JP5254189B2 - Moving picture coding method, moving picture coding apparatus, moving picture decoding method, and moving picture decoding apparatus - Google Patents

Description

  The present invention relates to a moving picture coding technique for coding a moving picture, and a moving picture decoding technique for decoding a coded moving picture.

  In moving picture coding, compression efficiency can be increased by using a prediction technique. As a typical prediction technique, an intra-screen prediction technique that generates a predicted image of a block to be encoded using pixels in the same screen, and a coding by performing motion compensation using pixels (reference images) of the previous and subsequent screens. An inter-screen prediction technique for generating a prediction image of a target block is known. The difference (prediction error) between the prediction image generated by these prediction techniques and the input image is calculated, and the moving image is compressed through orthogonal transformation and quantization.

  Here, in order to increase the compression rate while maintaining the quality of the moving image, it is necessary to reduce the prediction error. For such a problem, a method has been proposed in which a reference image used for motion compensation for inter-screen prediction is filtered to reduce the prediction error (see, for example, Patent Document 1). In addition, in the compression of still images, a method has been proposed in which a prediction error is reduced by replacing a still image such as the Internet or a hard disk with a predicted image (see, for example, Patent Document 2).

JP 2006-135376 A JP 2008-199587 A

  According to Patent Document 1, it is possible to reduce a prediction error by filtering a reference image used for motion compensation. However, this method filters a reference image used for motion compensation in inter-screen prediction, and does not support intra-screen prediction. In addition, since the difference signal of the filter coefficient is encoded, it is conceivable that the compression effect is reduced in a scene where the image changes rapidly (for example, a scene change).

  According to Patent Document 2, prediction errors can be reduced by using images recorded on a recording medium such as a hard disk or images stored in a server on the Internet as predicted images. However, in this case, in order to create an effective prediction image for various patterns, a huge storage capacity and a network connection environment for storing the image used as the prediction image are required, and calculation required for the search is required. The cost may be high.

  An object of the present invention is to provide a moving picture coding method, a moving picture coding apparatus, a moving picture decoding method, and a moving picture decoding apparatus capable of reducing a prediction error and reducing the calculation cost required for picture search. Is to provide.

  The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

  The following is a brief description of an outline of typical inventions disclosed in the present application.

  That is, the moving image encoding method includes a subtraction process for calculating a difference between an original image and a predicted image, an orthogonal transform process for orthogonally transforming the difference calculated by the subtraction process, and a result of the orthogonal transform process is quantized. The result of the quantization process, the inverse quantization process that inversely quantizes the result of the quantization process, the inverse orthogonal transform process that inversely transforms the result of the inverse quantization process, and the result of the inverse orthogonal transform process Addition processing for adding images. At this time, the high-frequency component image corresponding to the result of the addition process is searched using a database of high-frequency component images of the high-frequency component lost in the compression process for the specific pattern, and the searched high-frequency component image is added to the addition process. And a prediction process for generating the predicted image based on the image restored by the restoration process.

  The moving image encoding apparatus is obtained by the subtractor that calculates the difference between the original image and the predicted image, the orthogonal transform unit that converts the difference calculated by the subtractor into an orthogonal transform coefficient, and the orthogonal transform unit. A quantization unit that quantizes the orthogonal transform coefficient to obtain a quantization coefficient. Further, the moving picture coding apparatus includes an inverse quantization unit that inversely quantizes the quantization coefficient obtained by the quantization unit, and an inverse orthogonal transform that inversely transforms the inverse quantization coefficient obtained by the inverse quantization unit. An orthogonal transform unit, and an adder that adds a predicted image to the processing result of the inverse orthogonal transform unit. Then, a high frequency component image corresponding to the output of the adder is searched using a database of high frequency component images of high frequency components lost in the compression process for the specific pattern, and the searched high frequency component image is stored in the adder. A restoration unit that restores an image by combining with an output and a prediction unit that generates the predicted image based on the image restored by the restoration unit are provided.

  The moving picture decoding method includes an entropy decoding process for entropy decoding an input encoded stream, an inverse quantization process for inversely quantizing a result of the entropy decoding process, and a result of the inverse quantization process. Inverse orthogonal transformation processing for inverse orthogonal transformation, and addition processing for adding the result of the inverse orthogonal transformation processing and the predicted image to decode the image data. At this time, the high-frequency component image corresponding to the result of the addition process is searched using a database of high-frequency component images of the high-frequency component lost in the compression process for the specific pattern, and the searched high-frequency component image is added to the addition process. And a prediction process for generating the predicted image using the image data obtained by synthesizing the high-frequency components in the recovery process.

