a prediction image for the processing target block Bk. Since the position of the candidate area is limited to a part of the decoded image area in this procedure, the process required for the template matching process while maintaining the accuracy of the predicted image generated in the template matching as compared with the conventional technique. The amount can be reduced. In this procedure, a region having characteristics similar to the processing target block is set as a template matching search center by comparing peripheral feature vectors. Therefore, even when a region having an image similar to the processing target block is at a position spatially separated from the processing target block, the region can be used as a predicted image.

  Other recording methods, deletion of recorded information (slice-by-slice refresh), search range switching according to feature vector similarity, and other searches that can be employed in the prediction image generation processing in the first embodiment The center determination method will be described.

  As a recording method in the peripheral feature vector recording unit 32, other methods can be adopted. That is, in the predicted image generation process in the first embodiment described above, the peripheral feature vector PBk is recorded in the peripheral feature vector recording unit 32 in step S26. At this time, if a peripheral feature vector PBa that is the same as or similar to the peripheral feature vector PBk is already recorded, it is not always necessary to record both peripheral feature vectors. For example, the position information of the existing peripheral feature vector PBa and the block Ba may be overwritten with the position information of the peripheral feature vector PBk and the block Bk. In this case, there is an advantage that the memory capacity of the peripheral feature vector recording unit 32 can be reduced.

  The information recorded by the peripheral feature vector recording unit 32 may be deleted. In other words, in the predicted image generation processing in the first embodiment described above, the information recorded in the peripheral feature vector recording unit 32 is not deleted, but the present invention is not limited to this. For example, as an encoding unit / decoding unit, an area (slice) in which a plurality of processing target blocks in the processing order are grouped is considered, and an image is divided into a plurality of slices for encoding / decoding processing. Also good.

  Then, before starting (preferably immediately before) the encoding / decoding process of the processing target block having the earliest processing order in the slice, all information in the peripheral feature vector recording unit 32 is deleted. As a result, processing for each encoding unit / decoding unit becomes possible. In that case, by limiting the position of the candidate region in the template matching within the same slice, each slice can be independently encoded / decoded, which has the advantage that the parallelism of the encoding / decoding process is improved.

  The search range may be switched according to the similarity of the feature vectors. That is, in the search range determining means (explained as the process of the prediction region determining unit 34) in the first embodiment described above, the value LS2 representing the width of the search range is a predetermined constant. Not limited to. For example, the value LS2 may be determined according to the size of Smax with reference to the maximum value Smax of the similarity S obtained when the search center determination unit 33 determines the search center. Hereinafter, the determination procedure of LS2 will be described in detail.

First, two values LS2a and LS2b are prepared as candidates for LS2. Here, it is assumed that LS2a is larger than LS2b. Further, a threshold TH1 for determining the magnitude of Smax is determined in advance. Then, LS2 is determined by the following equation.
LS2 = Ls2a (Smax <TH1), Ls2b (Smax ≧ TH1)

  When Smax is small, it means that the degree of similarity between the characteristics of the peripheral area of the search center and the characteristics of the peripheral area of the processing target block is low. Conversely, when Smax is large, it means that the similarity is high. Further, when the similarity is high, there is a high possibility that there is an area close to the processing target block in the vicinity of the search center, so that the template matching search range can be set narrow. On the contrary, when the similarity is low, the possibility is low, so it is necessary to set a wide search range. Therefore, by setting LS2 according to the above-described equation, the search range can be narrowed when Smax is large, so that the processing amount required for template matching can be reduced.

  In this way, the search range determination means records the highest similarity among the similarities obtained by calculation, and determines the first distance when the lowest similarity value is less than a predetermined threshold value. If the distance is equal to or greater than a predetermined threshold, a second distance smaller than the first distance may be set as the predetermined distance.

  Further, as a search center determination method in the search center determination unit 33, other methods can be adopted. In other words, the search center determination unit 33 calculates the similarity S between all the peripheral feature vectors recorded in the peripheral feature vector recording unit 32 and the peripheral feature vector PBk of the processing target block Bk, respectively, thereby calculating the search center. Although the position is determined, the present invention is not limited to this.

  For example, step S23 (search center position determination process) in the search center determination unit 33 can be replaced with a process of sequentially performing the following steps S23-1 to S23-3 (not shown). First, in step S23-1, the similarity S0 between the peripheral feature vectors PBk and all the peripheral feature vectors recorded in the peripheral feature vector recording unit 32 is calculated. Here, the similarity S0 between the peripheral feature vector PBa and the peripheral feature vector PBb is calculated by the following equation.

