JP5017181B2 - Moving picture coding apparatus, moving picture coding method, moving picture coding program, moving picture decoding apparatus, moving picture decoding method, and moving picture decoding program - Google Patents

Description

  The present invention relates to a moving image encoding apparatus, a moving image encoding method, a moving image encoding program, a moving image decoding apparatus, a moving image decoding method, and a moving image decoding program.

  In the moving image encoding method using motion compensation, motion detection of the image to be encoded is performed using a reference image that has been encoded in the past and stored in the frame memory, and is referenced using the detected motion vector. A motion compensated image (predicted image) is created from the image. At this time, the accuracy of the reference image is changed from the pixel unit accuracy (integer pixel accuracy) originally existing in the reference image to the pixel unit accuracy (fractional pixel accuracy) between adjacent pixels of the reference image. The motion of the encoding target image can be compensated with high accuracy, and the encoding efficiency can be improved.

  In the conventional encoding method, high encoding efficiency is realized by performing motion detection and motion compensation using a reference image with 1/4 pixel accuracy.

  Specifically, first, a reference image with 1/2 pixel accuracy is obtained by applying a one-dimensional 6-tap filter with a coefficient (1, -5, 20, 20, -5, 1) / 32 to a reference image with integer pixel accuracy. An image is generated. Next, a one-dimensional two-tap average value filter with a coefficient (1, 1) / 2 is applied to the reference image with 1/2 pixel accuracy to generate a reference image with 1/4 pixel accuracy.

A method for generating a reference image with 1/4 pixel accuracy in the conventional coding method will be described in detail with reference to FIG. FIG. 1 is a layout diagram of pixels of a reference image having a 1/4 pixel accuracy. In FIG. 1, pixels corresponding to the integer pixel signal (for example, pixels A, B, C, etc.) correspond to pixels that have been subjected to diagonally downward slanting, and pixels corresponding to the 1/2 pixel signal (for example, Pixels aa, bb, cc, etc.) correspond to pixels that have been diagonally hatched to the right, and pixels that correspond to 1/4 pixel signals (for example, pixels a, c, d, etc.) are hatched. This corresponds to a pixel that is not. The 1/2 pixel signal at the intermediate position in the horizontal direction between the two integer pixel signals is generated by a horizontal 6-tap filter. For example, the pixel b is calculated as shown in the following equation (1) by applying a horizontal 6-tap filter to the integer pixels C1, C2, C3, C4, C5, and C6.
b = (C1-5 × C2 + 20 × C3 + 20 × C4-5 × C5 + C6) / 32 (1)
Further, a ½ pixel signal at a middle position in the vertical direction between two integer pixel signals is generated by a vertical 6-tap filter. For example, the pixel h is expressed by the following equation (2) by applying a vertical 6-tap filter to the integer pixels A3, B3, C3, D3, E3, and F3.
h = (A3-5 × B3 + 20 × C3 + 20 × D3-5 × E3 + F3) / 32 (2)
A 1/2 pixel signal at an intermediate position between the four integer pixel signals is generated by applying a 6-tap filter both horizontally and vertically. For example, the pixel j generates a half pixel signal aa, bb, b, hh, ii, jj by a horizontal 6-tap filter, and then applies a vertical 6-tap filter to these signals by It is calculated as 3).
j = (aa-5 * bb + 20 * b + 20 * hh-5 * ii + jj) / 32 (3)
Alternatively, after generating the 1/2 pixel signals cc, dd, h, ee, ff, and gg by vertical filtering, the pixel j may be generated by the horizontal filtering as shown in the following equation (4).
j = (cc-5 * dd + 20 * h + 20 * ee-5 * ff + gg) / 32 (4)
Next, after all the ½ pixel signals are calculated, a ¼ pixel signal is generated using an average value filter. Pixels a, c, i, and k in FIG. 1 are generated by applying an average value filter in the horizontal direction to adjacent integer pixel signals or ½ pixel signals. For example, the pixel a is calculated by the following equation (5).
a = (C3 + b) / 2 (5)
Pixels d, f, n, and q are generated by applying an average value filter in the vertical direction to adjacent integer pixel signals or ½ pixel signals. For example, the pixel f is calculated by the following equation (6).
f = (b + j) / 2 (6)
Pixels e, g, p, and r are calculated by applying an average value filter in an oblique direction. For example, the pixel r is calculated by the following equation (7).
r = (ee + hh) / 2 (7)
In this way, in the conventional encoding methods, a fixed one-dimensional 6-tap filter and a one-dimensional two-tap average value filter are always used to generate a 1/4 pixel accuracy from an integer pixel accuracy reference image. A reference image is generated.

  On the other hand, in general, video has a motion amount and frequency characteristics with different pixel accuracy for each type of video sequence and each frame, so a reference image with fractional pixel accuracy using a different filter for each type of video sequence and each frame. Is preferably generated.

Non-Patent Document 1 below discloses a technique for generating a ¼ pixel precision reference image using different filters for each frame. Specifically, the relative positions of the fractional pixels between integer pixels (a, b, c, d, e, f, g, h, i, j, k, n, p, q, r in FIG. 1). A two-dimensional 6-tap filter is prepared for each position), and each filter is applied to a reference image with integer pixel accuracy to directly generate a reference image with 1/4 pixel accuracy. Although this technique can change the filter coefficient for each frame, it is fixed so that all 36 filter coefficients of the two-dimensional 6-tap filter at every fractional pixel position between integer pixels are always significant coefficients. Has been. In addition, the filter coefficient has horizontal and vertical objectivity, and the objectivity is uniquely fixed for each fractional pixel position between integer pixels. In practice, each frame has a 1/4 pixel accuracy reference. A total of 54 filter coefficients are encoded and decoded to generate an image.
Y. Vatis, B. Edler, D. Nguyen, J. Ostermann, “Motion-and Aliasing-Compensated Prediction Using a Two-Dimensional Non-Separable Adaptive Wiener Interpolation Filter”, Proc. ICIP2005, IEEE International Conference on Image Processing, Genova , Italy, September 2005.

  In the conventional encoding method, since a reference image with fractional pixel accuracy is always generated using a fixed filter, it is not possible to generate a reference image that matches the characteristics of each frame to be encoded. In Non-Patent Document 1, the reference image can be changed by changing the filter coefficient of a two-dimensional 6-tap filter for each fractional pixel position between integer pixels for each frame to be encoded. Since all filter coefficients are fixed so that they are encoded as significant coefficients, there is a lot of filter coefficient information, which reduces the coding efficiency and reduces the filter coefficient information. Therefore, there is a problem in that the degree of freedom of the filter is reduced because the number of fractional pixels is limited to a certain filter with limited objectivity in the horizontal and vertical directions.

  The present invention has been made in view of the above points. In the case where a reference image with fractional pixel accuracy is generated using a filter for each frame to be encoded, and motion compensation is performed for encoding and decoding, a filter coefficient is obtained. Video encoding apparatus, video encoding method, video encoding program, and video decoding apparatus for generating a filter fractional accuracy reference image having a high degree of freedom suitable for video characteristics Another object is to provide a moving picture decoding method and a moving picture decoding program.

The moving picture coding apparatus according to the present invention is a moving picture coding apparatus that creates a fractional pixel precision reference image using a filter and performs motion compensation on a moving image sequence formed of a time sequence of frame images. A filter pattern information storage unit that stores a plurality of filter patterns that are basic patterns of the filter, a filter pattern selection unit that selects a filter pattern from the plurality of filter patterns stored in the filter pattern information storage unit, and the filter A filter pattern information encoding unit that encodes information about the filter pattern selected by the pattern selection unit, and a filter based on the information about the filter pattern selected by the filter pattern selection unit, and a reference image with fractional pixel accuracy Create a frame image to compensate for motion and encode the frame image Comprising an encoding unit, wherein the filter pattern is characterized by a binary pattern of the combination of significant coefficients and insignificant coefficient.

In addition, the moving picture coding method of the present invention is a moving picture coding apparatus that creates a reference picture with fractional pixel accuracy using a filter for a moving picture sequence composed of a time sequence of frame images and performs motion compensation. And a filter pattern information storage step for storing a plurality of filter patterns, which are basic filter patterns, and a plurality of filter patterns stored in the filter pattern information storage step. A filter pattern selection step for selecting, a filter pattern information encoding step for encoding information on the filter pattern selected in the filter pattern selection step, and information on the filter pattern selected in the filter pattern selection step Generate filter and fraction By creating a reference image containing precision motion compensation, comprising: a frame image encoding step of encoding the frame image, and in that the filter pattern is a binary pattern of the combination of significant coefficients and insignificant coefficient Features.

Also, the moving picture coding program of the present invention causes a computer to create a fractional pixel precision reference picture using a filter for a moving picture sequence composed of a time sequence of frame images so as to compensate for motion. A moving image encoding program for functioning, comprising: a filter pattern information storage unit that stores a plurality of filter patterns that are basic filter patterns; and a plurality of filter patterns stored in the filter pattern information storage unit, Based on information on a filter pattern selection unit that selects a pattern, a filter pattern information encoding unit that encodes information on a filter pattern selected by the filter pattern selection unit, and a filter pattern selected by the filter pattern selection unit Filter to generate fractional pixel accuracy Creating a reference image motion compensation, as a frame picture coding unit for coding a frame image, to function, the filter pattern is characterized by a binary pattern of the combination of significant coefficients and insignificant coefficient .

According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when a fractional pixel precision reference image is generated using a filter and encoded with motion compensation, A basic pattern is selected from a plurality of candidates. Therefore, the filter pattern can be changed according to the characteristics of the moving image and the code amount assigned as filter information, and only the filter coefficient matched to the basic pattern of the filter needs to be encoded. It is possible to perform motion compensation with high accuracy while reducing the amount. Further, by encoding and outputting the filter pattern independently of the frame image code, it is possible to encode the frame image code while changing the importance.
In addition, when generating a reference image with fractional pixel accuracy using a filter and performing motion compensation and encoding, the basic pattern of the filter, which is a binary pattern of a combination of a significant coefficient and an insignificant coefficient, is a plurality of candidates. Selected from. Therefore, the significant coefficient pattern of the filter to be encoded can be changed according to the characteristics of the moving image and the code amount assigned as filter information, and only the filter coefficient at the position of the significant coefficient of the filter is encoded. Therefore, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern selection unit selects a filter pattern in units of a moving image sequence from a plurality of filter patterns stored in the filter pattern information storage unit. Preferably, the filter pattern information encoding unit encodes information about the filter pattern selected by the filter pattern selection unit in units of moving image sequences.

  In the moving image encoding method of the present invention, in the filter pattern selection step, a filter pattern is selected in a moving image sequence unit from a plurality of filter patterns stored in the filter pattern information storage step, and the filter pattern information In the encoding step, it is preferable that the information related to the filter pattern selected in the filter pattern selection step is encoded in units of moving image sequences.

  According to such a moving image encoding apparatus, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, a moving image A basic filter pattern is selected from a plurality of candidates in sequence units. Therefore, the filter pattern can be changed according to the characteristics of the moving image sequence, and only the filter coefficient matched to the basic pattern of the filter needs to be encoded. Therefore, the code amount of the filter information can be reduced with high accuracy. Motion compensation can be performed.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern selection unit is a frame in which a plurality of frame images are collected from a plurality of filter patterns stored in the filter pattern information storage unit. A filter pattern is selected in units of image groups, and the filter pattern information encoding unit encodes information related to the filter pattern selected in the filter pattern selection unit in units of frame images grouping a plurality of frame images. preferable.

  In the moving image encoding method of the present invention, in the filter pattern selection step, a filter pattern is selected in units of frame images in which a plurality of frame images are collected from the plurality of filter patterns stored in the filter pattern information storage step. In the filter pattern information encoding step, it is preferable that the information regarding the filter pattern selected in the filter pattern selection step is encoded in units of a frame image group in which a plurality of frame images are collected.

  According to such a moving image encoding apparatus, a moving image encoding method, and a moving image encoding program, a frame image is generated when a reference image with fractional pixel accuracy is generated using a filter and is encoded with motion compensation. The basic pattern of the filter is selected from a plurality of candidates for each frame image group in which a plurality of the above are grouped. Therefore, the filter pattern can be changed according to the characteristics of the frame image group, and only the filter coefficient matched to the basic pattern of the filter needs to be encoded. Therefore, the code amount of the filter information can be reduced with high accuracy. Motion compensation can be performed.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern selection unit generates a filter pattern for each frame image from a plurality of filter patterns stored in the filter pattern information storage unit. Preferably, the filter pattern information encoding unit encodes information about the filter pattern selected by the filter pattern selection unit in units of frame images.

  In the moving image encoding method of the present invention, in the filter pattern selection step, a filter pattern is selected in a frame image unit from a plurality of filter patterns stored in the filter pattern information storage step, and the filter pattern information code In the converting step, it is preferable that the information regarding the filter pattern selected in the filter pattern selecting step is encoded in units of frame images.

  According to such a moving image encoding apparatus, a moving image encoding method, and a moving image encoding program, a frame image is generated when a reference image with fractional pixel accuracy is generated using a filter and is encoded with motion compensation. A basic filter pattern is selected from a plurality of candidates in units. For this reason, the filter pattern can be changed according to the characteristics of the frame image, and only the filter coefficient matched to the basic pattern of the filter needs to be encoded. It becomes possible to compensate.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the frame image encoding unit is configured to generate a frame image for each frame image to be encoded based on the filter pattern selected by the filter pattern selection unit. It is preferable to generate a filter, create a reference image with fractional pixel accuracy, perform motion compensation, and encode a frame image.

  In the moving image encoding method of the present invention, in the frame image encoding step, a filter is generated for each frame image to be encoded based on the filter pattern selected in the filter pattern selection step, and fractional pixel accuracy is It is preferable to generate a reference image and compensate for motion to encode a frame image.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when a reference image with fractional pixel accuracy is generated using a filter and encoded with motion compensation, encoding is performed. A filter is generated for each frame image to be performed. Therefore, the filter can be changed according to the characteristics of the frame image, and motion compensation can be performed with high accuracy.

  In the video encoding device of the present invention and the video encoding program of the present invention, the frame image encoding unit divides a frame image to be encoded based on the filter pattern selected by the filter pattern selection unit. It is preferable to generate a filter for each slice obtained by collecting a plurality of blocks, create a reference image with fractional pixel accuracy, perform motion compensation, and encode a frame image.

  In the moving image encoding method of the present invention, in the frame image encoding step, for each slice in which a plurality of blocks obtained by dividing the frame image to be encoded are grouped based on the filter pattern selected in the filter pattern selection step. Preferably, a filter is generated, a reference image with fractional pixel accuracy is created to compensate for motion, and a frame image is encoded.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when a reference image with fractional pixel accuracy is generated using a filter and encoded with motion compensation, encoding is performed. A filter is generated for each slice in which a plurality of blocks into which the frame image to be divided is divided. Therefore, the filter can be changed according to the feature of the slice, and motion compensation can be performed with high accuracy.

  In the moving image encoding apparatus of the present invention and the moving image encoding program of the present invention, the filter pattern selection unit is a unit of an image sequence to be encoded from a plurality of filter patterns stored in the filter pattern information storage unit. The filter pattern information encoding unit encodes information about the filter pattern selected by the filter pattern selection unit in a moving image sequence unit, and the frame image encoding unit It is preferable to generate a filter for each frame image to be encoded based on the filter pattern selected by the selection unit, create a reference image with fractional pixel accuracy, perform motion compensation, and encode the frame image.

  In the moving image encoding method of the present invention, in the filter pattern selection step, a filter pattern is selected for each image sequence to be encoded from a plurality of filter patterns stored in the filter pattern information storage step, In the filter pattern information encoding step, information on the filter pattern selected in the filter pattern selection step is encoded in units of moving image sequences, and in the frame image encoding step, the filter pattern selected in the filter pattern selection step Based on the above, it is preferable to generate a filter for each frame image to be encoded, create a reference image with fractional pixel accuracy, perform motion compensation, and encode the frame image.

  According to such a moving image encoding apparatus, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, a moving image A basic filter pattern is selected from a plurality of candidates in sequence units, and a filter is generated for each frame image to be encoded. Therefore, the filter pattern can be changed according to the characteristics of the moving image sequence, the filter can be changed according to the characteristics of the frame image, and only the filter coefficients matching the basic pattern of the filter need be encoded. Thus, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the moving picture coding apparatus of the present invention and the moving picture coding program of the present invention, the filter pattern is preferably a binary pattern of a combination of a two-dimensional significant coefficient and a non-significant coefficient.

  The shape of the filter pattern (that is, the position of “1” (significant coefficient) in each pattern) desirably has the following characteristics. In other words, the symmetry of the filter pattern in the horizontal, vertical, and diagonal directions matches the symmetry in the horizontal, vertical, and diagonal directions of the fractional pixel position in the rectangle surrounded by the neighboring integer pixels. ing. Here, “symmetry” means a characteristic that is symmetric or asymmetric in a corresponding direction. With the above feature, it is possible to generate an appropriate filter according to the fractional pixel position of each fractional pixel and perform encoding.

  In the video encoding device of the present invention and the video encoding program of the present invention, the filter pattern information storage unit stores the plurality of filter patterns in association with identifiers, and the filter pattern information encoding unit It is preferable that the identifier associated with the filter pattern is encoded as information on the filter pattern.