  The video decoding device includes an entropy decoding unit that performs entropy decoding on an input encoded stream, an inverse quantization unit that inversely quantizes a decoding result in the entropy decoding unit, and the inverse quantization unit An inverse orthogonal transform unit that performs inverse orthogonal transform on the result of inverse quantization in the above, and an adder that adds the inverse orthogonal transform result in the inverse orthogonal transform unit and the predicted image to decode the image data. At this time, a high-frequency component image corresponding to the output of the adder is searched using a database of high-frequency component images of high-frequency components lost in the compression process for the specific pattern, and the searched high-frequency component image is added to the adder. A restoration unit that restores an image by combining the output with the output of the image and a prediction unit that generates the predicted image using image data in which high-frequency components are synthesized by the restoration unit are provided.

  As described above, a high-frequency component image corresponding to the addition result is searched using a database of high-frequency component images due to high-frequency components lost in the compression process for the specific pattern, and the searched high-frequency component image is combined with the addition result. Thus, an image close to the original image is restored, and a predicted image is generated based on the restored image, thereby reducing a prediction error and a calculation cost required for the image search.

  The effects obtained by the representative ones of the inventions disclosed in the present application will be briefly described as follows.

  That is, it is possible to provide a moving image encoding method, a moving image encoding device, a moving image decoding method, and a moving image decoding device that can reduce the prediction error and reduce the calculation cost required for image search. Can do.

1 is a block diagram illustrating a configuration example of a moving image encoding device according to the present invention. FIG. 3 is a block diagram illustrating a configuration example of a restoration unit included in the moving image encoding device illustrated in FIG. 1. It is explanatory drawing of the code book contained in the decompression | restoration part shown by FIG. It is a flowchart which shows the flow of operation | movement of the decompression | restoration part shown by FIG. FIG. 2 is a block diagram illustrating a configuration example of a video decoding device for decoding a decoded stream output from the video encoding device illustrated in FIG. 1. FIG. 6 is a block diagram illustrating a configuration example of a restoration unit included in the video decoding device illustrated in FIG. 5. It is a flowchart which shows the flow of operation | movement of the decompression | restoration part shown by FIG. It is another structural example block diagram of the said moving image encoder. FIG. 9 is a block diagram illustrating a configuration example of a restoration unit included in the moving image encoding device illustrated in FIG. 8. It is a flowchart which shows the flow of operation | movement of the decompression | restoration part shown by FIG. It is explanatory drawing of the encoding stream output from the moving image encoding apparatus shown by FIG. FIG. 9 is a block diagram illustrating a configuration example of a moving picture decoding apparatus for decoding a decoded stream output from the moving picture encoding apparatus illustrated in FIG. 8. FIG. 13 is a block diagram illustrating a configuration example of a restoration unit included in the video decoding device illustrated in FIG. 12. It is a flowchart which shows the flow of operation | movement of the decompression | restoration part shown by FIG.

1. First, an outline of a typical embodiment of the invention disclosed in the present application will be described. Reference numerals in the drawings referred to in parentheses in the outline description of the representative embodiments merely exemplify what are included in the concept of the components to which the reference numerals are attached.

  [1] A video encoding device (100) according to a typical embodiment of the present invention includes a subtracter (101) that calculates a difference between an original image and a predicted image, and a difference calculated by the subtractor. And an orthogonal transform unit (102) for transforming the orthogonal transform coefficient into an orthogonal transform coefficient, and a quantization unit (103) for quantizing the orthogonal transform coefficient obtained by the orthogonal transform unit to obtain a quantized coefficient. Further, the moving image encoding device (100) includes an inverse quantization unit (105) that inversely quantizes the quantization coefficient obtained by the quantization unit, and an inverse quantization coefficient obtained by the inverse quantization unit. An inverse orthogonal transform unit (106) that performs inverse orthogonal transform, an adder (107) that adds a prediction image to the processing result of the inverse orthogonal transform unit, and an output stream obtained by entropy encoding the output of the quantization unit And an entropy encoding unit (104) for generating. Then, using the high-frequency component image database 202 by the high-frequency component lost in the compression process for the specific pattern, the high-frequency component image corresponding to the output of the adder is searched, and the searched high-frequency component image is A restoration unit (108) that restores an image by combining with the output of the adder and a prediction unit (110) that generates the predicted image based on the image restored by the restoration unit are provided.

  [2] In the above [1], the restoration unit is based on an image determination unit (201) that determines whether the output of the adder corresponds to the specific pattern, and a determination result of the image determination unit. And a synthesizing unit (203) for searching for a corresponding high frequency component from the database and synthesizing it with the output of the adder.

  [3] In the above [2], the image determination unit compares a variance value or an average deviation value as a result of the addition process or an average response by a directional filter with a predetermined threshold value, and based on the comparison result It can be configured to determine whether or not the addition result in the adder corresponds to a specific pattern.