  Then, peripheral feature vectors whose similarity S0 is equal to or less than a predetermined threshold TH0 are recorded as vectors having high similarity, and peripheral feature vectors whose similarity S0 is less than the threshold TH0 are recorded as vectors having low similarity.

  In step S23-2, the similarity S between the peripheral feature vectors PBk is calculated for all the peripheral feature vectors recorded as vectors having a high similarity in step S23-1. The similarity S is obtained by the calculation formula described in step S23 in the search center determination unit 33.

  In step S23-3, the peripheral feature vector that maximizes the similarity S obtained in step S23-2 is selected, and the position of the block associated with the peripheral feature vector is read from the peripheral feature vector recording unit 32. The position of the read block is set as a search center Vc in template matching.

  The processing in steps S23-1 to S23-3 described above is performed by narrowing down search center position candidates by comparing using partial vectors of the peripheral feature vectors, and then comparing the entire peripheral feature vectors. It corresponds to the process of determining. In this process, by calculating the similarity S0 with a small processing amount required for calculation first and narrowing down the search center position candidates, the number of calculations of the similarity S having a large processing amount can be reduced. Therefore, it is possible to reduce the amount of processing necessary for determining the search center position.

  As described above, the search center determining unit 33 compares the feature vector corresponding to the processing target block with the feature vector corresponding to the decoded block recorded in the peripheral feature vector recording unit 32 to obtain the first similarity. Only a feature vector corresponding to a decoded block whose first similarity is higher than a predetermined threshold set in advance is calculated as a second similarity between the feature vector corresponding to the processing target block (first (Similarity derived by calculation more complicated than the similarity of the second) may be calculated, and the position of the decoded block corresponding to the feature vector having the highest second similarity may be set at the center of the search range. .

(Second Embodiment)
An image encoding device and an image decoding device according to the second embodiment of the present invention will be described with reference to FIGS. In the image coding apparatus according to the first embodiment, the prediction region position is determined by performing template matching based on the determined search center position, but the image coding according to the second embodiment described here is performed. In the apparatus, the predicted region position is searched by block matching based on the determined search center position, and the displacement of the predicted region position from the search center position is encoded. As described above, first, the search center is determined based on the peripheral feature amount of the block, and the displacement from the search center to the prediction region position is encoded, thereby reducing the amount of code required for encoding the prediction region position. it can.

  FIG. 9 is a block diagram showing a configuration example of an image encoding device according to the second embodiment of the present invention, and FIG. 10 is a diagram for schematically explaining an operation example in the image encoding device of FIG. FIG.

  The image encoding device 4 illustrated in FIG. 9 includes an orthogonal transform unit 41 having the same functions as the orthogonal transform unit 11, the quantization unit 12, the inverse quantization unit 13, the inverse orthogonal transform unit 14, and the frame memory 15 in FIG. , A quantization unit 42, an inverse quantization unit 43, an inverse orthogonal transform unit 44, and a frame memory 45.

  Further, the image encoding device 4 determines the search center in the prediction region search using the input image and the decoded image as input, and generates a prediction image based on the region on the decoded image indicated by the input prediction region position An intra-screen prediction unit 55, an input image, a decoded image, and a search center as inputs, a prediction region search unit 47 that determines a prediction region position, and a variable-length encoding unit 48 that performs variable-length encoding on input data With. The prediction region search unit 47 and the intra-screen prediction unit 46 include a prediction region determination unit on the encoding side that selects a region similar to the processing target block from the search range as a prediction region by block matching, and an image of the prediction region. It is an example with the prediction image production | generation means which produces | generates a prediction image based on.

  Next, a schematic operation example of the image encoding device 4 will be described with reference to FIG. First, blocks constituting an image are selected in accordance with the encoding process order (step S31), and the processes of steps S32 to S37 are performed up to the end of the loop (step S38) with the selected block as the encoding target block Bk.

  In step S32, the in-screen prediction unit 46 determines a search center having characteristics close to the peripheral region of the processing target block Bk. This process is a characteristic process in the present invention, and details will be described later. In step S33, the prediction region search unit 47 uses the decoded image recorded in the frame memory 45 in the frame memory 45 as a region having a high correlation with the processing target block Bk on the input image, with the search center obtained in step S32 as the center. The position of the selected area is set as the predicted area position.