  In the moving image encoding method of the present invention, in the filter pattern information storage step, the plurality of filter patterns are stored in association with identifiers, respectively, and in the filter pattern information encoding step, information about the filter pattern is as follows: Preferably, the identifier associated with the filter pattern is encoded.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when a fractional pixel precision reference image is generated using a filter and encoded with motion compensation, A basic pattern is selected from a plurality of candidates, and only its identifier is encoded. Therefore, the filter pattern can be changed according to the feature of the moving image and the code amount assigned as filter information, and only the identifier needs to be encoded. It becomes possible to perform motion compensation with high accuracy.

  In the video encoding device of the present invention and the video encoding program of the present invention, the filter pattern information storage unit stores the plurality of filter patterns in association with identifiers for each fractional pixel position between integer pixels. Then, the filter pattern selection unit selects one filter pattern from the plurality of filter patterns for each fractional pixel position between the integer pixels, and the filter pattern information encoding unit includes information on the filter pattern, The identifier associated with the filter pattern selected by the filter pattern selection unit is encoded for each fractional pixel position between the integer pixels, and the frame image encoding unit is configured for each fractional pixel position between the integer pixels. A filter based on the filter pattern associated with the identifier And motion compensation to create a precision of the reference image, it is preferable to encode the frame image.

  In the moving image encoding method of the present invention, in the filter pattern information storage step, the plurality of filter patterns are stored in association with identifiers for each fractional pixel position between integer pixels, and in the filter pattern selection step. , One filter pattern is selected from the plurality of filter patterns for each fractional pixel position between the integer pixels, and in the filter pattern information encoding step, information on the fractional pixels between the integer pixels is used as information about the filter pattern. The identifier associated with the filter pattern selected in the filter pattern selection step for each position is encoded, and the filter associated with the identifier for each fractional pixel position between the integer pixels in the frame image encoding step. Based on pattern Generates a filter, and motion compensation to create a reference image of a fractional pixel accuracy, it is preferable to encode the frame image.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, an integer pixel A basic filter pattern is selected from a plurality of candidates for each fractional pixel position, and only the identifier is encoded. Therefore, the filter pattern can be changed for each fractional pixel position between integer pixels according to the characteristics of the moving image and the code amount assigned as filter information, and only the identifier need be encoded. Thus, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern information storage unit includes a plurality of filter patterns as identifiers for each group in which the positions of fractional pixels between integer pixels are grouped. The filter pattern selection unit selects one filter pattern from the plurality of filter patterns for each group, and the filter pattern information encoding unit stores information on the filter pattern for each group. The identifier associated with the filter pattern selected by the filter pattern selection unit is encoded, and the frame image encoding unit is based on a filter pattern associated with the identifier for each fractional pixel position between the integer pixels. To create a reference image with fractional pixel accuracy Motion compensation Te, it is preferable to encode the frame image.

  In the moving image encoding method of the present invention, in the filter pattern information storage step, a plurality of filter patterns are stored in association with identifiers for each group in which the positions of fractional pixels between integer pixels are grouped, and the filter pattern In the selection step, one filter pattern is selected from the plurality of filter patterns for each group, and in the filter pattern information encoding step, information about the filter pattern is selected in the filter pattern selection step for each group. The identifier associated with the filter pattern is encoded, and in the frame image encoding step, a filter is generated based on the filter pattern associated with the identifier for each fractional pixel position between the integer pixels. By creating a reference image containing precision motion compensation, it is preferable to encode the frame image.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, an integer pixel A basic filter pattern is selected from a plurality of candidates for each group in which the positions of fractional pixels in between are grouped, and only the identifier is encoded. Therefore, the filter pattern can be changed for each group in which the positions of fractional pixels between integer pixels are grouped according to the characteristics of the moving image and the code amount assigned as filter information. Since it is only necessary to encode the identifier, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the video encoding device of the present invention and the video encoding program of the present invention, the filter pattern information storage unit stores a plurality of filter pattern groups obtained by combining the plurality of filter patterns in association with identifiers, respectively. It is preferable that the filter pattern information encoding unit encodes the identifier associated with the filter pattern group as information on the filter pattern.

  In the moving image encoding method of the present invention, in the filter pattern information storage step, a plurality of filter pattern groups obtained by combining the plurality of filter patterns are stored in association with identifiers, and in the filter pattern information encoding step. It is preferable to encode the identifier associated with the filter pattern group as information on the filter pattern.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, A filter pattern group obtained by combining filter patterns is selected from a plurality of candidates, and only the identifier is encoded. Therefore, the filter pattern group can be changed according to the feature of the moving image and the code amount assigned as the filter information, and only the identifier only needs to be encoded for each filter pattern group. It is possible to perform motion compensation with high accuracy while reducing the code amount.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern information storage unit includes a plurality of filter patterns obtained by combining the positions of fractional pixels between integer pixels and the association of the filter patterns. Each group is stored in association with an identifier, the filter pattern selection unit selects one filter pattern group from the plurality of filter pattern groups, and the filter pattern information encoding unit includes the filter pattern information as the information about the filter pattern, The identifier associated with the filter pattern group selected by the filter pattern selection unit is encoded, and the frame image encoding unit is based on the filter pattern associated with the identifier for each fractional pixel position between the integer pixels. Filter to create a reference image with fractional pixel accuracy. Compensated, it is preferable to encode the frame image.

  In the moving image encoding method of the present invention, in the filter pattern information storing step, a plurality of filter pattern groups obtained by combining the positions of fractional pixels between integer pixels and the association of the filter patterns are stored in association with identifiers. In the filter pattern selection step, one filter pattern group is selected from the plurality of filter pattern groups. In the filter pattern information encoding step, the filter pattern information encoding step is selected in the filter pattern selection step. The identifier associated with the filter pattern group is encoded, and in the frame image encoding step, a filter is generated based on the filter pattern associated with the identifier for each fractional pixel position between the integer pixels, and the fractional pixel Spirit And motion compensation to create a reference image, it is preferable to encode the frame image.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, an integer pixel A filter pattern group that combines the positions of the fractional pixels and the correspondence between the filter patterns is selected from a plurality of candidates, and only the identifier is encoded. Therefore, the filter pattern group associated with the position of the fractional pixel between the integer pixels can be changed according to the feature of the moving image and the code amount assigned as filter information, and the filter pattern group can be changed for each filter pattern group. Since only the identifier needs to be encoded, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern information storage unit combines a group in which the positions of fractional pixels between integer pixels are grouped and the association of the filter pattern. A plurality of filter pattern groups are stored in association with respective identifiers, and the filter pattern selection unit selects one filter pattern group from the plurality of filter pattern groups, thereby corresponding to the position of a fractional pixel between the integer pixels. The filter pattern information encoding unit encodes the identifier associated with the filter pattern group selected by the filter pattern selection unit as information on the filter pattern, and encodes the frame image The unit uses the identifier for each fractional pixel position between the integer pixels. Generating a filter based on continuous Tagged filter patterns, and motion compensation to create a reference image of a fractional pixel accuracy, it is preferable to encode the frame image.

  In the moving image encoding method of the present invention, in the filter pattern information storage step, a plurality of filter pattern groups each combining a group of fractional pixel positions between integer pixels and a combination of the filter patterns are respectively identified. In the filter pattern selection step, by selecting one filter pattern group from the plurality of filter pattern groups, the filter pattern corresponding to the position of the fractional pixel between the integer pixels is selected, In the filter pattern information encoding step, the identifier associated with the filter pattern group selected in the filter pattern selection step is encoded as information on the filter pattern, and in the frame image encoding step, the integer pixels Fractional pixel To generate a filter based on the filter patterns associated with the identifier for each position, and motion compensation to create a reference image of a fractional pixel accuracy, it is preferable to encode the frame image.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, an integer pixel A group of filter patterns obtained by combining the grouping of the positions of fractional pixels and the association of the filter patterns is selected from a plurality of candidates, and only the identifier is encoded. Therefore, the filter pattern group associated with the group in which the positions of the fractional pixels between the integer pixels are combined can be changed according to the feature of the moving image and the code amount assigned as filter information. Since it is only necessary to encode the identifier for each group, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern information encoding unit uses the combination of the two-dimensional significant coefficient and the insignificant coefficient as information on the filter pattern. It is preferable to encode the value pattern.

  In the moving image encoding method of the present invention, in the filter pattern information encoding step, the binary pattern of the combination of the two-dimensional significant coefficient and the insignificant coefficient may be encoded as information about the filter pattern. preferable.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when a reference image with fractional pixel accuracy is generated using a filter and encoded with motion compensation, the two-dimensional A filter pattern, which is a binary pattern of a combination of significant coefficients and insignificant coefficients, is selected from a plurality of candidates, and the pattern is encoded. Therefore, the pattern of the significant coefficient of the two-dimensional filter to be encoded can be changed according to the feature of the moving image and the code amount assigned as filter information, and the filter pattern can be directly encoded. It is possible to perform motion compensation with high accuracy without associating with the identifier.

  In the moving image encoding device of the present invention and the moving image encoding program of the present invention, the filter pattern code encoded by the filter pattern information encoding unit and the frame image encoded by the frame image encoding unit It is preferable to further include a code output unit that outputs the combined codes.

  In the moving image encoding method of the present invention, the code output for outputting the filter pattern code encoded in the filter pattern information encoding step and the frame image code encoded in the frame image encoding step together. It is preferable to further comprise a step.

  According to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, when a fractional pixel precision reference image is generated using a filter and encoded with motion compensation, A basic pattern is selected from a plurality of candidates, and the filter pattern changed according to the feature of the moving image and the code amount assigned as filter information is encoded and output as part of the encoded bitstream. Thus, the association with the frame image code can be facilitated.

The moving picture decoding apparatus according to the present invention is a moving picture decoding apparatus that generates a reference image with fractional pixel accuracy using a filter, performs motion compensation, and decodes a moving picture sequence including a time series of frame images. A filter pattern information decoding unit that decodes a filter pattern code obtained by encoding information on a filter pattern, which is a basic pattern of the filter, and obtains information on the filter pattern; and a filter pattern information decoding unit A frame image decoding unit that generates a filter based on information on the decoded filter pattern, creates a reference image with fractional pixel accuracy, performs motion compensation, and decodes the frame image, and the filter pattern includes: The binary pattern is a combination of a significant coefficient and an insignificant coefficient .

In addition, the moving picture decoding method of the present invention creates a reference picture with fractional pixel accuracy using a filter, performs motion compensation, and decodes a moving picture sequence composed of a time sequence of frame images. A filter pattern information decoding step for decoding a filter pattern code obtained by encoding information on a filter pattern, which is a basic pattern of a filter, and obtaining information on the filter pattern, and a filter pattern information decoding to generate a filter based on the filter pattern relates information decoded in step and motion compensation to create a reference image of a fractional pixel accuracy, with a frame image decoding step for decoding the frame image, wherein the filter pattern it is characterized by a binary pattern of the combination of significant coefficients and insignificant coefficient

Also, the moving picture decoding program of the present invention generates a fractional pixel precision reference image using a filter to compensate for motion, and decodes a moving image formed of a time sequence of frame images. A moving picture decoding program for causing a computer to decode a filter pattern code obtained by encoding information related to a filter pattern, which is a basic filter pattern, and obtain information related to the filter pattern Frame image decoding for generating a filter based on the information on the filter pattern decoded by the conversion unit and the filter pattern information decoding unit, creating a reference image with fractional pixel accuracy, motion compensation, and decoding the frame image as part, to function, the filter pattern, combination of significant coefficients and insignificant coefficient Characterized in that it is a binary pattern of.

According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, a filter pattern is generated when a reference picture with fractional pixel accuracy is generated using a filter to perform motion compensation and decoding. A code is decoded, information about the filter pattern is obtained, a filter is generated based on the obtained information about the filter pattern, a reference image with fractional pixel accuracy is created, motion compensation is performed, and a frame image is decoded. The filter information can be decoded according to the characteristics of the moving image and the code amount assigned as filter information, and only the filter coefficients that match the basic pattern of the filter need to be decoded. It is possible to perform motion compensation with high accuracy while reducing the amount of codes. Furthermore, by inputting and decoding the filter pattern independently of the frame image code, it is possible to decode the frame image code while changing the importance.
In addition, when generating a reference image with fractional pixel accuracy using a filter and decoding with motion compensation, the basic pattern of the filter, which is a binary pattern of a combination of significant coefficients and insignificant coefficients, is decoded. Therefore, it is possible to decode the pattern of the significant coefficient of the filter according to the feature of the moving image and the code amount assigned as the filter information, and it is sufficient to decode only the filter coefficient at the position of the significant coefficient of the filter. Therefore, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the video decoding device of the present invention and the video decoding program of the present invention, it is preferable that the filter pattern information decoding unit decodes information related to the filter pattern in units of video sequences.

  In the moving picture decoding method of the present invention, it is preferable that in the filter pattern information decoding step, information relating to the filter pattern is decoded in units of moving picture sequences.

  According to such a moving image decoding apparatus, moving image decoding method, and moving image decoding program, when generating a reference image with fractional pixel accuracy using a filter and performing motion compensation for decoding, a moving image The basic pattern of the filter is decoded in sequence units. For this reason, the filter pattern can be changed according to the characteristics of the moving image sequence, and only the filter coefficient matched to the basic pattern of the filter needs to be decoded. Motion compensation can be performed.

  In the video decoding device of the present invention and the video decoding program of the present invention, the filter pattern information decoding unit decodes information related to the filter pattern in units of a frame image group in which a plurality of frame images are collected. It is preferable.

  In the moving picture decoding method of the present invention, it is preferable that in the filter pattern information decoding step, information relating to the filter pattern is decoded in units of a frame image group in which a plurality of frame images are collected.

  According to such a moving image decoding apparatus, moving image decoding method, and moving image decoding program, when generating a fractional pixel precision reference image using a filter and decoding with motion compensation, a frame image The basic pattern of the filter is decoded for each frame image group in which a plurality of frames are grouped. Therefore, the filter pattern can be changed in accordance with the characteristics of the frame image group, and only the filter coefficient matched to the basic pattern of the filter needs to be decoded. Therefore, the code amount of the filter information can be reduced with high accuracy. Motion compensation can be performed.

  In the video decoding device of the present invention and the video decoding program of the present invention, it is preferable that the filter pattern information decoding unit decodes information on the filter pattern in units of frame images.

  In the moving picture decoding method of the present invention, it is preferable that in the filter pattern information decoding step, information relating to the filter pattern is decoded on a frame image basis.

  According to such a moving image decoding apparatus, moving image decoding method, and moving image decoding program, when generating a fractional pixel precision reference image using a filter and decoding with motion compensation, a frame image The basic pattern of the filter is decoded in units. For this reason, the filter pattern can be changed according to the characteristics of the frame image, and only the filter coefficient matched to the basic pattern of the filter needs to be decoded. It becomes possible to compensate.

  In the video decoding device of the present invention and the video decoding program of the present invention, the frame image decoding unit performs decoding based on information about the filter pattern decoded by the filter pattern information decoding unit. It is preferable to generate a filter for each frame image, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  In the moving image decoding method of the present invention, in the frame image decoding step, a filter is generated for each frame image to be decoded based on information on the filter pattern decoded in the filter pattern information decoding step. It is preferable to generate a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, decoding is performed. A filter is generated for each frame image to be performed. Therefore, the filter can be changed according to the characteristics of the frame image, and motion compensation can be performed with high accuracy.

  In the video decoding device of the present invention and the video decoding program of the present invention, the frame image decoding unit performs decoding based on information about the filter pattern decoded by the filter pattern information decoding unit. It is preferable to generate a filter for each slice obtained by collecting a plurality of blocks obtained by dividing a frame image to be decoded, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  In the moving image decoding method of the present invention, in the frame image decoding step, a frame image to be decoded is divided based on information about the filter pattern decoded in the filter pattern information decoding step. It is preferable to generate a filter for each slice obtained by collecting a plurality of blocks, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, decoding is performed. A filter is generated for each slice in which a plurality of blocks into which the frame image to be divided is divided. Therefore, the filter can be changed according to the feature of the slice, and motion compensation can be performed with high accuracy.

  In the video decoding device of the present invention and the video decoding program of the present invention, the filter pattern information decoding unit decodes information on the filter pattern in units of a video sequence, and the frame image decoding unit Generates a filter for each frame image to be decoded based on information on the filter pattern decoded by the filter pattern information decoding unit, creates a reference image with fractional pixel accuracy, performs motion compensation, Decoding is preferred.

  In the moving picture decoding method of the present invention, in the filter pattern information decoding step, information on the filter pattern is decoded in a moving picture sequence unit, and in the frame image decoding step, the filter pattern information decoding is performed. It is preferable to generate a filter for each frame image to be decoded based on information about the filter pattern decoded in the step, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  According to such a moving image decoding apparatus, moving image decoding method, and moving image decoding program, when generating a reference image with fractional pixel accuracy using a filter and performing motion compensation for decoding, a moving image A basic filter pattern is decoded in sequence units, and a filter is generated for each frame image to be decoded. Therefore, the filter pattern can be changed according to the characteristics of the moving image sequence, the filter can be changed according to the characteristics of the frame image, and only the filter coefficients corresponding to the basic pattern of the filter need to be decoded. Thus, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  The moving picture decoding apparatus of the present invention and the moving picture decoding program of the present invention further include a filter pattern information storage unit that stores a plurality of filter patterns in association with identifiers of the plurality of filter patterns. It is preferable that the pattern information decoding unit decodes the identifier associated with the filter pattern as information on the filter pattern.