  [4] In the above [2], the image determination unit forms identification information based on a determination result of whether or not the addition result in the adder corresponds to the specific pattern. It can be configured to be synthesized with the encoded stream by the entropy encoding unit.

  [5] An entropy decoding unit (501) that entropy-decodes the input encoded stream, an inverse quantization unit (502) that inversely quantizes a decoding result in the entropy decoding unit, and the inverse quantum An inverse orthogonal transform unit (503) that performs inverse orthogonal transform on the inverse quantization result in the conversion unit, and an adder (504) that adds the inverse orthogonal transform result in the inverse orthogonal transform unit and the predicted image to decode the image data ) Is configured. At this time, a high-frequency component image corresponding to the output of the adder is searched using a database of high-frequency component images of high-frequency components lost in the compression process for the specific pattern, and the searched high-frequency component image is added to the adder. A restoration unit (505) that restores an image by synthesizing it with the output of the above and a prediction unit (507) that generates the prediction image using image data in which high-frequency components are synthesized by the restoration unit are provided.

  [6] In the above [5], the restoration unit is based on an image determination unit (601) that determines whether or not the output of the adder corresponds to the specific pattern, and a determination result of the image determination unit. And a synthesizing unit (602) for searching for a corresponding high frequency component from the database and synthesizing it with the output of the adder.

  [7] In the above [6], the image determination unit compares a variance value or an average deviation value of the output of the adder or an average response by a directional filter with a predetermined threshold value, and based on the comparison result In addition, it can be configured to determine whether or not the output of the adder corresponds to a specific pattern.

  [8] In the above [5], the entropy decoding unit decodes identification information included in the input encoded stream, and the restoration unit is based on the identification information decoded by the entropy decoding unit. Thus, a selection circuit (1301) for selecting an image to be restored can be provided. The entropy decoding unit searches for the corresponding high frequency component image from the database with respect to the image selected as the image to be restored by the selection circuit, and combines it with the output of the adder. A combining unit (1302) can be provided.

2. Details of Embodiments Embodiments will be further described in detail.

<Embodiment 1>
FIG. 1 shows a configuration example of a moving image encoding apparatus according to Embodiment 1.

  The moving image encoding apparatus 100 shown in FIG. The present invention is applied to a video engine based on MPEG4 (Moving Picture Experts Group phase 4) defined as H.264 / AVC. The moving image coding apparatus 100 includes a subtractor 101, an orthogonal transform unit (OTF) 102, a quantization unit (QUA) 103, an entropy coding unit (ENTC) 104, an inverse quantization unit (IQUA) 105, and an inverse orthogonal transform. Unit (IOTF) 106, adder 107, restoration unit (RES) 108, frame memory (FMEM) 109, and prediction unit (PRE) 110. Each of the above units may be configured by hardware or software. Further, it may be a module combining hardware and software.

  The subtractor 101 obtains a difference between the original image ORG-IMG and the predicted image PRE-IMG and creates a difference image. The orthogonal transform unit 102 transforms the difference image created by the subtractor 101 into an orthogonal transform coefficient using an orthogonal transform method such as discrete cosine transform. The quantization unit 103 quantizes the orthogonal transform coefficient obtained by the orthogonal transform unit 102 and outputs a quantization coefficient. The entropy encoding unit 104 encodes the quantization coefficient output from the quantization unit 103, the prediction mode output from the prediction unit 110, and the like using Huffman code, arithmetic code, and the like, and compresses image compression information (encoded stream). QSTR). The inverse quantization unit 105 inversely quantizes the quantization coefficient output from the quantization unit 103. The inverse orthogonal transform unit 106 performs inverse orthogonal transform on the inverse quantization coefficient output from the inverse quantization unit 105. The adder 107 adds the difference image output from the inverse orthogonal transform unit 106 and the prediction image output from the prediction unit 110, and outputs a locally decoded image. Here, the locally decoded image is an image block to be processed, and the size is not particularly limited, but is 4 × 4 pixels, 8 × 8 pixels, 16 × 16 pixels, or the like. The restoration unit 108 creates a second local decoded image (this is referred to as a “synthesized local decoded image”) using the code book. This creation method will be described later with reference to FIGS. A frame memory (FMEM) 109 stores the image output from the restoration unit 108 as a reference image. The prediction unit 110 performs prediction using the reference image stored in the frame memory 109, and outputs a predicted image (PRE-IMG). As the prediction method in the prediction unit 110, intra-screen prediction, inter-screen prediction, or both can be employed.

  FIG. 2 shows a configuration example of the restoration unit 108.