  In step S34, the predicted region position is input to the intra prediction unit 46, and a predicted image is generated based on the region on the decoded image indicated by the predicted region position. In this way, the prediction region search unit 47 and the in-screen prediction unit 46 select a region similar to the processing target block as a prediction region from the search range by block matching. In step S35, a prediction residual image that is a difference between the encoding target block Bk and the prediction image generated in step S34 is input in the order of the orthogonal transform unit 41 and the quantization unit 42, and orthogonal transform and quantization processing are performed. To obtain a transform coefficient level image.

  In step S36, the transform coefficient level image is input in the order of the inverse quantization unit 43 and the inverse orthogonal transform unit 44, subjected to inverse quantization processing and inverse orthogonal transform, and then added to the prediction image generated in step S34. By combining them, a locally decoded image of the encoding target block Bk is obtained and recorded in the frame memory 45.

  In step S37, the transform coefficient level image and the prediction region position are also input to the variable length encoding unit 48, subjected to variable length encoding, and then output as encoded data. Here, the predicted region position is encoded as a relative position from the search center. Thus, the displacement from the center of the search range to the region (prediction region) similar to the processing target block is also encoded and output by the image encoding device 4. Then, the process ends when K loops corresponding to all the encoding target blocks are completed. By the procedure described above, the image encoding device 4 can encode an input image and output encoded data corresponding to the input image.

  FIG. 11 is a block diagram showing a configuration example of an image decoding apparatus according to the second embodiment of the present invention, and FIG. 12 is a flowchart for schematically explaining an operation example in the image decoding apparatus of FIG. It is.

  The image decoding apparatus 5 illustrated in FIG. 11 includes an inverse quantization unit 52, an inverse orthogonal transform, which have the same functions as the inverse quantization unit 43, the inverse orthogonal transform unit 44, the frame memory 45, and the in-screen prediction unit 46 in FIG. Unit 53, frame memory 54, and in-screen prediction unit 55.

  Further, the image decoding device 5 receives the relative position from the search center of the prediction region and the search center as input, and the prediction region position calculation unit 56 that determines the prediction region position, and the variable length encoded input data And a variable length decoding unit 51 that performs a decoding process. The predicted region position calculation unit 56 determines (selects) the position of the region similar to the processing target block based on the displacement from the center of the search range to the region similar to the processing target block obtained by decoding from the encoded data. This is an example of a prediction area determination unit on the decoding side.

  Next, a schematic operation example of the image decoding device 5 will be described with reference to FIG. First, the decoding target block Bk is sequentially selected according to the encoding processing order (step S41), and the processing of steps S42 to S47 until the end of the loop (step S48) is performed with the selected block as the decoding target block Bk.

  In step S42, the variable length decoding unit 51 decodes the encoded data corresponding to the decoding target block Bk, and reproduces the relative position from the search center of the transform coefficient level image and the prediction region.

  In step S43, the in-screen prediction unit 55 determines the search center of the prediction region. This process is a characteristic process in the present invention, and details will be described later. In step S44, the prediction region position calculation unit 56 determines the prediction region position by adding the relative position from the search center of the prediction region reproduced in step S42 to the search center determined in step S43. To do. Thus, the prediction region position calculation unit 56 is based on the displacement (relative position) from the center of the search range to the region similar to the processing target block obtained by decoding from the encoded data in step S42. The position of a region (predicted region) similar to the processing target block is determined.

  In step S45, the predicted region position is input to the intra prediction unit 55, and a predicted image is generated based on the region on the decoded image indicated by the predicted region position. In step S46, the transform coefficient level image is input in the order of the inverse quantization unit 52 and the inverse orthogonal transform unit 53, subjected to inverse quantization processing and inverse orthogonal transform, and then added to the predicted image generated in step S42. Together, a decoded image of the decoding target block Bk is generated.

  In step S47, the generated decoded image is recorded in the frame memory 54 as a local decoded image and is output as a decoded image. Then, the process is terminated when K loops corresponding to all the decoding target blocks are completed. Through the procedure described above, the encoded data generated by the image encoding device 4 can be input, and the decoded image corresponding to the encoded data can be output by the image decoding device 5.

  Next, the in-screen prediction units 46 and 55 that are characteristic components of the image encoding device 4 shown in FIG. 9 and the image decoding device 5 shown in FIG. 11 will be described in detail with reference to FIGS. Explained. 13 is a block diagram showing an example of the configuration of the intra-screen prediction unit in FIGS. 9 and 11, and FIG. 14 is a flowchart for explaining an example of the search center determination operation in the intra-screen prediction unit of FIG. is there.

  The intra-screen prediction units 46 and 55 illustrated in FIG. 13 have peripheral functions having the same functions as the peripheral feature vector calculation unit 31, the peripheral feature vector recording unit 32, the search center determination unit 33, and the predicted image generation unit 35 in FIG. A vector calculation unit 61, a peripheral feature vector recording unit 62, a search center determination unit 63, and a predicted image generation unit 65 are provided.