  The moving image decoding method of the present invention further includes a filter pattern information storage step of storing a plurality of filter patterns in association with identifiers of the plurality of filter patterns, and in the filter pattern information decoding step, It is preferable to decode the identifier associated with the filter pattern as information on the filter pattern.

  According to such a moving image decoding apparatus, moving image decoding method, and moving image decoding program, when generating a reference image with fractional pixel accuracy using a filter and decoding with motion compensation, A basic pattern is selected from a plurality of candidates, and only its identifier is decoded. Therefore, the filter pattern can be changed according to the feature of the moving image and the code amount assigned as filter information, and only the identifier needs to be decoded, so reducing the code amount of the filter information, It becomes possible to perform motion compensation with high accuracy.

  In the video decoding device of the present invention and the video decoding program of the present invention, the filter pattern information storage unit stores the plurality of filter patterns in association with identifiers for each fractional pixel position between integer pixels. The filter pattern information decoding unit decodes the identifier associated with the filter pattern for each fractional pixel position between the integer pixels as information on the filter pattern, and the frame image decoding unit It is preferable to generate a filter based on a filter pattern associated with the identifier for each fractional pixel position between integer pixels, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  In the moving image decoding method of the present invention, in the filter pattern information storage step, the plurality of filter patterns are stored in association with identifiers for each fractional pixel position between integer pixels, and the filter pattern information decoding is performed. In the step, as the information on the filter pattern, the identifier associated with the filter pattern is decoded for each fractional pixel position between the integer pixels, and in the frame image decoding step, the fractional pixel position between the integer pixels Preferably, each time a filter is generated based on a filter pattern associated with the identifier, a reference image with fractional pixel accuracy is created, motion compensation is performed, and a frame image is decoded.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, an integer pixel A basic filter pattern is selected from a plurality of candidates for each fractional pixel position, and only the identifier is decoded. Therefore, the filter pattern can be changed for each fractional pixel position between the integer pixels according to the characteristics of the moving image and the code amount assigned as filter information, and only the identifier only has to be decoded. Thus, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the moving picture decoding apparatus of the present invention and the moving picture decoding program of the present invention, the filter pattern information storage unit includes a plurality of filter patterns as identifiers for each group in which positions of fractional pixels between integer pixels are grouped. The filter pattern information decoding unit decodes the identifier associated with the filter pattern for each group as information about the filter pattern, and the frame image decoding unit It is preferable to generate a filter based on the filter pattern associated with the identifier for each fractional pixel position, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  In the moving picture decoding method of the present invention, in the filter pattern information storing step, a plurality of filter patterns are stored in association with identifiers for each group in which the positions of fractional pixels between integer pixels are grouped, and the filter pattern In the information decoding step, as the information regarding the filter pattern, the identifier associated with the filter pattern is decoded for each group, and in the frame image decoding step, the identifier is determined for each fractional pixel position between the integer pixels. It is preferable to generate a filter based on the filter pattern associated with, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, an integer pixel A basic filter pattern is selected from a plurality of candidates for each group in which the positions of fractional pixels in between are grouped, and only the identifier is decoded. Therefore, the filter pattern can be changed for each group in which the positions of fractional pixels between integer pixels are grouped according to the characteristics of the moving image and the code amount assigned as filter information. Since only the identifier needs to be decoded, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the video decoding device of the present invention and the video decoding program of the present invention, the filter pattern information storage unit stores a plurality of filter pattern groups obtained by combining the plurality of filter patterns in association with identifiers, respectively. Preferably, the filter pattern information decoding unit decodes the identifier associated with the filter pattern group as information on the filter pattern.

  In the moving picture decoding method of the present invention, in the filter pattern information storage step, a plurality of filter pattern groups obtained by combining the plurality of filter patterns are stored in association with identifiers, respectively, and in the filter pattern information decoding step, It is preferable to decode the identifier associated with the filter pattern group as information on the filter pattern.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when a fractional pixel precision reference picture is generated using a filter and motion compensated for decoding, A group of filter patterns obtained by combining filter patterns is selected from a plurality of candidates, and only the identifier is decoded. Therefore, the filter pattern group can be changed according to the feature of the moving image and the code amount assigned as the filter information, and only the identifier for each filter pattern group needs to be decoded. It is possible to perform motion compensation with high accuracy while reducing the code amount.

  In the moving picture decoding apparatus of the present invention and the moving picture decoding program of the present invention, the filter pattern information storage unit includes a plurality of filter patterns in which the positions of fractional pixels between integer pixels and combinations of the filter patterns are combined. Each group is stored in association with an identifier, and the filter pattern information decoding unit decodes the identifier associated with the filter pattern group as information on the filter pattern, and the frame image decoding unit It is preferable to generate a filter based on a filter pattern associated with the identifier for each fractional pixel position between pixels, create a reference image with fractional pixel accuracy, perform motion compensation, and decode a frame image.

In the moving picture decoding method of the present invention, in the filter pattern information storing step, a plurality of filter pattern groups obtained by combining the positions of fractional pixels between integer pixels and the association of the filter patterns are stored in association with identifiers. And
In the filter pattern information decoding step, the identifier associated with the filter pattern group is decoded as information on the filter pattern, and in the frame image decoding step, the fractional pixel positions between the integer pixels are It is preferable to generate a filter based on the filter pattern associated with the identifier, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, an integer pixel A filter pattern group that combines the positions of the fractional pixels in between and the correspondence between the filter patterns is selected from a plurality of candidates, and only the identifier is decoded. Therefore, the filter pattern group associated with the position of the fractional pixel between the integer pixels can be changed according to the feature of the moving image and the code amount assigned as filter information, and the filter pattern group can be changed for each filter pattern group. Since only the identifier needs to be decoded, motion compensation can be performed with high accuracy while reducing the code amount of the filter information.

  In the moving picture decoding apparatus of the present invention and the moving picture decoding program of the present invention, the filter pattern information storage unit combines a group in which the positions of fractional pixels between integer pixels are grouped and the correspondence of the filter pattern. Each of the plurality of filter pattern groups is stored in association with an identifier, and the filter pattern information decoding unit decodes the identifier associated with the filter pattern group as information on the filter pattern, and the frame image decoding unit Generates a filter based on a filter pattern associated with the identifier for each fractional pixel position between the integer pixels, creates a fractional pixel precision reference image, performs motion compensation, and decodes a frame image Is preferred.

  In the moving picture decoding method of the present invention, in the filter pattern information storing step, a plurality of filter pattern groups each combining a group of fractional pixel positions between integer pixels and a combination of the filter patterns are respectively identified. In the filter pattern information decoding step, the identifier associated with the filter pattern group is decoded as information on the filter pattern, and in the frame image decoding step, the fraction between the integer pixels is stored. It is preferable to generate a filter based on a filter pattern associated with the identifier for each pixel position, create a reference image with fractional pixel accuracy, perform motion compensation, and decode the frame image.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, an integer pixel A group of filter patterns obtained by combining the grouping of the positions of the fractional pixels between them and the association of the filter patterns is selected from a plurality of candidates, and only the identifier is decoded. Therefore, the filter pattern group associated with the group in which the positions of the fractional pixels between the integer pixels are combined can be changed according to the feature of the moving image and the code amount assigned as filter information. Since only the identifier needs to be decoded for each group, it is possible to perform motion compensation with high accuracy while reducing the code amount of the filter information.

  In the video decoding device of the present invention and the video decoding program of the present invention, the filter pattern information decoding unit decodes a binary pattern of a filter pattern that is information related to the filter pattern, and the two-dimensional A combination of a significant coefficient and a non-significant coefficient is preferable. In this case, the moving picture decoding apparatus of the present invention and the moving picture decoding program of the present invention do not require the filter pattern information storage unit.

  In the moving image decoding method of the present invention, in the filter pattern information decoding step, a binary pattern of a filter pattern which is information on the filter pattern is decoded, and the combination of the two-dimensional significant coefficient and the non-significant coefficient It is preferable that In this case, the moving picture decoding method of the present invention does not require the filter pattern information storing step.

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a reference picture with fractional pixel accuracy using a filter and performing motion compensation for decoding, the two-dimensional A filter pattern which is a binary pattern of a combination of the significant coefficient and the insignificant coefficient is decoded. Therefore, the pattern of the significant coefficient of the two-dimensional filter to be decoded can be changed according to the feature of the moving image and the code amount assigned as filter information, and the filter pattern can be directly decoded. It is possible to perform motion compensation with high accuracy without associating with the identifier.

  In the moving picture decoding apparatus of the present invention and the moving picture decoding program of the present invention, an input code is encoded with a filter pattern code obtained by encoding information on the filter pattern, and the frame image is encoded. It is preferable to further include a code separation unit that separates the frame image code obtained in this way.

  In the moving image decoding method of the present invention, the input code is divided into a filter pattern code obtained by encoding information on the filter pattern, and a frame image code obtained by encoding the frame image. It is preferable to further include a code separation step for separating the data into

  According to such a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program, when generating a fractional pixel precision reference picture using a filter and decoding with motion compensation, the coding is performed. By decoding the filter pattern from a part of the bit stream, it is possible to facilitate the association with the frame image code.

  According to such a moving image encoding device, a moving image encoding method, a moving image encoding program, a moving image decoding device, a moving image decoding method, and a moving image decoding program, a fractional pixel accuracy is achieved using a filter. When a reference image is generated and motion-compensated for encoding and decoding, motion compensation can be performed with high accuracy using a filter that matches the video characteristics while reducing the code amount of the filter coefficient.

  Referring to the drawings, a moving picture coding apparatus, a moving picture coding method, a moving picture coding program, a moving picture decoding apparatus, a moving picture decoding method, and a moving picture decoding program according to the first embodiment of the present invention will be described. I will explain. In addition, in each figure, the same code | symbol is attached | subjected to the same element and the overlapping description is abbreviate | omitted.

  FIG. 2 is a block diagram showing a configuration of the moving picture encoding apparatus 20 according to the present embodiment. The moving image encoding device 20 includes, as functional components, a filter pattern information storage unit (filter pattern information storage unit) 201, a filter pattern selection unit (filter pattern selection unit) 202, and a filter pattern information encoding unit ( A filter pattern information encoding unit) 203 and a frame image encoding unit (frame image encoding unit) 204.

  The filter pattern information storage unit 201 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, in association with an identifier for each fractional pixel position between integer pixels.

  Specifically, the filter pattern information storage unit 201 has three types of two-dimensional 6-tap filters of the pattern a-00, the pattern a-01, and the pattern a-10 in FIG. A filter pattern is stored. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the pattern a-00, the pattern a-01, and the pattern a-10 in FIG. 3 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores the filter patterns of the three types of two-dimensional 6-tap filters of the pattern b-00, the pattern b-01, and the pattern b-10 in FIG. 4 at the fractional pixel position b in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern b-00, the pattern b-01, and the pattern b-10 in FIG. 4 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern c-00, the pattern c-01, and the pattern c-10 in FIG. 5 at the fractional pixel position c in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern c-00, the pattern c-01, and the pattern c-10 in FIG. 5 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern d-00, the pattern d-01, and the pattern d-10 in FIG. 6 at the fractional pixel position d in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern d-00, the pattern d-01, and the pattern d-10 in FIG. 6 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern e-00, the pattern e-01, and the pattern e-10 in FIG. 7 for the fractional pixel position e in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. In addition, identifiers corresponding to the filter patterns of the pattern e-00, the pattern e-01, and the pattern e-10 in FIG. 7 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern f-00, the pattern f-01, and the pattern f-10 in FIG. 8 for the fractional pixel position f in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, the identifiers corresponding to the filter patterns of the pattern f-00, the pattern f-01, and the pattern f-10 in FIG. 8 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern g-00, the pattern g-01, and the pattern g-10 in FIG. 9 for the fractional pixel position g in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the pattern g-00, the pattern g-01, and the pattern g-10 in FIG. 9 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern h-00, the pattern h-01, and the pattern h-10 in FIG. 10 for the fractional pixel position h in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern h-00, the pattern h-01, and the pattern h-10 in FIG. 10 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern i-00, the pattern i-01, and the pattern i-10 in FIG. 11 at the fractional pixel position i in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Further, identifiers corresponding to the filter patterns of the pattern i-00, the pattern i-01, and the pattern i-10 in FIG. 11 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern j-00, the pattern j-01, and the pattern j-10 in FIG. 12 for the fractional pixel position j in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern j-00, the pattern j-01, and the pattern j-10 in FIG. 12 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern k-00, the pattern k-01, and the pattern k-10 in FIG. 13 for the fractional pixel position k in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the pattern k-00, the pattern k-01, and the pattern k-10 in FIG. 13 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern n-00, the pattern n-01, and the pattern n-10 in FIG. 14 at the fractional pixel position n in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern n-00, the pattern n-01, and the pattern n-10 in FIG. 14 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern p-00, the pattern p-01, and the pattern p-10 in FIG. 15 for the fractional pixel position p in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Further, identifiers corresponding to the filter patterns of the pattern p-00, the pattern p-01, and the pattern p-10 in FIG. 15 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern q-00, the pattern q-01, and the pattern q-10 in FIG. 16 for the fractional pixel position q in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the patterns q-00, q-01, and pattern q-10 in FIG. 16 are 00, 01, and 10, respectively.

  In addition, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern r-00, the pattern r-01, and the pattern r-10 in FIG. 17 for the fractional pixel position r in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern r-00, the pattern r-01, and the pattern r-10 in FIG. 17 are 00, 01, and 10, respectively.

  The shape of the filter pattern shown in FIGS. 3 to 17 (that is, the position of “1” (significant coefficient) in each pattern) has the following characteristics. In other words, the symmetry of the filter pattern in the horizontal, vertical, and diagonal directions matches the symmetry in the horizontal, vertical, and diagonal directions of the fractional pixel position in the rectangle surrounded by the neighboring integer pixels. ing. Here, “symmetry” means a characteristic that is symmetric or asymmetric in a corresponding direction.

  For example, the filter pattern for the fractional pixel position b shown in FIG. 4 is symmetric in the horizontal direction and asymmetrical in the vertical and diagonal directions. In FIG. 1, the fractional pixel position b is a neighboring integer pixel (for example, the integer pixel G). , M, N, and H) are symmetric in the horizontal direction and asymmetric in the vertical and diagonal directions. As described above, the horizontal, vertical, and diagonal symmetries of the filter pattern with respect to the fractional pixel position b coincide with the horizontal, vertical, and diagonal symmetries of the fractional pixel position b. .

  Further, the filter pattern for the fractional pixel position e shown in FIG. 7 is asymmetric in the horizontal and vertical directions and symmetrical in the lower right diagonal direction. In FIG. The quadrangle surrounded by the pixels G, M, N, and H) is asymmetric in the horizontal and vertical directions and symmetric in the lower right diagonal direction. As described above, the horizontal, vertical, and diagonal symmetry of the filter pattern for the fractional pixel position e coincides with the horizontal, vertical, and diagonal symmetry of the fractional pixel position e. .

  Also, the filter pattern for the fractional pixel position a shown in FIG. 3 is asymmetric in the horizontal, vertical, and diagonal directions. In FIG. 1, the fractional pixel position a is a neighboring integer pixel (for example, integer pixels G, M, N, H) are asymmetric in the horizontal, vertical and diagonal directions. Thus, the horizontal, vertical, and diagonal symmetries of the filter pattern with respect to the fractional pixel position a coincide with the horizontal, vertical, and diagonal symmetries of the fractional pixel position a. .

  As described above, the symmetry of the filter pattern in the horizontal direction, the vertical direction, and the diagonal direction is different from the symmetry of the horizontal, vertical direction, and diagonal direction of the fractional pixel position in the rectangle surrounded by the neighboring integer pixels. By matching, it is possible to generate an appropriate filter according to the fractional pixel position of each fractional pixel and perform encoding.

  In the filter pattern for each fractional pixel position, the number of “1” s (significant coefficients) is the same. For example, in the filter patterns j-00, j-01, and j-10 for the fractional pixel position j shown in FIG. 12, the number of “1” s (significant coefficients) is 8 and the same. Further, in each of the filter patterns shown in FIGS. 3 to 11 and FIGS. 13 to 17, the number of “1” (significant coefficient) is 6, which is the same. Thus, when one filter pattern is selected from a plurality of filter pattern candidates for a certain fractional pixel position, the number of “1” s (significant coefficients) in each filter pattern candidate is the same, so that selection is easy. Become.

  The filter pattern selection unit 202 is stored in the filter pattern information storage unit 201 for each fractional pixel position between integer pixels in a unit of the image sequence to be encoded with respect to the moving image sequence 205 to be encoded input from the outside. One filter pattern is selected from the filter patterns, and the filter pattern identifier 206, which is the identifier of the selected filter pattern, is output to the filter pattern information encoding unit 203 and the frame image encoding unit 204.