The restoration unit 108 illustrated in FIG. 2 includes an image determination unit (DEC) 201, a code book (CODB) 202, a synthesis unit (SYN) 203, and a selection circuit 204. The image determination unit 201 analyzes the pattern of the input local decoded image IN-IMG, and determines whether the pattern corresponds to a specific pattern such as a line, an edge, a corner, or a T-shape. When the design of the local decoded image corresponds to a specific pattern, the local decoded image is transmitted from the image determination unit 201 to the synthesis unit 203. The synthesizing unit 203 restores the local decoded image to an image close to the original image using the code book 202, and generates a synthesized local decoded image. Here, the code book 202 is a database formed in advance based on a plurality of specific patterns 302 in the reference image 301 as shown in FIG. 3A, for example. That is, as shown in FIG. 3B, for example, the code book 202 includes a compressed image 303 for the specific pattern 302 and an image of the high-frequency component lost in the compression process of the specific pattern 302 (“high-frequency component image”). 304 ) are collected into a database by being grouped together with the same index, and stored in an appropriate semiconductor memory such as a ROM (Read Only Memory). The selection circuit 204 selects and outputs the local decoded image output from the image determination unit 201 and the synthesized local decoded image output from the synthesis unit 203 based on the determination result of the image determination unit 201. The image output via the selection circuit 204 is “OUT-IMG”.

  Next, the flow of processing in the restoration unit 108 will be described with reference to FIG.

  First, a locally decoded image block is input to the restoration unit 108 (S4001). Thereafter, it is determined whether or not the locally decoded image block corresponds to the specific pattern (S4002). For example, if the variance (Equation 1) or the average deviation (Equation 2) of the locally decoded image block is greater than or equal to a predetermined threshold, it is determined that the specific pattern is met.

Here, “S ² ” is the variance value of the locally decoded image block, “d” is the average deviation value, “n” is the number of pixels in the block, “x _ij ” is the pixel value at coordinates (i, j), “ “x ₀ ” is the average pixel value in the block. In addition to this, for example, a directional filter such as a canny filter or a Gabor filter may be used to determine that a specific pattern corresponds if the average response is equal to or greater than a predetermined threshold.

  If the image determination unit 201 determines that the pattern does not correspond to the specific pattern, the restoration is not performed. In such a case, the locally decoded image IN-IMG input to the image determination unit 201 is directly output to the frame memory 109 via the selection circuit 204 (S4003). On the other hand, if it is determined that it corresponds to the specific pattern, an index that minimizes the error from the compressed image 303 in the codebook 202 is searched using the input local decoded image block as a key, and the corresponding high frequency The component image 304 is extracted (S4004). As an error criterion, SAD (Sum of Absolute Difference) or SSD (Sum of Square Difference) may be used. Thereafter, the extracted high-frequency component image 304 is synthesized (added) to the local decoded image block to generate a synthesized local decoded image (S4005). The synthesized local decoded image thus created is output to the frame memory 109 (S4006).

  FIG. 5 shows a configuration example of a moving picture decoding apparatus 500 for decoding the decoded stream QSTR output from the moving picture encoding apparatus 100 shown in FIG.

  5 is not particularly limited, but an entropy decoding unit (ENTD) 501, an inverse quantization unit (IQUA) 502, an inverse orthogonal transform unit (IOTE) 503, an adder 504, and a restoration unit. (RES) 505, a frame memory (FMEM) 506, and a prediction unit (PRE) 507 are configured. As with the moving image encoding apparatus 100, each unit may be configured by hardware or software. Further, it may be a module combining hardware and software.

  The entropy decoding unit 501 decodes the transmitted image compression information. Here, the image compression information refers to the encoded stream QSTR output from the moving image encoding apparatus 100. The inverse quantization unit 502 inversely quantizes the quantization coefficient. The inverse orthogonal transform unit 503 performs inverse orthogonal transform on the inverse quantization coefficient output from the inverse quantization unit 502. The adder 504 adds the difference image output from the inverse orthogonal transform unit 503 and the prediction image output from the prediction unit 507, and outputs a decoded image. The restoration unit 505 generates a reference image by synthesizing the high-frequency component image with the decoded image output from the adder 504 using the code book having the same content as the code book 202 in the moving image coding apparatus 100. The frame memory 506 stores the reference image created by the restoration unit 505. The prediction unit 507 generates a predicted image using the reference image stored in the frame memory 506.

  FIG. 6 shows a configuration example of the restoration unit 505.

  6 includes an image determination unit (DEC) 601, a synthesis unit (SYN) 602, a code book (CODB) 603, and a selection circuit 604. The image determination unit 601 analyzes the pattern of the input decoded image, and passes the decoded image to the synthesis unit 602 when the pattern corresponds to a specific pattern such as a line, an edge, a corner, or a T shape. The synthesizing unit 602 restores the decoded image to an image close to the original image using the code book 603, and generates a synthesized decoded image. The selection circuit 604 selects a decoded image or a synthesized decoded image based on the determination result in the image determination unit 601 to obtain an output image OUT-IMG. The code book 603 is a database having the same content as the code book 202 in the moving image coding apparatus 100.