  Next, an example of the search center determination operation in the in-screen prediction units 46 and 55 will be described with reference to FIG. First, it is determined whether or not the processing target block Bk is positioned at the left end or the upper end of the screen (step S51). If it is positioned, the process proceeds to step S56. In other cases, steps S52 to S55 are sequentially executed.

  In step S52, the peripheral feature vector calculation unit 61 calculates a peripheral feature vector PBk for the processing target block Bk. At this time, the peripheral feature vector PBa for the block Ba is calculated by the method described in the first embodiment. In step S53, the peripheral feature vector PBk derived in step S52 is recorded in the peripheral feature vector recording unit 62.

  In step S54, the search center determination unit 63 first selects a peripheral feature vector having properties similar to the peripheral feature vector PBk obtained in step S52 from the peripheral feature vectors recorded in the peripheral feature vector recording unit 62. To do. The selection method at this time is the method described in the first embodiment.

  In step S55, the search center determining unit 63 reads the position of the block associated with the selected peripheral feature vector from the peripheral feature vector recording unit 62. Then, the read block position is output as the search center, and the process ends.

  On the other hand, in step S56, the position of the processing target block Bk is output as the search center, and the process ends. By the procedure described above, the in-screen prediction units 46 and 55 determine the search center based on the peripheral feature vector of the processing target block Bk and the peripheral feature vector recorded in the peripheral feature vector recording unit 62.

  As described above, the image coding apparatus according to the second embodiment can generate a predicted image based on a region at a position spatially separated from the processing target block. Further, in this image encoding device, information for specifying the prediction region position is encoded as a displacement from the search center to the prediction region position. Here, since the characteristics of the area near the processing target block are close to the characteristics of the area around the search center, there is a high probability that the displacement from the search center to the predicted area position is small. For this reason, since the variation of the displacement is reduced, the displacement can be encoded with a small amount of data. Therefore, the image encoding device according to the second embodiment can encode information representing the prediction region position with a small amount of data.

(Third embodiment)
An image encoding device and an image decoding device according to the third embodiment of the present invention will be described with reference to FIGS. In the image encoding / decoding devices according to the first and second embodiments, a predicted image is generated by the only intra-screen prediction method. However, the image encoding device according to the third embodiment described here has a plurality of screens. An optimal method is selected from the intra prediction methods, and a predicted image is generated by the selected method. That is, in this image encoding apparatus, a predicted image determination in which at least two reserved image generation units in total and one predicted image generation unit are selected by comparing the input image with the predicted image generated by each predicted image generation unit. A predicted image generated by one selected predicted image generating unit is used as a predicted image in the processing target block. Further, the image encoding device includes a prediction mode encoding unit that encodes a prediction mode indicating one selected prediction image generation unit so that decoding is possible.

  In the following description, information for specifying each intra-screen prediction method is referred to as a prediction mode. Unless otherwise specified, one prediction method is selected for each block from the following four intra-screen prediction methods to generate a predicted image.

Prediction method using template matching (TM prediction)
Prediction method using the average value of pixels adjacent to the processing target block (DC prediction)
Prediction method using pixel values of pixels adjacent to the left side of the processing target block (H prediction)
Prediction method (V prediction) using pixel values of pixels adjacent to the upper side of the processing target block

  FIG. 15 is a block diagram showing a configuration example of an image encoding device according to the third embodiment of the present invention, and FIG. 16 is a diagram for schematically explaining an operation example in the image encoding device of FIG. FIG.

  An image encoding device 7 illustrated in FIG. 15 includes an orthogonal transform unit 71 having the same functions as the orthogonal transform unit 11, the quantization unit 12, the inverse quantization unit 13, the inverse orthogonal transform unit 14, and the frame memory 15 in FIG. , A quantization unit 72, an inverse quantization unit 73, an inverse orthogonal transform unit 74, and a frame memory 75. Furthermore, the image encoding device 7 selects one prediction image by comparing each of the plurality of input prediction images with the input image, and outputs a prediction mode corresponding to the prediction image. 77, an intra-screen prediction unit 76 that generates a predicted image using a decoded image based on the input prediction mode, and a variable-length encoding unit 78 that performs variable-length encoding on input data. Prepare. Here, the intra-screen prediction unit 76 has a total of at least two reserved image generation means.