  The filter pattern information encoding unit 203 encodes the filter pattern identifier 206 for each fractional pixel position between integer pixels input from the filter pattern selection unit 202 in units of image sequences to be encoded, and outputs the filter pattern code 207 to the outside. Output.

  The frame image encoding unit 204 determines a filter pattern associated with the filter pattern identifier 206 for each fractional pixel position between integer pixels input from the filter pattern selection unit 202 with reference to the filter pattern information storage unit 201. The encoding target moving image sequence 205 input from the outside is encoded using the determined filter pattern. The frame image encoding unit 204 outputs the encoded frame image code 208 to the outside.

  Specifically, the frame image encoding unit 204 encodes a filter coefficient of one part that is a significant coefficient in the filter pattern associated with the filter pattern identifier 206 for each fractional pixel position between integer pixels. Generated for each image. At this time, when the frame image encoding unit 204 creates a reference image with fractional pixel accuracy by filtering a reference image encoded in the past, an error from the signal of the frame image to be encoded is minimized. The filter coefficient of the significant coefficient 1 part in the filter pattern is generated.

  For example, when 00 is input as the filter pattern identifier 206 for the fractional pixel position a in FIG. 1, the frame image encoding unit 204 refers to the filter pattern information storage unit 201 to determine the pattern a-00 in FIG. . Next, the frame image encoding unit 204 temporarily creates a reference image with fractional pixel accuracy for a reference image encoded in the past using a filter used in the previous encoding method, The motion search is performed using the frame image to be converted and the reference image temporarily generated with fractional pixel accuracy. Next, the frame image encoding unit 204 generates a pixel signal of a reference image having a fractional pixel accuracy that minimizes the error by the motion search at the fractional pixel position a in FIG. 1 with respect to the pixel signal in the frame image to be encoded. If there is a filter, apply a filter coefficient value to the 1 part which is a significant coefficient of the pattern a-00 to the 6 × 6 integer pixels around the fractional pixel position a and apply the filter to the pixel of the frame pixel to be encoded. The filter coefficient value of the 1 part which is a significant coefficient is determined so that the error between the signal and the filtered pixel signal is minimized in the entire frame to be encoded. Thereby, the filter of the fractional pixel position a with respect to the encoding target frame image is generated. Further, when 01 or 10 is input as the filter pattern identifier 206 for the fractional pixel position a in FIG. 1, the frame image encoding unit 204 similarly performs the part of the significant coefficient 1 of the pattern a-01 or the pattern a-10. Calculate the filter coefficient value of. In addition, for fractional pixel positions b, c, d, e, f, g, h, i, j, k, n, p, q, r in FIG. A filter coefficient value of a portion of 1 which is a significant coefficient of the pattern is generated.

  The frame image encoding unit 204 creates a reference image with fractional pixel accuracy using a filter coefficient generated for each fractional pixel position between integer pixels with respect to the reference image, and performs motion compensation to generate a frame image to be encoded. Encode. At this time, among the filter coefficients for each fractional pixel position between the integer pixels generated for each frame image, the filter coefficient value of the 1 part which is a significant coefficient is also encoded, and the frame image code 208 is transmitted to the outside. Output.

  The frame image encoding unit 204 generates, for each frame image to be encoded, one part of the filter coefficient that is a significant coefficient in the filter pattern associated with the filter pattern identifier 206 for each fractional pixel position between integer pixels. The method is not limited to a method that minimizes the error from the signal of the frame image to be encoded when a reference image with a fractional pixel accuracy is created by filtering a reference image encoded in the past. Other methods are also acceptable.

  Next, the operation of the moving picture coding method according to the present embodiment will be described with reference to FIG. FIG. 18 is a flowchart for explaining a moving picture coding method that is an operation of the moving picture coding apparatus 20 according to the present embodiment.

  First, the filter pattern information storage unit 201 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, in association with an identifier for each fractional pixel position between integer pixels (step S1801).

  Next, the filter pattern selection unit 202 stores the encoding target moving image sequence 205 in the filter pattern information storage unit 201 for each fractional pixel position between integer pixels in the encoding target image sequence unit. One filter pattern is selected from the stored filter patterns, and its identifier is determined (step S1802).

  Next, the filter pattern information encoding unit 203 encodes the identifier of the filter pattern for each fractional pixel position between the integer pixels selected by the filter pattern selection unit 202 for each image sequence to be encoded, and outputs the encoded image to the outside ( Step S1803).

  Further, the frame image encoding unit 204 determines the filter pattern of the filter pattern identifier for each fractional pixel position between the integer pixels selected by the filter pattern selection unit 202 with reference to the filter pattern information storage unit 201, and externally Is encoded using the determined filter pattern and output to the outside (step S1804).

  Next, step S1802, which is the operation of the filter pattern selection unit 202, will be described in detail with reference to FIG. FIG. 19 is a flowchart for explaining step S1802, which is the operation of the filter pattern selection unit 202 according to this embodiment.

  First, the filter pattern selection unit 202 inputs a moving image sequence 205 to be encoded from the outside (step S1901).

  Next, the filter pattern selection unit 202 selects one filter pattern from the filter patterns stored in the filter pattern information storage unit 201 at one fractional pixel position between integer pixels (step S1902), and the identifier of the selected filter pattern Is output to the filter pattern information encoding unit 203 and the frame image encoding unit 204 (step S1903).

  Next, the filter pattern selection unit 202 determines whether a filter pattern is selected at a fractional pixel position between all integer pixels (step S1904), and no filter pattern is selected at a fractional pixel position between all integer pixels. If yes (NO), repeat from step S1902 for the fractional pixel position between the next integer pixels. If a filter pattern is selected at a fractional pixel position between all integer pixels (YES), the process ends.

  Next, step S1803, which is the operation of the filter pattern information encoding unit 203, will be described in detail with reference to FIG. FIG. 20 is a flowchart for explaining step S1803 which is the operation of the filter pattern selection unit 202 according to this embodiment.

  First, the filter pattern information encoding unit 203 encodes the identifier of the filter pattern at the fractional pixel position between one integer pixel with respect to the moving image to be encoded (step S2001), and outputs it (step S2002).

  Next, the filter pattern information encoding unit 203 determines whether or not the identifier of the filter pattern has been encoded at the fractional pixel positions between all integer pixels (step S2003), and the filter pattern at the fractional pixel positions between all integer pixels. If the identifier is not encoded (NO), the process is repeated from step S2001 for the fractional pixel position between the next integer pixels. If the filter pattern identifier is encoded at the fractional pixel position between all integer pixels (YES), the process ends.

  Next, step S1804, which is the operation of the frame image encoding unit 204, will be described in detail with reference to FIG. FIG. 21 is a flowchart illustrating step S1804, which is the operation of the frame image encoding unit 204 according to the present embodiment.

  First, the frame image encoding unit 204 determines a filter pattern by referring to the filter pattern information storage unit 201 from the identifier of the filter pattern at the fractional pixel position between one integer pixel (step S2101), and the determined filter pattern The filter coefficient of 1 part, which is a significant coefficient, is generated for each frame image to be encoded, and the filter is determined (step S2102). In addition, among the determined filters, the filter coefficient of 1 part which is a significant coefficient in the filter pattern is encoded (step S2105).

  Next, the frame image encoding unit 204 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S2104), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), repeat from step S2101 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image encoding unit 204 applies a filter for each determined fractional pixel position between integer pixels to a reference image that has been encoded in the past, thereby generating a reference image with fractional pixel accuracy (step S2105).

  Next, the frame image encoding unit 204 encodes the moving image sequence 205 to be encoded by performing motion compensation using the reference image with fractional pixel accuracy (step S2106), and matches the encoded filter coefficient with the encoded code. To the outside (step S2107), and the process is terminated.

  Next, a video encoding program 220 for causing a computer to function as the above-described video encoding device 20 will be described. FIG. 22 is a diagram showing a configuration of the moving image encoding program 220.

  As shown in FIG. 22, the moving image encoding program 220 includes a main module program 2201 that supervises processing, a filter pattern information storage module 2202, a filter pattern selection module 2203, a filter pattern information encoding module 2204, A frame image encoding module 2205. The filter pattern information storage module 2202, the filter pattern selection module 2203, the filter pattern information encoding module 2204, and the frame image encoding module 2205 have functions corresponding to the above-described filter pattern information storage units 201. The filter pattern selection unit 202, the filter pattern information encoding unit 203, and the frame image encoding unit 204 are the same.

  FIG. 23 is a block diagram illustrating a configuration of the video decoding device 230 according to the present embodiment. The moving picture decoding apparatus 230 includes, as its functional components, a filter pattern information storage unit (filter pattern storage unit) 2301, a filter pattern information decoding unit (filter pattern information decoding unit) 2302, frame image decoding And a conversion unit (frame image decoding means) 2303.

  The filter pattern information storage unit 2301 stores a plurality of filter patterns that are patterns of significant coefficients of a filter of a two-dimensional matrix for each fractional pixel position between integer pixels in association with an identifier. The filter pattern information storage unit 2301 is the same as the filter pattern information storage unit 201 in the moving image encoding device 20.

  Specifically, the filter pattern information storage unit 2301 has three types of two-dimensional 6-tap filters of the pattern a-00, the pattern a-01, and the pattern a-10 in FIG. A filter pattern is stored. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the pattern a-00, the pattern a-01, and the pattern a-10 in FIG. 3 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores the filter patterns of the three types of two-dimensional 6-tap filters of the pattern b-00, the pattern b-01, and the pattern b-10 in FIG. 4 at the fractional pixel position b in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern b-00, the pattern b-01, and the pattern b-10 in FIG. 4 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern c-00, the pattern c-01, and the pattern c-10 in FIG. 5 at the fractional pixel position c in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern c-00, the pattern c-01, and the pattern c-10 in FIG. 5 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern d-00, the pattern d-01, and the pattern d-10 in FIG. 6 at the fractional pixel position d in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern d-00, the pattern d-01, and the pattern d-10 in FIG. 6 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern e-00, the pattern e-01, and the pattern e-10 in FIG. 7 for the fractional pixel position e in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. In addition, identifiers corresponding to the filter patterns of the pattern e-00, the pattern e-01, and the pattern e-10 in FIG. 7 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern f-00, the pattern f-01, and the pattern f-10 in FIG. 8 for the fractional pixel position f in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, the identifiers corresponding to the filter patterns of the pattern f-00, the pattern f-01, and the pattern f-10 in FIG. 8 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern g-00, the pattern g-01, and the pattern g-10 in FIG. 9 for the fractional pixel position g in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the pattern g-00, the pattern g-01, and the pattern g-10 in FIG. 9 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern h-00, the pattern h-01, and the pattern h-10 in FIG. 10 for the fractional pixel position h in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern h-00, the pattern h-01, and the pattern h-10 in FIG. 10 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern i-00, the pattern i-01, and the pattern i-10 in FIG. 11 at the fractional pixel position i in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Further, identifiers corresponding to the filter patterns of the pattern i-00, the pattern i-01, and the pattern i-10 in FIG. 11 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern j-00, the pattern j-01, and the pattern j-10 in FIG. 12 for the fractional pixel position j in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern j-00, the pattern j-01, and the pattern j-10 in FIG. 12 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern k-00, the pattern k-01, and the pattern k-10 in FIG. 13 for the fractional pixel position k in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the pattern k-00, the pattern k-01, and the pattern k-10 in FIG. 13 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern n-00, the pattern n-01, and the pattern n-10 in FIG. 14 at the fractional pixel position n in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern n-00, the pattern n-01, and the pattern n-10 in FIG. 14 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern p-00, the pattern p-01, and the pattern p-10 in FIG. 15 for the fractional pixel position p in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Further, identifiers corresponding to the filter patterns of the pattern p-00, the pattern p-01, and the pattern p-10 in FIG. 15 are 00, 01, and 10, respectively.

  Further, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern q-00, the pattern q-01, and the pattern q-10 in FIG. 16 for the fractional pixel position q in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. Also, identifiers corresponding to the filter patterns of the patterns q-00, q-01, and pattern q-10 in FIG. 16 are 00, 01, and 10, respectively.

  In addition, the filter pattern information storage unit 201 stores three types of two-dimensional 6-tap filter patterns of the pattern r-00, the pattern r-01, and the pattern r-10 in FIG. 17 for the fractional pixel position r in FIG. I remember it. Here, 1 in the figure is a significant coefficient and 0 is an insignificant coefficient. The identifiers corresponding to the filter patterns of the pattern r-00, the pattern r-01, and the pattern r-10 in FIG. 17 are 00, 01, and 10, respectively.

  The shape of the filter pattern shown in FIGS. 3 to 17 (that is, the position of “1” (significant coefficient) in each pattern) has the following characteristics. In other words, the symmetry of the filter pattern in the horizontal, vertical, and diagonal directions matches the symmetry in the horizontal, vertical, and diagonal directions of the fractional pixel position in the rectangle surrounded by the neighboring integer pixels. ing. Here, “symmetry” means a characteristic that is symmetric or asymmetric in a corresponding direction.

  For example, the filter pattern for the fractional pixel position b shown in FIG. 4 is symmetric in the horizontal direction and asymmetrical in the vertical and diagonal directions. In FIG. 1, the fractional pixel position b is a neighboring integer pixel (for example, the integer pixel G). , M, N, and H) are symmetric in the horizontal direction and asymmetric in the vertical and diagonal directions. As described above, the horizontal, vertical, and diagonal symmetries of the filter pattern with respect to the fractional pixel position b coincide with the horizontal, vertical, and diagonal symmetries of the fractional pixel position b. .

  Further, the filter pattern for the fractional pixel position e shown in FIG. 7 is asymmetric in the horizontal and vertical directions and symmetrical in the lower right diagonal direction. In FIG. The quadrangle surrounded by the pixels G, M, N, and H) is asymmetric in the horizontal and vertical directions and symmetric in the lower right diagonal direction. As described above, the horizontal, vertical, and diagonal symmetry of the filter pattern for the fractional pixel position e coincides with the horizontal, vertical, and diagonal symmetry of the fractional pixel position e. .

  Also, the filter pattern for the fractional pixel position a shown in FIG. 3 is asymmetric in the horizontal, vertical, and diagonal directions. In FIG. 1, the fractional pixel position a is a neighboring integer pixel (for example, integer pixels G, M, N, H) are asymmetric in the horizontal, vertical and diagonal directions. Thus, the horizontal, vertical, and diagonal symmetries of the filter pattern with respect to the fractional pixel position a coincide with the horizontal, vertical, and diagonal symmetries of the fractional pixel position a. .

  As described above, the symmetry of the filter pattern in the horizontal direction, the vertical direction, and the diagonal direction is different from the symmetry of the horizontal, vertical direction, and diagonal direction of the fractional pixel position in the rectangle surrounded by the neighboring integer pixels. By matching, it is possible to generate an appropriate filter according to the fractional pixel position of each fractional pixel and perform encoding.

  In the filter pattern for each fractional pixel position, the number of “1” s (significant coefficients) is the same. For example, in the filter patterns j-00, j-01, and j-10 for the fractional pixel position j shown in FIG. 12, the number of “1” s (significant coefficients) is 8 and the same. Further, in each of the filter patterns shown in FIGS. 3 to 11 and FIGS. 13 to 17, the number of “1” (significant coefficient) is 6, which is the same. Thus, when one filter pattern is selected from a plurality of filter pattern candidates for a certain fractional pixel position, the number of “1” s (significant coefficients) in each filter pattern candidate is the same, so that selection is easy. Become.

  The filter pattern information decoding unit 2302 decodes a filter pattern code 2304 for each fractional pixel position between integer pixels input from outside in units of an image sequence to be encoded, and a frame image decoding unit 2303 as a filter pattern identifier 2305. Output to.

  The frame image decoding unit 2303 refers to the filter pattern information storage unit 201 for the filter pattern associated with the filter pattern identifier 2305 for each fractional pixel position between integer pixels input from the filter pattern information decoding unit 2302. Using the determined filter pattern, the frame image code 2306 input from the outside is decoded and output to the outside such as a display device (not specified).

  Specifically, the frame image decoding unit 2303 uses the frame image code 2306 as a filter coefficient of one part that is a significant coefficient in the filter pattern associated with the filter pattern identifier 2305 for each fractional pixel position between integer pixels. To generate a filter. Next, the frame image decoding unit 2303 creates a reference image with fractional pixel accuracy using a filter generated for each fractional pixel position between integer pixels, with respect to a reference image that has been decoded in the past, and performs motion compensation. Thus, the frame image to be decoded is decoded.

  For example, when 00 is input as the filter pattern identifier 2305 for the fractional pixel position a in FIG. 1, the frame image decoding unit 2303 refers to the filter pattern information storage unit 2301 to determine the pattern a-00 in FIG. . Next, the frame image decoding unit 2303 decodes the filter coefficient value of 1 part, which is a significant coefficient of the filter pattern a-00, with respect to the reference image encoded in the past. Thereby, the filter of the fractional pixel position a with respect to the encoding target frame image is generated. Further, when 01 or 10 is input as the filter pattern identifier 2305 for the fractional pixel position a in FIG. 1, the frame image decoding unit 2303 similarly performs the significant coefficient 1 portion of the pattern a-01 or the pattern a-10. The filter coefficient value of is decoded. In addition, for fractional pixel positions b, c, d, e, f, g, h, i, j, k, n, p, q, r in FIG. A filter coefficient is generated by decoding the filter coefficient value of 1 part which is a significant coefficient of the pattern.