  Next, the processing flow of the restoration unit 505 will be described with reference to FIG.

  First, a decoded image block is input to the image determination unit 601 as the input image IN-IMG (S7001). Thereafter, the image determination unit 601 determines whether the decoded image block corresponds to the specific pattern (S7002). For example, if the variance value (Equation 1) and the average deviation value (Equation 2) of the decoded image block are equal to or greater than a predetermined threshold value, it is determined that the pattern corresponds to the specific pattern. In addition to this, for example, a directional filter such as a Canny filter or a Gabor filter may be used to determine that the pattern corresponds to a specific pattern if the average response is equal to or greater than a threshold value. If it is determined by the image determination unit 601 that the pattern does not correspond to the specific pattern, the decoded image block is output as it is via the selection circuit 604 (S7003). On the other hand, if it is determined by the image determination unit 601 that the image corresponds to the specific pattern, an index with the smallest average error from the compressed image 303 in the codebook is searched using the input decoded image block as a key. The corresponding high frequency component image 304 is obtained (S7004). The extracted high-frequency component image 304 is combined with the decoded image block in the combining unit 602, and a combined decoded image IMGD is generated (S7005). The generated synthesized decoded image is output to a subsequent circuit (not shown) and the frame memory 506 (S7006).

  According to the first embodiment described above, the following operational effects can be obtained.

  (1) The moving image encoding apparatus 100 is provided with a prediction unit 110 that generates a predicted image based on the image restored by the restoration unit 108. In the above configuration, when a local decoded image block is input to the restoration unit 108, it is determined whether or not the local decoded image block corresponds to the specific pattern (S4002). In this determination, if it is determined that the pattern corresponds to the specific pattern, an index that minimizes the error from the compressed image 303 in the codebook 202 is searched using the input local decoded image block as a key, and the corresponding high-frequency component is searched. The image 304 is extracted (S4004). Thereafter, the extracted high frequency component image 304 is synthesized (added) to the local decoded image block, and a synthesized local decoded image is generated (S4005).

  In order to increase the compression rate while maintaining the quality of the moving image, it is necessary to reduce the prediction error. Here, according to the technique of Patent Document 2, it is necessary to prepare an enormous number of images in order to create an effective prediction image for various patterns, and if that is not possible, an unnatural reference image is generated. There is also a possibility that it will be generated. In addition, when an enormous image is required, an enormous storage capacity and network connection environment for storing the image are required, and the calculation cost required for the search increases.

  On the other hand, in the first embodiment, the code book 202, which is a database of high-frequency component images based on high-frequency components lost in the compression process for the specific pattern as described above, is used to deal with the output of the adder. By searching for a high-frequency component image and adding the searched high-frequency component image to the output of the adder 107, an image close to the original image can be restored, and a predicted image can be generated based on the restored image. Can be reduced, thereby increasing the compression ratio. In addition, since the restoration is limited to a specific pattern in the code book 202, the number of comparison targets can be greatly reduced as compared with the case where the restoration is performed on all images. The size of is also reduced. As a result, it is possible to reduce the storage capacity necessary for forming the code book 202 and the amount of calculation when searching for the code book, and to reduce the matching error.

  (2) In Embodiment 1, unlike the case of Embodiment 2 to be described later, it is not necessary to add identification information to the encoded stream, so that the compression rate can be increased accordingly. Furthermore, by using the restored reference image not only for inter-screen prediction but also for intra-screen prediction, the compression rate can be further increased. Further, unlike Patent Document 2, since the differential signal of the filter coefficient is not encoded, the compression effect does not change even in a scene where the image changes abruptly (for example, a scene change).

  (3) The restoration unit 108 determines whether or not the output of the adder 107 corresponds to a specific pattern and the code book 202 based on the determination result of the image determination unit 201. This can be easily realized by providing a synthesizing unit 203 that searches for a high-frequency component to be added and adds it to the output of the adder 107.

  (4) The image determination unit 201 can be easily realized by being configured to compare the variance value or the average deviation value of the output of the adder or the average response by the directional filter with a predetermined threshold value. Can do.

  (5) The moving picture decoding apparatus 500 uses the code book 603, which is a database of high-frequency component images based on high-frequency components lost in the compression process for a specific pattern, and uses the high-frequency component corresponding to the processing result in the addition process. A restoration unit 505 that searches for an image and restores the image by adding the searched high-frequency component image to the output of the adder 504 is included. In addition, the moving picture decoding apparatus 500 includes a prediction unit 507 that generates a prediction image using the image data in which the high-frequency components are synthesized by the restoration unit 505. Thus, the high frequency component image searched using the code book 603 is added to the output of the adder 504, so that an image close to the original image can be restored, and a predicted image is generated based on the restored image. As a result, the prediction error can be reduced. In addition, since the restoration unit 505 performs restoration only for a specific pattern, the number of comparison targets can be greatly reduced as compared with restoration for all images. The size of 603 is also reduced. As a result, it is possible to reduce the storage capacity necessary to form the code book 603 and the amount of calculation when searching for the code book, and to expect the effect of reducing matching errors.