  Next, a schematic operation example of the image encoding device 7 will be described with reference to FIG. First, blocks constituting an image are selected in accordance with the encoding processing order (step S61), and the processing of steps S62 to S67 up to the end of the loop (step S68) is performed with the selected block as the encoding target block Bk.

  In step S62, the in-screen prediction determination unit 77 inputs each prediction mode of TM prediction, DC prediction, H prediction, and V prediction to the in-screen prediction unit 76, and obtains predicted images Im_TM, Im_DC, Im_H, and Im_V. The details of the prediction image generation process corresponding to each prediction mode in the in-screen prediction unit 76 will be described later.

  In step S63, the in-screen prediction determination unit 77 calculates an evaluation value R for the input image of each prediction image obtained in step S62 by the following formula, and selects a prediction mode corresponding to the prediction image that minimizes the evaluation value. Set as the prediction mode to be applied to the processing target block.

  Here, lIm (i, j) represents the pixel value of each pixel in the predicted image (4 × 4 pixel block) to be evaluated. LBk (i, j) represents the pixel value of each pixel in the processing target block Bk in the input image.

  In step S64, the in-screen prediction unit 76 generates a predicted image corresponding to the prediction mode set in step S63. In step S65, the prediction residual image, which is the difference between the encoding target block Bk and the prediction image generated in step S64, is input in the order of the orthogonal transform unit 71 and the quantization unit 72, and orthogonal transform and quantization processing are performed. To obtain a transform coefficient level image.

  In step S66, the transform coefficient level image is input in the order of the inverse quantization unit 73 and the inverse orthogonal transform unit 74, and after the inverse quantization process and the inverse orthogonal transform are performed, the prediction image generated in step S64 and By adding together, a locally decoded image of the encoding target block Bk is obtained and recorded in the frame memory 75.

  In step S67, the transform coefficient level image and the prediction mode set in step S63 are also input to the variable length encoding unit 78, subjected to variable length encoding, and then output as encoded data. Then, the process ends when K loops corresponding to all the encoding target blocks are completed. Through the procedure described above, the image encoding device 7 can encode an input image and output encoded data corresponding to the input image.

  FIG. 17 is a block diagram showing a configuration example of an image decoding apparatus according to the third embodiment of the present invention, and FIG. 18 is a flowchart for schematically explaining an operation example in the image decoding apparatus of FIG. It is.

  The image decoding apparatus 8 illustrated in FIG. 17 includes an inverse quantization unit 82, an inverse orthogonal transform each having the same functions as the inverse quantization unit 73, the inverse orthogonal transform unit 74, the frame memory 75, and the in-screen prediction unit 76 in FIG. Unit 83, frame memory 84, and intra prediction unit 85, and further includes a variable length decoding unit 81 having the same function as the variable length decoding unit 21 in FIG. Here, the intra-screen prediction unit 85 has a total of at least two reserved image generation means.

  Next, a schematic operation example of the image decoding device 8 will be described with reference to FIG. First, the decoding target block Bk is sequentially selected in accordance with the encoding processing order (step S71), and the processing of steps S72 to S75 up to the end of the loop (step S76) is performed using the selected block as the decoding target block Bk.

  In step S72, the variable length decoding unit 81 decodes the encoded data corresponding to the decoding target block Bk, and reproduces the transform coefficient level image and the prediction mode. As described above, the image decoding device 8 includes prediction mode decoding means for decoding the prediction mode. In step S73, the in-screen prediction unit 85 uses the decoded image recorded in the frame memory 84 to generate a prediction image corresponding to the prediction mode reproduced in step S72. As described above, the image decoding device 8 selects a prediction image generation unit to be used according to the prediction mode decoded from the encoded data, and generates a prediction image in the selected prediction mode.

  In step S74, the transform coefficient level image reproduced in step S72 is input in the order of the inverse quantization unit 82 and the inverse orthogonal transform unit 83, subjected to inverse quantization processing and inverse orthogonal transform, and then generated in step S73. The decoded image of the decoding target block Bk is generated by adding the predicted image.

  In step S75, the generated decoded image is recorded in the frame memory 84 as a local decoded image and is output as a decoded image. Then, the process is terminated when K loops corresponding to all the decoding target blocks are completed. According to the procedure described above, the encoded data generated by the image encoding device 7 can be input, and the decoded image corresponding to the encoded data can be output by the image decoding device 8.

  Next, the in-screen prediction units 76 and 85 which are characteristic components in the image encoding device 7 shown in FIG. 15 and the image decoding device 8 shown in FIG. 17 will be described in detail with reference to FIGS. 19 and 20. Explained. 19 is a block diagram showing an example of the configuration of the intra-screen prediction unit in FIGS. 15 and 17, and FIG. 20 is a flowchart for explaining an example of the predicted image generation operation in the intra-screen prediction unit of FIG. is there.