  Next, the operation of the video decoding method according to this embodiment will be described with reference to FIG. FIG. 24 is a flowchart for explaining a moving picture decoding method which is an operation of the moving picture decoding apparatus 230 according to the present embodiment.

  First, the filter pattern information storage unit 2301 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, for each fractional pixel position between integer pixels in association with an identifier (step S2401).

  Next, the filter pattern information decoding unit 2302 decodes the filter pattern identifier for each fractional pixel position between integer pixels in units of an image sequence to be decoded from the filter pattern information code input from the outside (step S2402).

  Next, the frame image decoding unit 2303 refers to the filter pattern information storage unit 2301 for the filter pattern associated with the filter pattern identifier for each fractional pixel position between integer pixels input from the filter pattern information decoding unit 2302. Using the determined filter pattern, the frame image code 2306 input from the outside is decoded and output to the outside such as a display device (not specified) (step S2403).

  Next, step S2402, which is the operation of the filter pattern information decoding unit 2302, will be described in detail with reference to FIG. FIG. 25 is a flowchart for explaining step S2402, which is the operation of the filter pattern information decoding unit 2302 according to this embodiment.

  First, the filter pattern information decoding unit 2302 inputs a filter pattern code at a fractional pixel position between one integer pixel from the outside to the decoding target moving image (step S2501).

Next, the filter pattern information decoding unit 2302 decodes a filter pattern identifier, which is a filter pattern identifier, from the input filter pattern code. (Step S2502)
Next, the filter pattern information decoding unit 2302 determines whether the filter pattern identifier has been decoded at the fractional pixel positions between all the integer pixels (step S2503), and the filter pattern identifier at all the fractional pixel positions between the integer pixels. Is not decoded (NO), the process is repeated from step S2501 for the fractional pixel position between the next integer pixels. If the filter pattern identifier is decoded at the fractional pixel position between all the integer pixels (YES), the process ends.

  Next, step S2403, which is the operation of the frame image decoding unit 2303, will be described in detail with reference to FIG. FIG. 26 is a flowchart for explaining step S2403 which is the operation of the frame image decoding unit 2303 according to this embodiment.

  First, the frame image decoding unit 2303 inputs a frame image code from the outside (step S2601).

  Next, the frame image decoding unit 2303 discriminates the filter pattern by referring to the filter pattern information storage unit 201 from the filter pattern identifier at the fractional pixel position between one integer pixel (step S2602), and the discriminated filter. The filter coefficient of 1 part which is a significant coefficient in the pattern is decoded for each frame image to be decoded (step S2603), and a filter is determined (step S2604).

  Next, the frame image decoding unit 2303 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S2605), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), repeat from step S2602 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image decoding unit 2303 applies a filter for each determined fractional pixel position between integer pixels to a reference image that has been encoded in the past, thereby creating a reference image with fractional pixel accuracy (step S2606).

  Next, the frame image decoding unit 2303 uses the reference image with fractional pixel accuracy to decode and output the video sequence to be decoded from the frame image code by performing motion compensation (step S2607). finish.

  Next, a video decoding program 270 for causing a computer to function as the above-described video decoding device 230 will be described. FIG. 27 is a diagram showing the configuration of the moving picture decoding program 270.

  As shown in FIG. 27, the moving picture decoding program 270 includes a main module program 2701 that supervises processing, a filter pattern information storage module 2702, a filter pattern information decoding module 2703, and a frame image decoding module 2704. Is provided. The filter pattern information storage module 2702, the filter pattern information decoding module 2703, and the frame image decoding module 2704 cause the computer to perform functions corresponding to the above-described filter pattern information storage unit 2301 and filter pattern information decoding, respectively. The same as the unit 2302 and the frame image decoding unit 2303.

  According to the moving image encoding apparatus, the moving image encoding method, the moving image encoding program, the moving image decoding apparatus, the moving image decoding method, and the moving image decoding program as described above, the fractional pixel accuracy using the filter When the reference image is generated and motion compensated for encoding and decoding, the filter pattern of the filter coefficient is encoded for each moving image sequence to be encoded, and the significant coefficient of the filter pattern is encoded for each frame image to be encoded. Since it is only necessary to encode the filter coefficient, it is possible to perform motion compensation with high accuracy by using a filter suitable for the feature of the video while reducing the code amount of the filter coefficient.

  The filter pattern stored in the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 for each fractional pixel position between integer pixels is not limited to the illustrated filter pattern, and other patterns may be used. Further, the number of taps of the filter pattern is not limited to the two-dimensional 6 × 6 taps, and other tap numbers may be used. Furthermore, the values of the significant coefficient and the insignificant coefficient of the filter pattern are not limited to 1 and 0, respectively, and other values may be used. The filter pattern stored by the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 for each fractional pixel position between integer pixels can take any binary pattern that is a combination of a two-dimensional significant coefficient and an insignificant coefficient.

  In addition, the number of filter patterns that the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 store for each fractional pixel position between integer pixels is not limited to three as illustrated, but may be two or more other numbers. good.

  Further, the identifiers of the filter patterns that the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 store for each fractional pixel position between integer pixels are not limited to 00, 01, and 10 illustrated, but may be other identifiers. .

  In addition, the filter pattern selection unit 202 does not select a filter pattern for each fractional pixel position between integer pixels in units of a moving image sequence to be encoded, but in units of frame images in which a plurality of frame images to be encoded are collected. A filter pattern may be selected for each fractional pixel position between integer pixels. In this case, the filter pattern information encoding unit 203 encodes the filter pattern code 207 for each frame image group in which a plurality of encoding target frame images are collected. Also, the filter pattern information decoding unit 2302 decodes the filter pattern code 2304 in units of frame image groups in which a plurality of decoding target frame images are collected. Thus, when generating a fractional pixel precision reference image using a filter and performing motion compensation for encoding and decoding, a filter pattern of filter coefficients in units of frame images in which a plurality of frame images to be encoded are grouped together Since only the filter coefficient of the significant coefficient of the filter pattern is encoded and decoded in units of frame images to be encoded, a frame group of a plurality of video frame images combined while reducing the code amount of the filter coefficient. Motion compensation can be performed with high accuracy using a filter suitable for the feature.

  Also, the filter pattern selection unit 202 may select a filter pattern for each frame image to be encoded, instead of selecting a filter pattern for each fractional pixel position between integer pixels in units of a moving image sequence to be encoded. good. In this case, the filter pattern information encoding unit 203 encodes the filter pattern code 207 for each frame image to be encoded. The filter pattern information decoding unit 2302 decodes the filter pattern code 2304 for each frame image to be decoded. As a result, when a reference image with fractional pixel accuracy is generated using a filter, and motion compensation is performed for encoding and decoding, the filter pattern of the filter coefficient is encoded and encoded for each frame image to be encoded. Since only the filter coefficient of the significant coefficient of the filter pattern is encoded and decoded for each frame image, motion compensation is performed with high accuracy using a filter that matches the characteristics of the video frame image while reducing the code amount of the filter coefficient. can do.

  The frame image encoding unit 204 does not generate a filter coefficient of a significant coefficient of the filter pattern for each frame image to be encoded, but for each slice in which a plurality of blocks obtained by dividing the frame image to be encoded are collected. Filter coefficients may be generated. At this time, the frame image decoding unit 2303 does not decode the filter coefficient of the significant coefficient of the filter pattern for each frame image to be decoded, but a slice in which a plurality of blocks obtained by dividing the frame image to be decoded are collected. The filter coefficients may be decoded every time. Thus, when generating a fractional pixel precision reference image using a filter and performing motion compensation and encoding, the filter coefficient is determined for each slice and encoded and decoded. It is possible to perform motion compensation with high accuracy using a filter suitable for.

  In addition, the filter patterns stored in the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 and their identifiers are not associated with each fractional pixel position between integer pixels, but are a group of fractional pixel positions between integer pixels. You may match every. For example, positions a, c, d, n in FIG. 1, positions b, h in FIG. 1, positions e, g, p, r in FIG. 1, positions f, i, k, q in FIG. The positions j in FIG. 1 may be grouped together as a group (group 1, group 2, group 3, group 4, group 5), and a filter pattern and its identifier may be stored in association with each group. Thus, when a fractional pixel precision reference image is generated using a filter and motion compensation is performed for encoding and decoding, only for each group in which fractional pixel positions between integer pixels of a frame image to be encoded are collected. In addition, it is only necessary to encode and decode the filter pattern identifier of the filter coefficient, so that it is possible to perform motion compensation with high accuracy using a filter that matches the characteristics of the video frame image while reducing the code amount of the filter coefficient. it can.

  In addition, the filter patterns stored in the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 and their identifiers do not associate an identifier for each fractional pixel position between integer pixels, but instead of a fractional pixel position between integer pixels and its identifier. An identifier may be associated with each combination of filter patterns. For example, a combination identifier that uses pattern 00 at all fractional pixel positions between integer pixels is 100, a combination identifier that uses pattern 01 at all fractional pixel positions between integer pixels is 101, and a pattern at all fractional pixel positions between integer pixels The identifiers of combinations using 10 may be collectively stored as 110. As a result, when a reference image with fractional pixel accuracy is generated using a filter and motion compensation is performed for encoding and decoding, the fractional pixel position between integer pixels of the frame image to be encoded and its filter pattern are matched. Therefore, it is only necessary to encode and decode the identifiers of the combinations, so that motion compensation can be performed with high accuracy using a filter that matches the characteristics of the video frame image while reducing the code amount of the filter coefficient.

  In addition, the filter patterns stored in the filter pattern information storage unit 201 and the filter pattern information storage unit 2301 and their identifiers are not associated with each fractional pixel position between integer pixels, but are combined with fractional pixel positions between integer pixels. An identifier may be associated with each combination of the group and its filter pattern. For example, the combination identifier using pattern 00 at all fractional pixel positions in group 1 is 00000, the combination identifier using pattern 01 is at all fractional pixel positions in group 1, 00001, and the pattern is at all fractional pixel positions in group 1 The identifiers of combinations using 10 may be stored together as 00010. Also, the identifier of the combination using the pattern 00 at all fractional pixel positions in the group 2 is 00100, the identifier of the combination using the pattern 01 is at all fractional pixel positions in the group 2, 00101, and the pattern is at the all fractional pixel positions in the group 2. The identifiers of combinations using 10 may be stored together as 00110. Also, the combination identifier using pattern 00 at all fractional pixel positions in group 3 is 01000, the combination identifier using pattern 01 at all fractional pixel positions in group 3 is 01001, and the pattern at all fractional pixel positions in group 3 is pattern. The identifiers of combinations using 10 may be collectively stored as 01010. Also, the identifier of the combination using the pattern 00 at all the fractional pixel positions in the group 4 is 01100, the identifier of the combination using the pattern 01 at the all fractional pixel positions in the group 4 is 01101, and the pattern is at the all fractional pixel positions in the group 4 The identifiers of combinations using 10 may be collectively stored as 01110. Further, the identifier of the combination that uses the pattern 00 at all the fractional pixel positions in the group 5 is 10000, the identifier of the combination that uses the pattern 01 at the all fractional pixel positions in the group 5 is 10001, and the pattern at the all fractional pixel positions in the group 5 The identifiers of combinations using 10 may be stored together as 10010. Thus, when a fractional pixel precision reference image is generated using a filter and motion compensated for encoding and decoding, a group of fractional pixel positions between integer pixels of a frame image to be encoded and its group Since identifiers need only be encoded and decoded for each combination of filter patterns, motion compensation can be performed with high accuracy using a filter that matches the characteristics of the video frame image while reducing the amount of code of the filter coefficient. be able to.

  Further, the filter pattern code output from the filter pattern information encoding unit and the frame image code output from the frame image encoding unit may be output to different transmission paths. Also, the filter pattern code input by the filter pattern information decoding unit and the frame image code input by the frame image decoding unit may be input from different transmission paths. Accordingly, the filter pattern information and the frame image information can be transmitted with different degrees of importance.

(Modification 1 of Embodiment 1)
Regarding a video encoding device, a video encoding method, a video encoding program, a video decoding device, a video decoding method, and a video decoding program according to the first modification of the first embodiment of the present invention This will be described with reference to the drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same element and the overlapping description is abbreviate | omitted.

  FIG. 28 is a block diagram showing a configuration of a video encoding device 280 according to the first modification of the present embodiment. The moving image encoding device 280 includes, as its functional components, a filter pattern information storage unit (filter pattern information storage unit) 2801, a filter pattern selection unit (filter pattern selection unit) 2802, and a filter pattern information encoding unit ( A filter pattern information encoding unit) 2803 and a frame image encoding unit (frame image encoding unit) 2804. Note that the filter pattern information storage unit (filter pattern information storage unit) 2801 is the same as the filter pattern information storage unit (filter pattern information storage unit) 201 in the first embodiment of the present invention, and a description thereof will be omitted.

  The filter pattern selection unit 2802 stores, in the filter pattern information storage unit 2801, for each fractional pixel position between integer pixels in an encoding target image sequence unit with respect to an encoding target moving image sequence 2805. One filter pattern is selected from the filtered filter patterns, and a filter pattern identifier 2806 that is an identifier of the selected filter pattern is output to the filter pattern information encoding unit 2803. Further, the filter pattern selection unit 2802 outputs the filter pattern 2807 selected for each fractional pixel position between integer pixels in the image sequence unit to be encoded to the frame image encoding unit 204.

  The filter pattern information encoding unit 2803 encodes the filter pattern identifier 2806 for each fractional pixel position between integer pixels input from the filter pattern selection unit 2802 for each image sequence to be encoded, and outputs the filter pattern code 2808 to the outside. Output.

  The frame image encoding unit 2804 encodes an encoding target moving image sequence 2805 input from the outside using a filter pattern 2807 for each fractional pixel position between integer pixels input from the filter pattern selection unit 2802. . The frame image encoding unit 2804 outputs the encoded frame image code 2809 to the outside.

  Specifically, the frame image encoding unit 2804 generates, for each frame image to be encoded, one portion of the filter coefficient that is a significant coefficient in the filter pattern code 2808 for each fractional pixel position between integer pixels. At this time, when the frame image encoding unit 2804 creates a reference image with fractional pixel accuracy by filtering a reference image encoded in the past, an error from the signal of the frame image to be encoded is minimized. The filter coefficient of the significant coefficient 1 part in the filter pattern is generated.

  The frame image encoding unit 2804 creates a reference image with fractional pixel accuracy using a filter coefficient generated for each fractional pixel position between integer pixels with respect to the reference image, and performs motion compensation to generate a frame image to be encoded. Encode. Also, at this time, among the filter coefficients for each fractional pixel position between the integer pixels generated for each frame image, the filter coefficient value of the 1 part which is a significant coefficient is also encoded, and the frame image code 2809 is output to the outside. Output.

  Note that a method in which the frame image encoding unit 2804 generates a filter coefficient of one portion, which is a significant coefficient in the filter pattern 2807 for each fractional pixel position between integer pixels, for each frame image to be encoded is encoded in the past. The method is not limited to a method in which an error with a signal of a frame image to be encoded is minimized when a reference image with a fractional pixel accuracy is generated by filtering the converted reference image, and other methods may be used.

  Next, the operation of the moving picture coding method according to the first modification of the present embodiment will be described with reference to FIG. FIG. 29 is a flowchart illustrating a video encoding method that is an operation of the video encoding device 280 according to the first modification of the present embodiment.

  First, the filter pattern information storage unit 2801 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, for each fractional pixel position between integer pixels in association with an identifier (step S2901).

  Next, with respect to the moving image sequence 2805 to be encoded that is input from the outside by the filter pattern selection unit 2802, the filter pattern information storage unit 2801 stores, for each fractional pixel position between integer pixels, in units of the encoding target image sequence. One filter pattern is selected from the stored filter patterns, and its identifier is discriminated (step S2902).

  Next, the filter pattern information encoding unit 2803 encodes the identifier of the filter pattern for each fractional pixel position between the integer pixels selected by the filter pattern selection unit 2802 for each image sequence to be encoded, and outputs the encoded image to the outside ( Step S2903).

  In addition, the frame image encoding unit 2804 encodes a moving image sequence 2805 to be encoded input from the outside using a filter pattern for each fractional pixel position between integer pixels selected by the filter pattern selection unit 2802, Output to the outside (step S2904).

  Next, step S2902 as the operation of the filter pattern selection unit 2802 will be described in detail with reference to FIG. FIG. 30 is a flowchart illustrating step S2902 that is the operation of the filter pattern selection unit 2802 according to the first modification of the present embodiment.

  First, the filter pattern selection unit 2802 inputs a moving image sequence 2805 to be encoded from the outside (step S3001).

  Next, the filter pattern selection unit 2802 selects one filter pattern from the filter patterns stored in the filter pattern information storage unit 2801 at one fractional pixel position between integer pixels (step S3002), and the identifier of the selected filter pattern Is output to the filter pattern information encoding unit 2803 (step S3003). Further, the filter pattern selection unit 2802 outputs the selected filter pattern to the frame image encoding unit 2804 (step S3004).