  (6) The restoration unit 505 corresponds from the database based on the determination result of the image determination unit 601 that determines whether the addition result of the addition process corresponds to the specific pattern and the determination result of the image determination unit 601. This can be easily realized by providing a synthesizing unit 602 that searches for a high-frequency component to be added and adds it to the output of the adder 504.

  (7) The image determination unit 601 can be easily realized by being configured to compare the variance value or average deviation value of the output of the adder or the average response by the directional filter with a predetermined threshold value. Can do.

<Embodiment 2>
Next, the second embodiment will be described with reference to FIGS.

  FIG. 8 shows a configuration example of the moving picture encoding apparatus according to the second embodiment.

  The video encoding device 800 shown in FIG. 8 is greatly different from that shown in FIG. 1 in that identification information IDEN indicating whether or not the restoration unit 801 performs restoration is output. The identification information IDEN is transmitted to the entropy encoding unit 104. The entropy encoding unit 104 incorporates the identification information IDEN into the encoded stream QSTR.

  FIG. 9 shows a configuration example of the restoration unit 801. The restoration unit 801 includes an image determination unit 201, a code book 202, a synthesis unit 203, and a selection circuit 204, similar to the restoration unit 108 in the video encoding device 100. The image determination unit 201 forms identification information IDEN based on the image determination result. The identification information IDEN is transmitted to the selection circuit 204 and the entropy encoding unit 104.

  A processing flow of the restoration unit 801 will be described with reference to FIG.

  First, the local decoded image block IN-IMG is input to the image determination unit 201 (S10001). Thereafter, the image determination unit 201 determines whether or not the locally decoded image block IN-IMG corresponds to the specific pattern (S10002). This determination method is the same as that of the first embodiment. For example, based on the variance value (Equation 1) and the average deviation value (Equation 2) of the locally decoded image block, it is determined whether or not the specific pattern is met. Done. When the image determination unit 201 determines that the input locally decoded image block does not correspond to the specific pattern, the block is not restored, and identification information indicating non-restoration is output (S10003). The input local decoded image block is output as it is through the selection circuit 204 (S10004). When the image determination unit 201 determines that the input local decoded image IN-IMG corresponds to the block specific pattern, identification information indicating restoration is output (S10005). Thereafter, using the input local decoded image block as a key, an index that minimizes the error from the compressed image 303 in the codebook is searched, and the corresponding high-frequency component image 304 is obtained (S10006). As an error standard, SAD (Sum of Absolute Difference) or SSD (Sum of Square Difference) may be used. The high-frequency component image 304 extracted in this way is synthesized into a locally decoded image block, and a synthesized locally decoded image is generated (S10007). The synthesized local decoded image is output to the frame memory 109 (S10008).

  FIG. 11 shows an example of the encoded stream QSTR output from the entropy encoding unit 104 in FIG.

  The encoded stream QSTR is a digital signal 1101 output in a serial format, and the digital signal 1101 has identification information 1102, 1104 for determining whether or not to perform restoration, as shown partially enlarged. 1106, 1108, and information 1103, 1105, 1107, 1109 such as motion vectors and orthogonal transform coefficients. Each of the identification information 1102, 1104, 1106, and 1108 is 1-bit information. For example, when the restoration is not performed, the logical value is “0”, and when the restoration is performed, the logical value is “1”.

  FIG. 12 shows a configuration example of a moving picture decoding apparatus 1200 for decoding the encoded stream QSTR output from the moving picture encoding apparatus 800 shown in FIG.

  The moving picture decoding apparatus 1200 shown in FIG. 12 is greatly different from that shown in FIG. 5 in that the identification information IDEN included in the encoded stream QSTR is decoded by the entropy decoding unit 501, which is restored to the restoration unit 1201. It is a point transmitted to.

  FIG. 13 shows a configuration example of the restoration unit 1201.

  The restoration unit 1201 shown in FIG. 13 is significantly different from that shown in FIG. 6 in that no equivalent to the image determination unit 601 in FIG. 6 is provided, and instead based on the identification information IDEN. This is that selection circuits 1301 and 1303 for performing path switching are provided. In particular, the selection circuit 1301 selects a decoded image to be restored and supplies it to the synthesis unit 1302 by performing path switching based on the identification information IDEN. Based on the decoded image supplied via the selection circuit 1301, the synthesizer 1302 searches the codebook 1201 for a corresponding high frequency component image and adds it to the input image IN-IMG, thereby generating a synthesized decoded image. Generate.