  The intra-screen prediction units 76 and 85 illustrated in FIG. 19 are respectively the peripheral feature vector calculation unit 31, the peripheral feature vector recording unit 32, the search center determination unit 33, the prediction region determination unit 34, and the predicted image generation unit 35 in FIG. A peripheral feature vector calculation unit 91, a peripheral feature vector recording unit 92, a search center determination unit 93, a prediction region determination unit 94, and a predicted image generation unit 95 having the same function are provided.

  Furthermore, the intra-screen prediction units 76 and 85 generate a predicted image by DC prediction using a local decoded image, a DC predicted image generation unit 96 that generates a predicted image by DC prediction, and generate a predicted image by H prediction using the local decoded image. A predicted image generation unit 97, a V predicted image generation unit 98 that generates a predicted image based on V prediction using a locally decoded image, and a switch 99 for selecting a corresponding predicted image according to an input prediction mode. With.

  Next, an example of a predicted image generation operation in the in-screen prediction units 76 and 85 will be described with reference to FIG. First, the peripheral feature vector calculation unit 91 calculates a peripheral feature vector PBk for the processing target block Bk (step S81). Since a specific calculation method of the peripheral feature vector PBk uses the method described in the first embodiment, the details are omitted. Next, the peripheral feature vector calculation unit 91 passes the peripheral feature vector PBk to the peripheral feature vector recording unit 92, and the peripheral feature vector recording unit 92 records the peripheral feature vector PBk in association with the position of the processing target block Bk. (Step S82).

  Subsequently, the internal switch is set so that the switch 99 outputs an input from any one of the predicted image generation units 95 to 98 according to the prediction mode input as a result selected by the evaluation of the prediction mode value. Switching (step S83). In step S83, when the prediction mode is TM prediction, the output is switched so that the output of the predicted image generation unit 95 becomes a predicted image, and the process proceeds to step S84. When the prediction mode is DC prediction, the output is switched so that the output of the DC predicted image generation unit 96 becomes a predicted image, and the process proceeds to step S85. If the prediction mode is H prediction, the output is switched so that the output of the H predicted image generation unit 97 becomes a predicted image, and the process proceeds to step S86. If the prediction mode is V prediction, the output is switched so that the output of the V predicted image generation unit 98 becomes a predicted image, and the process proceeds to step S87.

  In step S <b> 84, the search center determination unit 93 determines the search center Vc in template matching and outputs it to the prediction region determination unit 94. Then, the prediction region determination unit 94 determines the prediction region position VR from the search center Vc and outputs it to the prediction image generation unit 95. Further, the predicted image generation unit 95 generates a predicted image from the predicted region position VR, outputs the generated predicted image via the switch 99, and ends the process. In addition, about the specific process in step S84, since it is the same as that of step S22-S25 (FIG. 7) of the estimated image generation process in 1st Embodiment, it abbreviate | omits for details.

  In step S85, the DC predicted image generation unit 96 generates a predicted image Im_DC using the following equation, outputs the generated predicted image Im_DC via the switch 99, and ends the process. In the following equation, lIm_DC (i, j) represents the pixel value of each pixel in the predicted image Im_DC composed of 4 × 4 pixels.

In step S86, the H predicted image generation unit 97 generates a predicted image Im_H by the following equation, outputs the generated predicted image Im_H via the switch 99, and ends the process. In the following equation, i = 0, 1, 2, 3, j = 0, 1, 2, 3, and lIm_H (i, j) represents the pixel value of each pixel in the predicted image Im_H composed of 4 × 4 pixels. .
lIm_H (i, j) = lBk (−1, j)

In step S87, the V predicted image generation unit 98 generates a predicted image Im_V according to the following equation, outputs the generated predicted image Im_V via the switch 99, and ends the process. In the following expression, i = 0, 1, 2, 3, j = 0, 1, 2, 3, and lIm_V (i, j) represents the pixel value of each pixel in the predicted image Im_V composed of 4 × 4 pixels. .
lIm_V (i, j) = lBk (i, -1)

  Through the above processing, the in-screen prediction units 76 and 85 can generate and output a prediction image corresponding to the input prediction mode. Thus, in the image coding device according to the third embodiment, one prediction method is selected from a plurality of prediction methods including prediction using template matching, and a prediction image based on the selected prediction method is used. Thus, the input image can be encoded. The prediction method using template matching is a method that is highly effective when the correlation between an area located spatially away from the target block and the image on the target block is high. On the other hand, when the correlation between the region adjacent to the processing target block and the image on the processing target block is high, the prediction method based on the image of the adjacent block can be expected to have a higher effect. Therefore, a high effect can be expected by selecting an optimum method from both methods for each block. That is, it is possible to encode the input image with a high compression rate.