  Next, the filter pattern selection unit 2802 determines whether or not a filter pattern has been selected at fractional pixel positions between all integer pixels (step S3005), and has not selected a filter pattern at all fractional pixel positions between integer pixels. If yes (NO), repeat from step S3002 for the fractional pixel position between the next integer pixels. If a filter pattern is selected at a fractional pixel position between all integer pixels (YES), the process ends.

  Next, step S2903, which is the operation of the filter pattern information encoding unit 2803, will be described in detail with reference to FIG. FIG. 31 is a flowchart illustrating step S2903, which is the operation of the filter pattern selection unit 2802 according to the first modification of the present embodiment.

  First, the filter pattern information encoding unit 2803 encodes an identifier of a filter pattern at a fractional pixel position between one integer pixel with respect to a moving image to be encoded (step S3101) and outputs it (step S3102).

  Next, the filter pattern information encoding unit 2803 determines whether the filter pattern identifier has been encoded at the fractional pixel positions between all the integer pixels (step S3103), and the filter pattern at all the fractional pixel positions between the integer pixels. If the identifier is not encoded (NO), the process is repeated from step S3101 for the fractional pixel position between the next integer pixels. If the filter pattern identifier is encoded at the fractional pixel position between all integer pixels (YES), the process ends.

  Next, step S2904, which is the operation of the frame image encoding unit 2804, will be described in detail with reference to FIG. FIG. 32 is a flowchart illustrating step S2904, which is the operation of the frame image encoding unit 2804 according to the first modification of the present embodiment.

  First, the frame image encoding unit 2804 generates a filter coefficient of one part, which is a significant coefficient in the filter pattern at a fractional pixel position between one integer pixel, for each frame image to be encoded, and determines a filter. (Step S3201). In addition, among the determined filters, the filter coefficient of 1 part which is a significant coefficient in the filter pattern is encoded (step S3202).

  Next, the frame image encoding unit 2804 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S3203), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), the process is repeated from step S3201 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image encoding unit 2804 creates a reference image with fractional pixel accuracy by applying the filter for each fractional pixel position between integer pixels thus determined to a reference image that has been encoded in the past (step S3204).

  Next, the frame image encoding unit 2804 encodes the moving image sequence 2805 to be encoded by performing motion compensation using the reference image with fractional pixel accuracy (step S3205), and matches the filter coefficient with the encoded code. Output to the outside (step S3206), and the process ends.

  Next, a video encoding program 330 for causing a computer to function as the above-described video encoding device 280 will be described. FIG. 33 is a diagram showing a configuration of the moving image encoding program 330.

  As shown in FIG. 33, the moving image encoding program 330 includes a main module program 3301 that supervises processing, a filter pattern information storage module 3302, a filter pattern selection module 3303, a filter pattern information encoding module 3304, A frame image encoding module 3305. The filter pattern information storage module 3302, the filter pattern selection module 3303, the filter pattern information encoding module 3304, and the frame image encoding module 3305 have functions corresponding to the above-described filter pattern information storage units 2801 respectively. These are the same as the filter pattern selection unit 2802, the filter pattern information encoding unit 2803, and the frame image encoding unit 2804.

  FIG. 34 is a block diagram illustrating a configuration of a video decoding device 340 according to the first modification of the present embodiment. The moving picture decoding apparatus 340 includes, as its functional components, a filter pattern information storage unit (filter pattern storage unit) 3401, a filter pattern information decoding unit (filter pattern information decoding unit) 3402, and frame image decoding. And a conversion unit (frame image decoding means) 3403. Since the filter pattern information storage unit (filter pattern information storage unit) 3401 is the same as the filter pattern information storage unit (filter pattern information storage unit) 2301 in the first embodiment of the present invention, the description thereof is omitted.

  The filter pattern information decoding unit 3402 decodes a filter pattern code 3404 for each fractional pixel position between integer pixels input from the outside in units of an image sequence to be encoded, and acquires a filter pattern identifier. The filter pattern information decoding unit 3402 refers to the filter pattern information storage unit 3401, determines the filter pattern associated with the filter pattern identifier for each fractional pixel position between integer pixels, and decodes the frame image as the filter pattern 3405. Output to the unit 3403.

  The frame image decoding unit 3403 uses the filter pattern 3405 for each fractional pixel position between integer pixels input from the filter pattern information decoding unit 3402 to decode the frame image code 3406 input from the outside, and displays the display device Output to the outside (not specified).

  Specifically, the frame image decoding unit 3403 decodes the filter coefficient of 1 part, which is a significant coefficient in the filter pattern 3405 for each fractional pixel position between integer pixels, from the frame image code 3406 to generate a filter. To do. Next, the frame image decoding unit 3403 creates a reference image with fractional pixel accuracy using a filter generated for each fractional pixel position between integer pixels, with respect to a reference image that has been decoded in the past, and performs motion compensation. Thus, the frame image to be decoded is decoded.

  Next, the operation of the moving picture decoding method according to the first modification of the present embodiment will be described using FIG. FIG. 35 is a flowchart for explaining a video decoding method that is an operation of the video decoding device 340 according to the first modification of the present embodiment.

  First, the filter pattern information storage unit 3401 stores a plurality of filter patterns, which are patterns of significant coefficients of a filter of a two-dimensional matrix for each fractional pixel position between integer pixels, in association with an identifier (step S3501).

  Next, the filter pattern information decoding unit 3402 decodes the filter pattern identifier for each fractional pixel position between integer pixels in units of the image sequence to be decoded from the filter pattern information code input from the outside, and the filter pattern A filter pattern associated with the identifier of the filter pattern is acquired with reference to the information storage unit 3401 (step S3502).

  Next, the frame image decoding unit 3403 uses the filter pattern for each fractional pixel position between integer pixels input from the filter pattern information decoding unit 3402 to decode the frame image code 3406 input from the outside, Output to the outside such as a display device (not specified) (step S3503).

  Next, step S3502 which is the operation of the filter pattern information decoding unit 3402 will be described in detail with reference to FIG. FIG. 36 is a flowchart for explaining step S3502 which is the operation of the filter pattern information decoding unit 3402 according to the first modification of the present embodiment.

  First, the filter pattern information decoding unit 3402 inputs a filter pattern code at a fractional pixel position between one integer pixel from the outside to the decoding target moving image (step S3601).

Next, the filter pattern information decoding unit 3402 decodes a filter pattern identifier, which is a filter pattern identifier, from the input filter pattern code. (Step S3602)
Next, the filter pattern information decoding unit 3402 refers to the filter pattern information storage unit 3401 and acquires a filter pattern associated with the identifier of the decoded filter pattern. (Step S3603)
Next, the filter pattern information decoding unit 3402 determines whether or not a filter pattern has been acquired at fractional pixel positions between all integer pixels (step S3604), and acquires a filter pattern at all fractional pixel positions between integer pixels. If not (NO), the process repeats from step S3601 for the fractional pixel position between the next integer pixels. If the filter pattern is acquired at the fractional pixel position between all the integer pixels (YES), the process ends.

  Next, step S3503, which is the operation of the frame image decoding unit 3403, will be described in detail with reference to FIG. FIG. 37 is a flowchart for explaining step S3503 which is the operation of the frame image decoding unit 3403 according to the first modification of the present embodiment.

  First, the frame image decoding unit 3403 inputs a frame image code from the outside (step S3701).

  Next, the frame image decoding unit 3403 decodes the filter coefficient of one part, which is a significant coefficient in the filter pattern at the fractional pixel position between one integer pixel, for each frame image to be decoded (step S3702). The filter is determined (step S3703).

  Next, the frame image decoding unit 3403 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S3704), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), the process is repeated from step S3702 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image decoding unit 3403 applies a filter for each fractional pixel position between the determined integer pixels to a reference image that has been encoded in the past, and creates a reference image with fractional pixel accuracy (step S3705).

  Next, the frame image decoding unit 3403 performs motion compensation on the moving image sequence to be decoded from the frame image code using the reference image with fractional pixel accuracy, and outputs the decoded image sequence (step S3706). finish.

  Next, a video decoding program 380 for causing a computer to function as the above-described video decoding device 340 will be described. FIG. 38 is a diagram showing the configuration of the video decoding program 380.

  As shown in FIG. 38, the moving picture decoding program 380 includes a main module program 3801 that supervises processing, a filter pattern information storage module 3802, a filter pattern information decoding module 3803, and a frame image decoding module 3804. Is provided. The filter pattern information storage module 3802, the filter pattern information decoding module 3803, and the frame image decoding module 3804 allow the computer to perform functions corresponding to the above-described filter pattern information storage unit 3401 and filter pattern information decoding, respectively. The same as the unit 3402 and the frame image decoding unit 3403.

(Modification 2 of Embodiment 1)
Regarding a video encoding device, a video encoding method, a video encoding program, a video decoding device, a video decoding method, and a video decoding program according to Modification 2 of the first embodiment of the present invention This will be described with reference to the drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same element and the overlapping description is abbreviate | omitted.

  FIG. 39 is a block diagram showing a configuration of a video encoding device 390 according to the second modification of the present embodiment. The moving image encoding apparatus 390 includes, as functional components, a filter pattern information storage unit (filter pattern information storage unit) 3901, a filter pattern selection unit (filter pattern selection unit) 3902, and a filter pattern information encoding unit ( A filter pattern information encoding unit) 3903 and a frame image encoding unit (frame image encoding unit) 3904. Note that the filter pattern information storage unit (filter pattern information storage unit) 3901 is the same as the filter pattern information storage unit (filter pattern information storage unit) 201 in the first embodiment of the present invention, and a description thereof will be omitted.

  The filter pattern selection unit 3902 stores the encoding target video sequence 3905 input from the outside in the filter pattern information storage unit 3901 for each fractional pixel position between integer pixels in units of the encoding target image sequence. One filter pattern is selected from the filtered filter patterns, and the selected filter pattern 3906 is output to the filter pattern information encoding unit 3903 and the frame image encoding unit 3904.

  The filter pattern information encoding unit 3903 encodes the filter pattern 3906 for each fractional pixel position between integer pixels input from the filter pattern selection unit 3902 in units of image sequences to be encoded, and outputs the result as a filter pattern code 3907 to the outside. To do.

  Specifically, the filter pattern information encoding unit 3903 receives a two-dimensional binary pattern of a significant coefficient and a non-significant coefficient of the filter pattern 3906 for each fractional pixel position between integer pixels input from the filter pattern selection unit 3902. The data is rearranged into a one-dimensional binary data string, encoded as information about the filter pattern in units of an image sequence to be encoded, and output to the outside as a filter pattern code 3907.

  The frame image encoding unit 3904 encodes an encoding target moving image sequence 3905 input from the outside using a filter pattern 3906 for each fractional pixel position between integer pixels input from the filter pattern selection unit 3902. . The frame image encoding unit 3904 outputs the encoded frame image code 3908 to the outside.

  Specifically, the frame image encoding unit 3904 generates, for each frame image to be encoded, one portion of the filter coefficient that is a significant coefficient in the filter pattern code 3907 for each fractional pixel position between integer pixels. At this time, when the frame image encoding unit 3904 creates a reference image with fractional pixel accuracy by filtering a reference image encoded in the past, an error from the signal of the frame image to be encoded is minimized. The filter coefficient of the significant coefficient 1 part in the filter pattern is generated.

  The frame image encoding unit 3904 creates a reference image with fractional pixel accuracy using a filter coefficient generated for each fractional pixel position between integer pixels with respect to the reference image, and performs motion compensation to generate a frame image to be encoded. Encode. Also, at this time, among the filter coefficients for each fractional pixel position between the integer pixels generated for each frame image, the filter coefficient value of the 1 part which is a significant coefficient is also encoded, and the frame image code 3908 is output to the outside. Output.

  Note that a method in which the frame image encoding unit 3904 generates a filter coefficient of one part, which is a significant coefficient in the filter pattern 3906 for each fractional pixel position between integer pixels, for each frame image to be encoded is encoded in the past. The method is not limited to a method in which an error with a signal of a frame image to be encoded is minimized when a reference image with a fractional pixel accuracy is generated by filtering the converted reference image, and other methods may be used.

  Next, the operation of the moving picture coding method according to the second modification of the present embodiment will be described with reference to FIG. FIG. 40 is a flowchart illustrating a video encoding method that is an operation of the video encoding device 390 according to the second modification of the present embodiment.

  First, the filter pattern information storage unit 3901 stores a plurality of filter patterns, which are patterns of significant coefficients of a filter of a two-dimensional matrix for each fractional pixel position between integer pixels, in association with an identifier (step S4001).

  Next, with respect to the moving image sequence 3905 to be encoded that is input from the outside by the filter pattern selection unit 3902, the filter pattern information is stored in the filter pattern information storage unit 3901 for each fractional pixel position between integer pixels in units of the encoding target image sequence. One filter pattern is selected from the stored filter patterns (step S4002).

  Next, the filter pattern information encoding unit 3903 encodes the filter pattern for each fractional pixel position between the integer pixels selected by the filter pattern selection unit 3902 for each image sequence to be encoded, and outputs the encoded image to the outside (step S4003). ).

  Further, the frame image encoding unit 3904 encodes a moving image sequence 3905 to be encoded input from the outside using a filter pattern for each fractional pixel position between integer pixels selected by the filter pattern selection unit 3902, Output to the outside (step S4004).

  Next, step S4002 which is the operation of the filter pattern selection unit 3902 will be described in detail with reference to FIG. FIG. 41 is a flowchart illustrating step S4002 which is the operation of the filter pattern selection unit 3902 according to the second modification of the present embodiment.

  First, the filter pattern selection unit 3902 inputs a moving image sequence 3905 to be encoded from the outside (step S4101).

  Next, the filter pattern selection unit 3902 selects one filter pattern from the filter patterns stored in the filter pattern information storage unit 3901 at one fractional pixel position between integer pixels (step S4102), and filters the selected filter pattern. The data is output to the pattern information encoding unit 3903 and the frame image encoding unit 3904 (step S4103).

  Next, the filter pattern selection unit 3902 determines whether or not a filter pattern is selected at a fractional pixel position between all integer pixels (step S4104), and does not select a filter pattern at a fractional pixel position between all integer pixels. If yes (NO), repeat from step S4102 for the fractional pixel position between the next integer pixels. If a filter pattern is selected at a fractional pixel position between all integer pixels (YES), the process ends.

  Next, step S4003 that is the operation of the filter pattern information encoding unit 3903 will be described in detail with reference to FIG. FIG. 42 is a flowchart for explaining step S4003 which is the operation of the filter pattern selection unit 3902 according to the second modification of the present embodiment.

  First, the filter pattern information encoding unit 3903 encodes a filter pattern at a fractional pixel position between one integer pixel for a moving image to be encoded (step S4201), and outputs the code (step S4202). .

  Next, the filter pattern information encoding unit 3903 determines whether or not the filter pattern is encoded at the fractional pixel positions between all integer pixels (step S4203), and encodes the filter pattern at the fractional pixel positions between all integer pixels. If not (NO), the process repeats from step S4201 for the fractional pixel position between the next integer pixels. If the filter pattern is encoded at the fractional pixel positions between all the integer pixels (YES), the process ends.

  Next, step S4004 that is the operation of the frame image encoding unit 3904 will be described in detail with reference to FIG. FIG. 43 is a flowchart for explaining step S4004 which is the operation of the frame image encoding unit 3904 according to the second modification of the present embodiment.

  First, the frame image encoding unit 3904 generates a filter coefficient of 1 part, which is a significant coefficient in the filter pattern at a fractional pixel position between one integer pixel, for each frame image to be encoded, and determines a filter. (Step S4301). In addition, among the determined filters, the filter coefficient of 1 part which is a significant coefficient in the filter pattern is encoded (step S4302).

  Next, the frame image encoding unit 3904 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S4303), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), the process is repeated from step S4301 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image encoding unit 3904 creates a reference image with fractional pixel accuracy by applying the determined filter for each fractional pixel position between integer pixels to a reference image that has been encoded in the past (step S4304).

  Next, the frame image encoding unit 3904 encodes the video sequence 3905 to be encoded with motion compensation using the reference image with fractional pixel accuracy (step S4305), and matches the encoded filter coefficient with the encoded code. To the outside (step S4306), and the process is terminated.

  Next, a video encoding program 440 for causing a computer to function as the above-described video encoding device 390 will be described. FIG. 44 is a diagram showing a configuration of the moving image encoding program 440.

  As shown in FIG. 44, the moving image encoding program 440 includes a main module program 4401 that supervises processing, a filter pattern information storage module 4402, a filter pattern selection module 4403, a filter pattern information encoding module 4404, A frame image encoding module 4405. The filter pattern information storage module 4402, the filter pattern selection module 4403, the filter pattern information encoding module 4404, and the frame image encoding module 4405 have functions corresponding to the above-described filter pattern information storage units 3901 respectively. The same as the filter pattern selection unit 3902, the filter pattern information encoding unit 3903, and the frame image encoding unit 3904.

  Note that the filter pattern stored in the filter pattern information storage unit 3901 does not necessarily have to be associated with the identifier.