  The processing flow of the restoration unit 1201 will be described with reference to FIG.

  First, a decoded image block is input to the decoding unit 1201 (S14001). Thereafter, based on the identification information IDEN, it is determined whether or not to execute restoration (S14002). According to the above-described example, if the identification information IDEN is the logical value “0”, the restoration is not performed and the input decoded image block is output as it is (S14003). If the identification information IDEN is a logical value “1”, an index with the smallest error from the compressed image 303 in the codebook is searched using the input decoded image block as a key, and the corresponding high-frequency component image 304 is obtained. (S14004). The extracted high frequency component image 304 is combined with the decoded image block, and a combined decoded image is generated (S14005). The synthesized decoded image created in this way is output to a subsequent circuit (not shown) and the frame memory 506 (S14006).

  According to the second embodiment described above, the restoration unit 801 is provided in the video encoding device 800, and the restoration unit 1201 is provided in the video decoding device 1200. Therefore, the same as in the case of the first embodiment. Further, according to the second embodiment, the identification information IDEN as to whether or not to perform restoration is included in the encoded stream. Therefore, in the moving picture decoding apparatus 1200, The thing corresponding to the image determination unit 601 in FIG. 6 is unnecessary, and as a result, the circuit scale and the processing amount of the restoration unit 1201 can be reduced.

  Although the invention made by the present inventor has been specifically described based on the embodiments, it is needless to say that the present invention is not limited thereto and can be variously modified without departing from the gist thereof.

  For example, the code book 202 may be arranged outside the restoration units 108 and 801. Similarly, the code book 603 may be arranged outside the restoration units 505 and 1201.

  Further, the present invention relates to H.264. The present invention is not limited to the moving image encoding and decoding technology defined as H.264 / AVC, and can be widely applied to encoding and decoding technologies using other methods.

DESCRIPTION OF SYMBOLS 100,800 moving image encoder 101 Subtractor 102 Orthogonal transformation part 103 Quantization part 104 Entropy encoding part 105,503 Inverse quantization part 106,502 Inverse orthogonal transformation part 107,504 Adder 108,505,801,1201 Restoration unit 109, 506 Frame memory 110, 507 Prediction unit 201, 601 Image determination unit 202, 603 Codebook 203, 602, 1302 Composition unit 204, 604, 1301, 1303 Selection circuit 500, 1200 Video decoding device

Claims (16)