  Examples of other prediction modes that can be employed in the predicted image generation processing in the third embodiment will be described. The image encoding device 7 according to the third embodiment selects four prediction methods of TM prediction, DC prediction, H prediction, and V prediction as candidates, and selects one method for each block from the prediction method candidates. Although the predicted image is generated, the present invention is not limited to this. For example, the prediction method in the image encoding device 4 in FIG. 9 described in the second embodiment or the image encoding device 100 in FIG. 23 described in the related art may be added as a prediction method candidate. Or H. An intra-screen prediction method defined in H.264 / AVC may be added as a prediction method candidate.

  In the predictive image generation process according to the third embodiment, it has been described that the variable length encoding unit 78 encodes the prediction mode for each block in the image encoding device 7. For example, the prediction mode is generated by the following procedure. Information may be encoded.

  First, an estimated prediction mode for estimating the prediction mode of the processing target block is derived based on the prediction mode of the adjacent block. The method of determining the estimated prediction mode is as follows (a) to (d).

(A) When the prediction mode of the adjacent block located above or on the left is TM prediction, TM prediction is set as the estimated prediction mode.
(B) On the other hand, when the prediction mode of the adjacent block located above or to the left is H prediction, the H prediction is set as the estimated prediction mode.
(C) On the other hand, when the prediction mode of the adjacent block located above or to the left is V prediction, V prediction is set as the estimated prediction mode.
(D) On the other hand, DC prediction is set to the estimated prediction mode.

  Next, when the prediction mode and the estimated prediction mode match, the bit “0” is encoded, and the process ends. On the other hand, if they do not match, bit “1” is encoded. If the prediction mode is TM prediction, bit string “00”, if DC prediction, bit string “11”, if H prediction, bit string “10”, if V prediction, bit string “01”. Is encoded.

  As described above, the prediction mode encoding means derives an estimated prediction mode for estimating the prediction mode of the processing target block, encodes information on whether or not the prediction mode of the processing target block matches the estimated prediction mode, and performs prediction. When the mode does not coincide with the estimated prediction mode, it is preferable to encode the information of one selected prediction mode.

  The prediction mode information can be encoded by the above procedure. Since the correlation between prediction modes between adjacent blocks is high, there is a high possibility that the estimated prediction mode matches the prediction mode. Therefore, according to the above method, there is a high possibility that a short code is encoded. Therefore, according to the above method, the amount of code required for encoding the prediction mode information can be reduced as compared with the case where the prediction mode is directly encoded.

  In addition, the prediction mode can be decoded from the encoded data by the image decoding device 8 by similarly deriving the estimated prediction mode. More specifically, the prediction mode decoding unit derives an estimated prediction mode for estimating the prediction mode of the processing target block, and information indicating whether the prediction mode of the processing target block matches the estimated prediction mode is encoded data. Read from. The prediction mode decoding means decodes the estimated prediction mode as the prediction mode of the processing target block if they match, and reads information indicating the prediction mode from the encoded data if they do not match, as the prediction mode of the processing target block The read prediction mode may be decoded.

  As described above, the embodiments of the image encoding device and the image decoding device according to the present invention have been described with reference to FIGS. 1 to 20. As described above, the present invention is A form as an image encoding method or an image decoding method consisting of processing procedures may also be adopted.

  In addition, the image encoding device and the image decoding device according to the present invention described above are preferably configured by hardware from the viewpoint of processing speed, but an existing image encoder or a part of the function is used as firmware. It may be incorporated into an image decoder in an executable manner, or may be incorporated into an LSI (Large Scale Integrated Circuit) chip, for example. Furthermore, all the functions of the image encoding device and the image decoding device according to the present invention may be installed as software in an executable manner on a general-purpose computer. A general-purpose computer generally has a storage area for data such as an arithmetic processing unit such as a CPU (Central Processing Unit), a ROM (Read Only Memory) as a program storage area, a RAM (Random Access Memory) as a work area, etc. Hardware such as a hard disk drive (HDD). The above-described software may be stored in a ROM (or rewritable ROM), HDD, or the like, and the software stored in the CPU may be read onto the RAM and executed. In either case, the image encoding and image decoding functions according to the present invention can be achieved.