  FIG. 45 is a block diagram illustrating a configuration of a video decoding device 450 according to the second modification of the present embodiment. The moving image decoding apparatus 450 includes a filter pattern information decoding unit (filter pattern information decoding unit) 4501 and a frame image decoding unit (frame image decoding unit) 4502 as functional components. Composed.

  The filter pattern information decoding unit 4501 decodes the filter pattern code 4503 for each fractional pixel position between integer pixels input from the outside in units of an image sequence to be encoded, and acquires a filter pattern. The filter pattern information decoding unit 4501 outputs the filter pattern decoded for each fractional pixel position between integer pixels to the frame image decoding unit 4502 as a filter pattern 4504.

  Specifically, the filter pattern information decoding unit 4501 decodes a one-dimensional binary data sequence from the filter pattern code 4503 for each fractional pixel position between integer pixels input from the outside, and converts them into significant coefficients and non-significant coefficients. The result is rearranged into a two-dimensional binary pattern of significant coefficients and output as a filter pattern 4504 to the frame image decoding unit 4502.

  The frame image decoding unit 4502 decodes the frame image code 4505 input from the outside using the filter pattern 4504 for each fractional pixel position between integer pixels input from the filter pattern information decoding unit 4501, and displays the display device Output to the outside (not specified).

  Specifically, the frame image decoding unit 4502 decodes the filter coefficient of one part, which is a significant coefficient in the filter pattern 4504 for each fractional pixel position between integer pixels, from the frame image code 4505 to generate a filter. To do. Next, the frame image decoding unit 4502 creates a reference image with fractional pixel accuracy for a reference image that has been decoded in the past, using a filter that is generated for each fractional pixel position between integer pixels, and performs motion compensation. Thus, the frame image to be decoded is decoded.

  Next, the operation of the video decoding method according to the second modification of the present embodiment will be described using FIG. FIG. 46 is a flowchart for explaining a video decoding method that is an operation of the video decoding device 450 according to the second modification of the present embodiment.

  First, the filter pattern information decoding unit 4501 decodes a filter pattern for each fractional pixel position between integer pixels in units of an image sequence to be decoded from an externally input filter pattern information code, and acquires a filter pattern. (Step S4601).

  Next, the frame image decoding unit 4502 decodes the frame image code 4505 input from the outside using the filter pattern for each fractional pixel position between the integer pixels input from the filter pattern information decoding unit 4501, Output to the outside such as a display device (not specified) (step S4602).

  Next, step S4601 which is the operation of the filter pattern information decoding unit 4501 will be described in detail with reference to FIG. FIG. 47 is a flowchart for explaining step S4601, which is the operation of the filter pattern information decoding unit 4501 according to the second modification of the present embodiment.

  First, the filter pattern information decoding unit 4501 inputs a filter pattern code at a fractional pixel position between one integer pixel from the outside to the decoding target moving image (step S4701).

Next, the filter pattern information decoding unit 4501 decodes the filter pattern from the input filter pattern code, and acquires the filter pattern. (Step S4702)
Next, the filter pattern information decoding unit 4501 determines whether the filter pattern is decoded at the fractional pixel positions between all integer pixels (step S4703), and decodes the filter pattern at the fractional pixel positions between all integer pixels. If not (NO), the process repeats from step S4701 for the fractional pixel position between the next integer pixels. When the filter pattern is decoded at the fractional pixel positions between all the integer pixels (YES), the process is terminated.

  Next, step S4602 which is the operation of the frame image decoding unit 4502 will be described in detail with reference to FIG. FIG. 48 is a flowchart for explaining step S4602, which is the operation of the frame image decoding unit 4502 according to the second modification of the present embodiment.

  First, the frame image decoding unit 4502 inputs a frame image code from the outside (step S4801).

  Next, the frame image decoding unit 4502 decodes the filter coefficient of one part, which is a significant coefficient in the filter pattern at the fractional pixel position between one integer pixel, for each frame image to be decoded (step S4802). The filter is determined (step S4803).

  Next, the frame image decoding unit 4502 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S4804), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), the process is repeated from step S4802 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image decoding unit 4502 creates a reference image with fractional pixel accuracy by applying a filter for each fractional pixel position between integer pixels thus determined to a reference image that has been encoded in the past (step S4805).

  Next, the frame image decoding unit 4502 uses the reference image with fractional pixel accuracy to decode and output the moving image sequence to be decoded from the frame image code by performing motion compensation (step S4806). finish.

  Next, a video decoding program 490 for causing a computer to function as the above-described video decoding device 450 will be described. FIG. 49 is a diagram showing a configuration of the moving picture decoding program 490.

  As shown in FIG. 49, the moving picture decoding program 490 includes a main module program 4901 that supervises processing, a filter pattern information decoding module 4902, and a frame image decoding module 4903. The functions that the filter pattern information decoding module 4902 and the frame image decoding module 4903 perform on the computer are the same as those of the corresponding filter pattern information decoding unit 4501 and frame image decoding unit 4502, respectively.

  According to the moving image encoding apparatus, the moving image encoding method, the moving image encoding program, the moving image decoding apparatus, the moving image decoding method, and the moving image decoding program as described above, the fractional pixel accuracy using the filter When the reference image is generated and motion compensated for encoding and decoding, the filter pattern of the filter coefficient is encoded for each moving image sequence to be encoded, and the significant coefficient of the filter pattern is encoded for each frame image to be encoded. Since it is only necessary to encode the filter coefficient, it is possible to perform motion compensation with high accuracy by using a filter suitable for the feature of the video while reducing the code amount of the filter coefficient. Further, since the filter pattern selection unit directly encodes the filter pattern and the filter pattern information decoding unit directly decodes the frame pattern, it is not necessary for the frame image decoding unit to include a filter pattern information storage unit. , A moving picture decoding method, and a moving picture decoding program can be simplified.

(Modification 3 of Embodiment 1)
Regarding a video encoding device, a video encoding method, a video encoding program, a video decoding device, a video decoding method, and a video decoding program according to Modification 3 of the first embodiment of the present invention This will be described with reference to the drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same element and the overlapping description is abbreviate | omitted.

  FIG. 50 is a block diagram illustrating a configuration of a moving image encoding apparatus 500 according to the third modification of the present embodiment. The moving image encoding apparatus 500 includes, as functional components, a filter pattern information storage unit (filter pattern information storage unit) 5001, a filter pattern selection unit (filter pattern selection unit) 5002, and a filter pattern information encoding unit ( Filter pattern information encoding unit) 5003, a frame image encoding unit (frame image encoding unit) 5004, and a code synthesis unit (code synthesis unit) 5005 are configured. The filter pattern information storage unit (filter pattern information storage unit) 5001 and the filter pattern selection unit (filter pattern selection unit) 5002 are respectively a filter pattern information storage unit (filter pattern information storage unit) in the first embodiment of the present invention. ) 201 and the filter pattern selection unit (filter pattern selection means) 202, the detailed description thereof will be omitted.

  The filter pattern information storage unit 5001 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, in association with an identifier for each fractional pixel position between integer pixels.

  The filter pattern selection unit 5002 stores the encoding target video sequence 5006 input from the outside in the filter pattern information storage unit 5001 for each fractional pixel position between integer pixels in the encoding target image sequence unit. One filter pattern is selected from the filtered filter patterns, and a filter pattern identifier 5007 that is an identifier of the selected filter pattern is output to the filter pattern information encoding unit 5003 and the frame image encoding unit 5004.

  The filter pattern information encoding unit 5003 encodes the filter pattern identifier 5007 for each fractional pixel position between integer pixels input from the filter pattern selection unit 5002 for each image sequence to be encoded, and performs code synthesis as a filter pattern code 5008. Output to the unit 5005.

  The frame image encoding unit 5004 discriminates the filter pattern associated with the filter pattern identifier 5007 for each fractional pixel position between integer pixels input from the filter pattern selection unit 5002 with reference to the filter pattern information storage unit 5001. The encoding target moving image sequence 5006 input from the outside is encoded using the determined filter pattern. The frame image encoding unit 5004 outputs the encoded frame image code 5009 to the code synthesis unit 5005.

  Specifically, the frame image encoding unit 5004 encodes a filter coefficient of one part that is a significant coefficient in the filter pattern associated with the filter pattern identifier 5007 for each fractional pixel position between integer pixels. Generated for each image. At this time, when the frame image encoding unit 5004 creates a reference image with fractional pixel accuracy by filtering a reference image encoded in the past, an error from the signal of the frame image to be encoded is minimized. The filter coefficient of the significant coefficient 1 part in the filter pattern is generated.

  The frame image encoding unit 5004 generates, for each frame image to be encoded, one portion of the filter coefficient that is a significant coefficient in the filter pattern associated with the filter pattern identifier 5007 for each fractional pixel position between integer pixels. The method is not limited to a method that minimizes the error from the signal of the frame image to be encoded when a reference image with a fractional pixel accuracy is created by filtering a reference image encoded in the past. Other methods are also acceptable.

  The code synthesis unit 5005 combines the filter pattern code 5008 input from the filter pattern information encoding unit 5003 and the frame image code 5009 input from the frame image encoding unit 5004, and outputs the combined code 5010 to the outside.

  Note that a method in which the code synthesis unit 5005 combines the filter pattern code 5008 and the frame image code 5009 includes a method in which the filter pattern code 5008 is used as a sequence header and the same bit stream as the frame image code 5009 is used. In addition, the method in which the code synthesis unit 5005 matches the filter pattern code 5008 and the frame image code 5009 is not limited to this method, and any synthesis method is possible.

  Next, the operation of the moving picture coding method according to the third modification of the present embodiment will be described using FIG. FIG. 51 is a flowchart illustrating a video encoding method that is an operation of the video encoding apparatus 500 according to the third modification of the present embodiment.

  First, the filter pattern information storage unit 5001 stores a plurality of filter patterns, which are patterns of significant coefficients of a filter of a two-dimensional matrix for each fractional pixel position between integer pixels, in association with an identifier (step S5101).

  Next, with respect to the moving image sequence 5006 to be encoded input from the outside by the filter pattern selection unit 5002, the filter pattern information storage unit 5001 stores the fractional pixel positions between integer pixels in units of the encoding target image sequence. One filter pattern is selected from the stored filter patterns and its identifier is determined (step S5102).

  Next, the filter pattern information encoding unit 5003 encodes the filter pattern identifier for each fractional pixel position between the integer pixels selected by the filter pattern selection unit 5002 in units of image sequences to be encoded (step S5103).

  Also, the frame image encoding unit 5004 determines the filter pattern of the filter pattern identifier for each fractional pixel position between the integer pixels selected by the filter pattern selection unit 5002 with reference to the filter pattern information storage unit 5001, and the external Is encoded using the determined filter pattern (step S5104).

  Next, the code composition unit 5005 outputs the filter pattern code and the frame image code together by the code composition unit 5005 (step S5105).

  Step S5102 which is the operation of the filter pattern selection unit 5002 is the same as step S1802 which is the operation of the filter pattern selection unit 202 in the first embodiment of the present invention, and thus description thereof is omitted.

  Next, step S5103, which is the operation of the filter pattern information encoding unit 5003, will be described in detail with reference to FIG. FIG. 52 is a flowchart illustrating step S5103, which is the operation of the filter pattern selection unit 5002 according to Modification 3 of the present embodiment.

  First, the filter pattern information encoding unit 5003 encodes the identifier of the filter pattern at the fractional pixel position between one integer pixel for the moving image to be encoded (step S5201).

  Next, the filter pattern information encoding unit 5003 determines whether or not the filter pattern identifier has been encoded at the fractional pixel positions between all the integer pixels (step S5202), and the filter pattern at all the fractional pixel positions between the integer pixels. Is not encoded (NO), the process is repeated from step S5001 for the fractional pixel position between the next integer pixels. If the filter pattern identifier is encoded at the fractional pixel position between all integer pixels (YES), the process ends.

  Next, step S5104, which is the operation of the frame image encoding unit 5004, will be described in detail with reference to FIG. FIG. 53 is a flowchart for explaining step S5104 which is the operation of the frame image encoding unit 5004 according to the third modification of the present embodiment.

  First, the frame image encoding unit 5004 discriminates the filter pattern by referring to the filter pattern information storage unit 5001 from the filter pattern identifier at the fractional pixel position between one integer pixel (step S5301), and the discriminated filter pattern. The filter coefficient of 1 part, which is a significant coefficient, is generated for each frame image to be encoded, and the filter is determined (step S5302). In addition, among the determined filters, the filter coefficient of 1 part which is a significant coefficient in the filter pattern is encoded (step S5305).

  Next, the frame image encoding unit 5004 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S5304), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), the process is repeated from step S5301 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image encoding unit 5004 creates a reference image with fractional pixel accuracy by applying the filter for each fractional pixel position between integer pixels thus determined to a reference image that has been encoded in the past (step S5305).

  Next, the frame image encoding unit 5004 encodes the moving image sequence 5006 to be encoded by performing motion compensation using the reference image with fractional pixel accuracy (step S5306), and ends the process.

  Next, step S5105, which is the operation of the code synthesis unit 5005, will be described in detail with reference to FIG. FIG. 54 is a flowchart illustrating step S5105, which is the operation of the code synthesis unit 5005 according to the third modification of the present embodiment.

  First, the code synthesis unit 5005 synthesizes the code of the input filter pattern identifier and the code obtained by encoding the input image (step S5401).

  Next, the code synthesis unit 5005 outputs a code obtained by combining the code of the identifier of the filter pattern and the code obtained by encoding the input image (step S5402), and ends the process.

  Next, a moving picture coding program 550 for causing a computer to function as the above-described moving picture coding apparatus 500 will be described. FIG. 55 is a diagram showing a configuration of the moving image encoding program 550.

  As shown in FIG. 55, the moving image encoding program 550 includes a main module program 5501 that supervises processing, a filter pattern information storage module 5502, a filter pattern selection module 5503, a filter pattern information encoding module 5504, A frame image encoding module 5505 and a code synthesis module 5506 are provided. The filter pattern information storage module 5502, the filter pattern selection module 5503, the filter pattern information encoding module 5504, the frame image encoding module 5505, and the code synthesizing module 5506 have the functions that the computer performs, respectively. This is the same as the pattern information storage unit 5001, the filter pattern selection unit 5002, the filter pattern information encoding unit 5003, the frame image encoding unit 5004, and the code synthesis unit 5005.

  FIG. 56 is a block diagram illustrating a configuration of a video decoding device 560 according to Modification 3 of the present embodiment. The moving picture decoding apparatus 560 includes, as its functional components, a filter pattern information storage unit (filter pattern storage unit) 5601, a code separation unit (code separation unit) 5602, and a filter pattern information decoding unit (filter pattern). (Information decoding means) 5603 and a frame image decoding unit (frame image decoding means) 5604. Since the filter pattern information storage unit (filter pattern information storage unit) 5601 is the same as the filter pattern information storage unit (filter pattern information storage unit) 201 in the first embodiment of the present invention, a detailed description thereof will be given. Omitted.

  The filter pattern information storage unit 5601 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, in association with an identifier for each fractional pixel position between integer pixels. The filter pattern information storage unit 5601 is the same as the filter pattern information storage unit 5001 in the moving image encoding apparatus 500.

  The code separation unit 5602 receives a composite code 5605 input from the outside, and separates it into a filter pattern code 5606 and a frame image code 5608. The code separation unit 5602 outputs the filter pattern code 5606 to the filter pattern information decoding unit 5603. Also, the code separation unit 5602 outputs the frame image code 5608 to the frame image decoding unit 5604.

  Note that the code separation unit 5602 separates the composite code 5605 into the filter pattern code 5606 and the frame image code 5608 by separating the filter pattern code 5606 from the bit stream as a sequence header. Further, the method by which the code separation unit 5602 separates the filter pattern code 5606 and the frame image code 5608 is not limited to this method, and any separation method is possible.

  The filter pattern information decoding unit 5603 decodes the filter pattern code 5606 for each fractional pixel position between integer pixels input from the code separation unit 5602 in units of image sequences to be encoded, and performs frame image decoding as the filter pattern identifier 5607. To the conversion unit 5604.

  The frame image decoding unit 5604 refers to the filter pattern information storage unit 5601 for the filter pattern associated with the filter pattern identifier 5607 for each fractional pixel position between integer pixels input from the filter pattern information decoding unit 5603. Using the determined filter pattern, the frame image code 5608 input from the code separation unit 5602 is decoded and output to the outside such as a display device (not specified).

  Specifically, the frame image decoding unit 5604 uses a frame image code 5608 as a filter coefficient of one part which is a significant coefficient in the filter pattern associated with the filter pattern identifier 5607 for each fractional pixel position between integer pixels. To generate a filter. Next, the frame image decoding unit 5604 creates a reference image with fractional pixel accuracy using a filter generated for each fractional pixel position between integer pixels with respect to a reference image that has been decoded in the past, and performs motion compensation. Thus, the frame image to be decoded is decoded.

  Next, the operation of the moving picture decoding method according to the third modification of the present embodiment will be described using FIG. FIG. 57 is a flowchart illustrating a video decoding method that is an operation of the video decoding device 560 according to Modification 3 of the present embodiment.

  First, the filter pattern information storage unit 5601 stores a plurality of filter patterns, which are patterns of significant coefficients of a two-dimensional matrix filter, for each fractional pixel position between integer pixels in association with an identifier (step S5701).