A subtraction process for calculating a difference between the original image and the predicted image;
An orthogonal transformation process for orthogonal transformation of the difference calculated in the subtraction process;
A quantization process for quantizing the result of the orthogonal transformation process;
An inverse quantization process for inversely quantizing the result of the quantization process;
An inverse orthogonal transform process for performing an inverse orthogonal transform on the result of the inverse quantization process;
An addition process for adding a predicted image to the result of the inverse orthogonal transform process;
An entropy encoding process that entropy encodes the result of the quantization process to generate an encoded stream,
A compressed image obtained by compressing a plurality of specific patterns in an image used as a reference for creating a code book and a high-frequency component image lost in the compression process of the specific pattern are collected in the same index and stored in a database in advance. Referring to the codebook, search for an index that minimizes the error between the result of the addition process and the compressed image in the codebook, and extract the corresponding high-frequency component image. Restoration processing to restore the image by compositing with the result of
And a prediction process for generating the predicted image based on the image restored by the restoration process. The restoration process includes a determination process for determining whether a result of the addition process corresponds to the specific pattern, and restores the image when the result of the addition process corresponds to the specific pattern. The moving image encoding method described. The above determination process, the mean and the response by the dispersion value or the average deviation or directional filter result of the addition process, with a predetermined threshold value comparison, on the basis of the comparison result, the processing result of the above addition process is the The moving image encoding method according to claim 2, further comprising a process of determining whether or not the specific pattern is met. The determination process further includes a process of forming identification information based on a determination result of whether or not the result of the addition process corresponds to the specific pattern,
The moving image encoding method according to claim 2, wherein the identification information is combined with the encoded stream in the entropy encoding process. A subtractor for calculating a difference between the original image and the predicted image;
An orthogonal transform unit that transforms the difference calculated by the subtractor into an orthogonal transform coefficient;
A quantizing unit that obtains a quantized coefficient by quantizing the orthogonal transform coefficient obtained by the orthogonal transform unit;
An inverse quantization unit that inversely quantizes the quantization coefficient obtained by the quantization unit;
An inverse orthogonal transform unit that performs inverse orthogonal transform on the inverse quantization coefficient obtained by the inverse quantization unit;
An adder for adding a predicted image to the processing result in the inverse orthogonal transform unit;
An entropy encoding unit that entropy encodes the output of the quantization unit to generate an encoded stream,
A compressed image obtained by compressing a plurality of specific patterns in an image used as a reference for creating a code book and a high-frequency component image lost in the compression process of the specific pattern are collected in the same index and stored in a database in advance. Referring to the code book, search for an index that minimizes an error between the result of the addition process and the compressed image in the code book, and extract the corresponding high-frequency component image. A restoration unit that restores the image by compositing with the addition result in,
And a prediction unit that generates the predicted image based on the image restored by the restoration unit. The restoration unit includes an image determination unit that determines whether an output of the adder corresponds to the specific pattern, and restores the image when the output of the adder corresponds to the specific pattern. 5. The moving image encoding device according to 5.   The image determination unit compares a dispersion value or an average deviation value as a result of the addition process or an average response by a directional filter with a predetermined threshold value, and based on the comparison result, the addition result in the adder is The moving picture encoding apparatus according to claim 6, wherein it is determined whether or not the specific pattern is satisfied. The image determination unit forms identification information based on a determination result of whether or not the addition result in the adder corresponds to the specific pattern,
The moving image encoding apparatus according to claim 6, wherein the identification information is combined with the encoded stream by the entropy encoding unit. An entropy decoding process for entropy decoding the input encoded stream;
An inverse quantization process for inversely quantizing the result of the entropy decoding process;
An inverse orthogonal transform process for performing an inverse orthogonal transform on the result of the inverse quantization process;
A moving picture decoding method comprising: a summing process of decoding the images by adding the results and the prediction image of the inverse orthogonal transform processing,
A compressed image obtained by compressing a plurality of specific patterns in an image used as a reference for creating a code book and a high-frequency component image lost in the compression process of the specific pattern are collected in the same index and stored in a database in advance. Referring to the codebook, the search is made for an index that minimizes the error between the image decoded by the addition process and the compressed image in the codebook, and the corresponding high-frequency component image is extracted. A restoration process for restoring an image by adding to the image decoded by the addition process ;
Moving picture decoding method characterized by including: a prediction process for generating the predicted image by using the image to which the high frequency components are synthesized by the recovery process. The restoration process includes a determination process for determining whether the image decoded by the addition process corresponds to the specific pattern,
The moving image decoding method according to claim 9 , wherein the image is restored when the image decoded by the addition processing corresponds to the specific pattern . The determination process compares a variance value or an average deviation value of the result of the image decoded by the addition process or an average response by a directional filter with a predetermined threshold, and based on the comparison result , The moving image decoding method according to claim 10, further comprising a process of determining whether or not the decoded image corresponds to the specific pattern. The entropy decoding process includes a process of decoding identification information included in the input encoded stream,
The moving picture decoding method according to claim 9, wherein in the restoration process, the restoration process is performed based on the identification information extracted in the entropy decoding process. An entropy decoding unit for entropy decoding the input encoded stream;
An inverse quantization unit that inversely quantizes a decoding result in the entropy decoding unit;
An inverse orthogonal transform unit that performs inverse orthogonal transform on the inverse quantization result in the inverse quantization unit;
A moving picture decoding apparatus comprising, an adder for decoding the images by adding the inverse orthogonal transform result and the predicted image in the inverse orthogonal transform unit,
A compressed image obtained by compressing a plurality of specific patterns in an image used as a reference for creating a code book and a high-frequency component image lost in the compression process of the specific pattern are collected in the same index and stored in a database in advance. The codebook is searched for an index that minimizes the error between the image decoded by the addition and addition process and the compressed image in the codebook, and the corresponding high-frequency component image is extracted. A restoration unit that restores the image by combining the decoded image with the image decoded by the adder ,
Video decoding apparatus characterized by comprising: a prediction unit for generating the predicted image by using the image to which the high frequency components are synthesized by the recovery unit. The recovery unit, said if the image decoded by the adder comprises determining image determining unit whether corresponds to the specific pattern, the decoded image in the adder corresponds to the specific pattern The moving picture decoding apparatus according to claim 13, which restores an image . The image determination unit compares a variance value or an average deviation value of the image decoded by the adder or an average response by a directional filter with a predetermined threshold, and decodes by the adder based on the comparison result 15. The moving picture decoding apparatus according to claim 14, wherein it is determined whether or not the processed image corresponds to a specific pattern. The entropy decoding unit decodes identification information included in the input encoded stream,
The restoration unit includes a selection circuit that selects an image to be restored based on the identification information decoded by the entropy decoding unit;
A synthesizing unit that searches the code book for a corresponding high-frequency component image for the image selected as an image to be restored by the selection circuit, and synthesizes the image with the image decoded by the adder; The moving picture decoding apparatus according to claim 13, further comprising:

Images