  In addition, such a program (firmware or software) can be distributed as a computer-readable recording medium on which the program is recorded, or can be distributed via a network or broadcast wave, and can be executed on a corresponding device. Can be incorporated.

DESCRIPTION OF SYMBOLS 1, 4, 7 ... Image coding apparatus, 2, 5, 8 ... Image decoding apparatus, 11, 41, 71 ... Orthogonal transformation part, 12, 42, 72 ... Quantization part, 13, 22, 43, 52, 73 , 82 ... Inverse quantization unit, 14, 23, 44, 53, 74, 83 ... Inverse orthogonal transform unit, 15, 24, 45, 54, 75, 84 ... Frame memory, 16, 25, 46, 55, 76, 85: In-screen prediction unit, 17, 48, 78 ... Variable length coding unit, 21, 51, 81: Variable length decoding unit, 31, 61, 91 ... Peripheral feature vector calculation unit, 32, 62, 92 ... Peripheral feature Vector recording unit, 33, 63, 93 ... search center determining unit, 34, 94 ... predicted region determining unit, 35, 65, 95 ... predicted image generating unit, 47 ... predicted region searching unit, 56 ... predicted region position calculating unit, 77 ... In-screen prediction determination unit, 96 ... DC prediction image generation unit, 97 ... H prediction Image generating unit, 98 ... V prediction image generating unit, 99 ... switch.

Claims (6)

In an image decoding apparatus that reproduces an image of each block in an image divided into a plurality of blocks by adding a difference image decoded from encoded data to a predicted image,
Feature vector deriving means for deriving a feature vector of the processing target block based on a pixel value in a neighborhood region of the processing target block;
The feature vector corresponding to the block to be processed and the feature vector corresponding to the decoded block are compared to calculate the similarity, and the position of the decoded block corresponding to the feature vector having the highest similarity is used as the center. A search range determining means for determining a region within a predetermined distance from the center as a search range;
A prediction region determining means for selecting a region similar to the processing target block from the search range as a prediction region;
An image decoding apparatus comprising: a predicted image generation unit configured to generate a predicted image for the processing target block based on an image of the prediction region. The image decoding apparatus further includes one or a plurality of different prediction image generation means for generating a prediction image for the processing target block, and a prediction mode decoding means for decoding a prediction mode, separately from the prediction image generation means,
The image decoding according to claim 1, wherein a prediction image generation unit to be used is selected according to a prediction mode decoded from the encoded data, and a prediction image is generated by the selected prediction image generation unit. apparatus. The prediction mode decoding means includes
Deriving an estimated prediction mode for estimating the prediction mode of the processing target block,
Read from the encoded data whether the prediction mode of the block to be processed and the estimated prediction mode match,
If they match, decode the estimated prediction mode as the prediction mode of the processing target block,
3. The image decoding apparatus according to claim 2, wherein, when they do not match, information indicating a prediction mode is read from encoded data, and the read prediction mode is decoded as a prediction mode of the processing target block. In an image encoding device that encodes a difference image between an input image and a predicted image for each block in an image divided into a plurality of blocks,
Feature vector deriving means for deriving a feature vector of the processing target block based on a pixel value in a neighborhood region of the processing target block;
The feature vector corresponding to the block to be processed and the feature vector corresponding to the decoded block are compared to calculate the similarity, and the position of the decoded block corresponding to the feature vector having the highest similarity is used as the center. A search range determining means for determining a region within a predetermined distance from the center as a search range;
A prediction region determining means for selecting a region similar to the processing target block from the search range as a prediction region;
An image coding apparatus comprising: a predicted image generation unit configured to generate a predicted image for the processing target block based on an image of the prediction region. In addition to the predicted image generating unit, the image encoding device includes one or a plurality of different predicted image generating units that generate a predicted image for the processing target block, and each predicted image generating unit included in the image encoding device. A prediction image determination unit that selects one prediction image generation unit by comparing the generated prediction image and an input image, and a prediction mode encoding that encodes a prediction mode indicating the selected one prediction image generation unit And further comprising means,
The image coding apparatus according to claim 4, wherein the predicted image generated by the selected one predicted image generation unit is a predicted image in the processing target block. The prediction mode encoding means includes
Deriving an estimated prediction mode for estimating the prediction mode of the processing target block,
Encode information on whether or not the prediction mode of the block to be processed and the estimated prediction mode match,
6. The image encoding apparatus according to claim 5, wherein when the prediction mode does not coincide with the estimated prediction mode, information of a prediction mode indicating the selected one prediction image generation unit is encoded.

Images