  Next, the code separation unit 5602 separates the composite code input from the outside into a filter pattern code and a frame image code. (Step S5702).

  Next, the filter pattern information decoding unit 5603 decodes the filter pattern identifier for each fractional pixel position between integer pixels in units of the decoding target image sequence from the filter pattern information code (step S5703).

  Next, the frame image decoding unit 5604 refers to the filter pattern information storage unit 5601 for the filter pattern associated with the filter pattern identifier for each fractional pixel position between the integer pixels input from the filter pattern information decoding unit 5603. The frame image code is decoded using the determined filter pattern and output to the outside such as a display device (not specified) (step S5704).

  Next, step S5702 which is the operation of the code separation unit 5602 will be described in detail with reference to FIG. FIG. 58 is a flowchart illustrating step S5702 which is the operation of the code separation unit 5602 according to the third modification of the present embodiment.

  First, the code separation unit 5602 receives a code obtained by combining the code of the filter pattern identifier and the code obtained by encoding the input image from the outside (step S5801).

  Next, the code separation unit 5602 separates the code obtained by combining the code of the filter pattern identifier and the code obtained by encoding the input image into the filter pattern code and the frame image code (step S5802), and ends the process.

  Next, step S5703, which is the operation of the filter pattern information decoding unit 5603, will be described in detail with reference to FIG. FIG. 59 is a flowchart illustrating step S5703, which is the operation of the filter pattern information decoding unit 5603 according to the third modification of the present embodiment.

  First, the filter pattern information decoding unit 5603 inputs the filter pattern code at the fractional pixel position between one integer pixel from the code separation unit to the decoding target moving image (step S5901).

Next, the filter pattern information decoding unit 5603 decodes a filter pattern identifier, which is a filter pattern identifier, from the input filter pattern code. (Step S5902)
Next, the filter pattern information decoding unit 5603 determines whether or not the filter pattern identifier has been decoded at the fractional pixel positions between all the integer pixels (step S5903), and the filter pattern identifier at all the fractional pixel positions between the integer pixels. Is not decoded (NO), the process is repeated from step S5901 for the fractional pixel position between the next integer pixels. If the filter pattern identifier is decoded at the fractional pixel position between all the integer pixels (YES), the process ends.

  Next, step S5704 which is the operation of the frame image decoding unit 5604 will be described in detail with reference to FIG. FIG. 60 is a flowchart illustrating step S5704, which is the operation of the frame image decoding unit 5604 according to the third modification of the present embodiment.

  First, the frame image decoding unit 5604 receives the frame image code from the code separation unit (step S6001).

  Next, the frame image decoding unit 5604 discriminates the filter pattern by referring to the filter pattern information storage unit 5601 from the filter pattern identifier at the fractional pixel position between one integer pixel (step S6002), and the discriminated filter. The filter coefficient of 1 part which is a significant coefficient in the pattern is decoded for each frame image to be decoded (step S6003), and a filter is determined (step S6004).

  Next, the frame image decoding unit 5604 determines whether or not filters at fractional pixel positions between all integer pixels have been determined (step S6005), and filters at fractional pixel positions between all integer pixels are determined. If not (NO), the process is repeated from step S6002 for the fractional pixel position between the next integer pixels. When the filter at the fractional pixel position between all the integer pixels is determined (YES), the process proceeds to the next step.

  Next, the frame image decoding unit 5604 creates a reference image with fractional pixel accuracy by applying the determined filter for each fractional pixel position between integer pixels to a reference image that has been encoded in the past (step S6006).

  Next, the frame image decoding unit 5604 performs motion compensation on the moving image sequence to be decoded from the frame image code using the reference image with fractional pixel accuracy, and outputs the decoded sequence (step S6007). finish.

  Next, a video decoding program 610 for causing a computer to function as the above-described video decoding device 560 will be described. FIG. 61 is a diagram showing a configuration of the moving picture decoding program 610.

  As shown in FIG. 61, the moving picture decoding program 610 includes a main module program 6101 that controls processing, a filter pattern information storage module 6102, a code separation module 6103, a filter pattern information decoding module 6104, a frame, An image decoding module 6105. The filter pattern information storage module 6102, code separation module 6103, filter pattern information decoding module 6104, and frame image decoding module 6105 have the functions that the computer performs respectively corresponding to the above-described filter pattern information storage unit 5601. The code separation unit 5602, the filter pattern information decoding unit 5603, and the frame image decoding unit 5604 are the same.

  According to the moving image encoding apparatus, the moving image encoding method, the moving image encoding program, the moving image decoding apparatus, the moving image decoding method, and the moving image decoding program as described above, the fractional pixel accuracy using the filter When the reference image is generated and motion compensated for encoding and decoding, the filter pattern of the filter coefficient is encoded for each moving image sequence to be encoded, and the significant coefficient of the filter pattern is encoded for each frame image to be encoded. Since it is only necessary to encode the filter coefficient, it is possible to perform motion compensation with high accuracy by using a filter suitable for the feature of the video while reducing the code amount of the filter coefficient. Further, according to such a moving image encoding device, a moving image encoding method, and a moving image encoding program, by encoding and outputting a filter pattern as a part of an encoded bit stream, The association can be facilitated. In addition, according to such a moving image decoding apparatus, moving image decoding method, and moving image decoding program, the filter pattern is decoded from a part of the encoded bitstream, thereby associating with the frame image code. Can be facilitated.

  In addition, the filter pattern selection unit 5002 does not select a filter pattern for each fractional pixel position between integer pixels in a moving image sequence unit to be encoded, but in a frame image group unit in which a plurality of frame images to be encoded are collected. A filter pattern may be selected for each fractional pixel position between integer pixels. In this case, the filter pattern information encoding unit 5003 encodes the filter pattern code 5008 for each frame image group in which a plurality of encoding target frame images are collected. In addition, the code synthesis unit 5005 synthesizes a filter pattern code 5008 and a frame image code 5009 as header information for each frame image group in which a plurality of frame images to be encoded are collected. In addition, the code separation unit 5602 separates the filter pattern code 5606 as header information for each frame image group in which a plurality of frame images to be encoded are collected from the input composite code. Further, the filter pattern information decoding unit 5603 decodes the filter pattern code 5606 in units of frame image groups in which a plurality of decoding target frame images are collected. Thus, when generating a fractional pixel precision reference image using a filter and performing motion compensation for encoding and decoding, a filter pattern of filter coefficients in units of frame images in which a plurality of frame images to be encoded are grouped together Since only the filter coefficient of the significant coefficient of the filter pattern is encoded and decoded in units of frame images to be encoded, a frame group of a plurality of video frame images combined while reducing the code amount of the filter coefficient. Motion compensation can be performed with high accuracy using a filter suitable for the feature.

  Further, the filter pattern selection unit 5002 may select a filter pattern for each frame image to be encoded instead of selecting a filter pattern for each fractional pixel position between integer pixels in units of a moving image sequence to be encoded. good. In this case, the filter pattern information encoding unit 5003 encodes the filter pattern code 5008 for each frame image to be encoded. Also, the code synthesis unit 5005 synthesizes the filter pattern code 5008 with the frame image code 5009 as frame header information for each frame image to be encoded. Also, the code separation unit 5602 separates the filter pattern code 5606 as frame header information for each frame image to be encoded from the input composite code. Also, the filter pattern information decoding unit 5603 decodes the filter pattern code 5606 for each frame image to be decoded. As a result, when a reference image with fractional pixel accuracy is generated using a filter, and motion compensation is performed for encoding and decoding, the filter pattern of the filter coefficient is encoded and encoded for each frame image to be encoded. Since only the filter coefficient of the significant coefficient of the filter pattern is encoded and decoded for each frame image, motion compensation is performed with high accuracy using a filter that matches the characteristics of the video frame image while reducing the code amount of the filter coefficient. can do.

It is an arrangement diagram of pixels of a reference image having a 1/4 pixel accuracy in a conventional encoding method. It is a block diagram which shows the structure of the moving image encoder 20 which concerns on this embodiment. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position a of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position b of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position c of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position d of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position e of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position f of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position g of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position h of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position i of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position j of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position k of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position n of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position p of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position q of FIG. It is a figure which shows the filter pattern of three types of two-dimensional 6 tap filters about the fraction pixel position r of FIG. It is a flowchart explaining the moving image encoding method which is operation | movement of the moving image encoder 20 which concerns on this embodiment. It is a flowchart explaining step S1802 which is operation | movement of the filter pattern selection part 202 which concerns on this embodiment. It is a flowchart explaining step S1803 which is operation | movement of the filter pattern selection part 202 which concerns on this embodiment. It is a flowchart explaining step S1804 which is operation | movement of the frame image coding part 204 which concerns on this embodiment. It is a figure which shows the structure of the moving image encoding program 220. FIG. It is a block diagram which shows the structure of the moving image decoding apparatus 230 which concerns on this embodiment. It is a flowchart explaining the moving image decoding method which is operation | movement of the moving image decoding apparatus 230 which concerns on this embodiment. It is a flowchart explaining step S2402 which is operation | movement of the filter pattern information decoding part 2302 which concerns on this embodiment. It is a flowchart explaining step S2403 which is operation | movement of the frame image decoding part 2303 which concerns on this embodiment. It is a figure which shows the structure of the moving image decoding program 270. FIG. It is a block diagram which shows the structure of the moving image encoder 280 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining the moving image encoding method which is operation | movement of the moving image encoder 280 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining step S2902 which is operation | movement of the filter pattern selection part 2802 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining step S2903 which is operation | movement of the filter pattern selection part 2802 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining step S2904 which is operation | movement of the frame image encoding part 2804 which concerns on the example 1 of a change of this embodiment. It is a figure which shows the structure of the moving image encoding program. It is a block diagram which shows the structure of the moving image decoding apparatus 340 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining the moving image decoding method which is operation | movement of the moving image decoding apparatus 340 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining step S3502 which is operation | movement of the filter pattern information decoding part 3402 which concerns on the example 1 of a change of this embodiment. It is a flowchart explaining step S3503 which is operation | movement of the frame image decoding part 3403 which concerns on the example 1 of a change of this embodiment. It is a figure which shows the structure of the moving image decoding program 380. FIG. It is a block diagram which shows the structure of the moving image encoder 390 which concerns on the example 2 of a change of this embodiment. It is a flowchart explaining the moving image encoding method which is operation | movement of the moving image encoder 390 which concerns on the modification 2 of this embodiment. It is a flowchart explaining step S4002 which is operation | movement of the filter pattern selection part 3902 which concerns on the example 2 of a change of this embodiment. It is a flowchart explaining step S4003 which is operation | movement of the filter pattern selection part 3902 which concerns on the example 2 of a change of this embodiment. It is a flowchart explaining step S4004 which is operation | movement of the frame image encoding part 3904 which concerns on the example 2 of a change of this embodiment. It is a figure which shows the structure of the moving image encoding program. It is a block diagram which shows the structure of the moving image decoding apparatus 450 which concerns on the modification 2 of this embodiment. It is a flowchart explaining the moving image decoding method which is operation | movement of the moving image decoding apparatus 450 which concerns on the modification 2 of this embodiment. It is a flowchart explaining step S4601 which is operation | movement of the filter pattern information decoding part 4501 which concerns on the modification 2 of this embodiment. It is a flowchart explaining step S4602 which is operation | movement of the frame image decoding part 4502 which concerns on the modification 2 of this embodiment. It is a figure which shows the structure of the moving image decoding program 490. FIG. It is a block diagram which shows the structure of the moving image encoder 500 which concerns on the modification 3 of this embodiment. It is a flowchart explaining the moving image encoding method which is operation | movement of the moving image encoding apparatus 500 which concerns on the example 3 of a change of this embodiment. It is a flowchart explaining step S5103 which is operation | movement of the filter pattern selection part 5002 which concerns on the modification 3 of this embodiment. It is a flowchart explaining step S5104 which is operation | movement of the frame image encoding part 5004 which concerns on the modification 3 of this embodiment. It is a flowchart explaining step S5105 which is operation | movement of the code | symbol synthesis part 5005 which concerns on the example 3 of a change of this embodiment. It is a figure which shows the structure of the moving image encoding program 550. FIG. It is a block diagram which shows the structure of the moving image decoding apparatus 560 which concerns on the modification 3 of this embodiment. It is a flowchart explaining the moving image decoding method which is operation | movement of the moving image decoding apparatus 560 which concerns on the example 3 of a change of this embodiment. It is a flowchart explaining step S5702 which is operation | movement of the code separation part 5602 which concerns on the modification 3 of this embodiment. It is a flowchart explaining step S5703 which is operation | movement of the filter pattern information decoding part 5603 which concerns on the modification 3 of this embodiment. It is a flowchart explaining step S5704 which is operation | movement of the frame image decoding part 5604 which concerns on the modification 3 of this embodiment. It is a figure which shows the structure of the moving image decoding program 610.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 20 ... Moving image encoding apparatus, 201 ... Filter pattern information storage part, 202 ... Filter pattern selection part, 203 ... Filter pattern information encoding part, 204 ... Frame image encoding part, 205 ... Moving image sequence, 206 ... Filter pattern Identifier: 207: Filter pattern code, 208: Frame image code, 220: Video encoding program, 2201: Main module program, 2202: Filter pattern information storage module, 2203: Filter pattern selection module, 2204: Filter pattern information encoding Modules, 2205 ... Frame image encoding module, 230 ... Moving picture decoding device, 2301 ... Filter pattern information storage unit, 2302 ... Filter pattern information decoding unit, 2303 ... Frame image decoding unit, 2304 ... Filter pattern code, 2305 ... Filter pattern identifier, 2306 ... Frame image code, 270 ... Moving picture decoding program, 2701 ... Main module program, 2702 ... Filter pattern information storage module, 2703 ... Filter pattern information decoding module, 2704 ... Frame Image decoding module, 280 ... moving image encoding device, 2801 ... filter pattern information storage unit, 2802 ... filter pattern selection unit, 2803 ... filter pattern information encoding unit, 2804 ... frame image encoding unit, 2805 ... moving image sequence 2806 ... Filter pattern identifier, 2807 ... Filter pattern, 2808 ... Filter pattern code, 2809 ... Frame image code, 330 ... Moving picture encoding program, 3301 ... Main module Program, 3302 ... Filter pattern information storage module, 3303 ... Filter pattern selection module, 3304 ... Filter pattern information encoding module, 3305 ... Frame image encoding module, 340 ... Moving picture decoding device, 3401 ... Filter pattern information storage unit, 3402 ... Filter pattern information decoding unit, 3403 ... Frame image decoding unit, 3404 ... Filter pattern code, 3405 ... Filter pattern, 3406 ... Frame image code, 380 ... Moving picture decoding program, 3801 ... Main module program, 3802 ... Filter pattern information storage module, 3803 ... Filter pattern information decoding module, 3804 ... Frame image decoding module, 390 ... Moving picture encoding device, 3901 ... Filter pattern Information storage unit, 3902 ... Filter pattern selection unit, 3903 ... Filter pattern information encoding unit, 3904 ... Frame image encoding unit, 3905 ... Moving image sequence, 3906 ... Filter pattern, 3907 ... Filter pattern code, 3908 ... Frame image code 440: moving image encoding program, 4401: main module program, 4402: filter pattern information storage module, 4403: filter pattern selection module, 4404: filter pattern information encoding module, 4405: frame image encoding module, 450: moving image Image decoding apparatus, 4501 ... Filter pattern information decoding unit, 4502 ... Frame image decoding unit, 4503 ... Filter pattern code, 4504 ... Filter pattern, 4505 ... Frame image code, 90 ... moving image decoding program, 4901 ... main module program, 4902 ... filter pattern information decoding module, 4903 ... frame image decoding module, 500 ... moving image encoding device, 5001 ... filter pattern information storage unit, 5002 ... filter Pattern selection unit, 5003 ... Filter pattern information encoding unit, 5004 ... Frame image encoding unit, 5005 ... Code synthesis unit, 5006 ... Video sequence, 5007 ... Filter pattern identifier, 5008 ... Filter pattern code, 5009 ... Frame image code , 5010 ... composite code, 550 ... moving picture encoding program, 5501 ... main module program, 5502 ... filter pattern information storage module, 5503 ... filter pattern selection module, 5504 ... fill Pattern information encoding module, 5505 ... frame image encoding module, 5506 ... code synthesis module, 560 ... moving picture decoding device, 5601 ... filter pattern information storage unit, 5602 ... code separation unit, 5603 ... filter pattern information decoding unit 5604 ... frame image decoding unit, 5605 ... composite code, 5606 ... filter pattern code, 5607 ... filter pattern identifier, 5608 ... frame image code, 610 ... moving picture decoding program, 6101 ... main module program, 6102 ... filter pattern Information storage module, 6103 ... code separation module, 6104 ... filter pattern information decoding module, 6105 ... frame image decoding module.

Claims (30)

A moving image encoding device that performs motion compensation by creating a reference image with fractional pixel accuracy using a filter for a moving image sequence formed of a time sequence of frame images,
A filter pattern information storage unit that stores a plurality of filter patterns that are basic patterns of filters;
A filter pattern selection unit that selects a filter pattern from a plurality of filter patterns stored in the filter pattern information storage unit;
A filter pattern information encoding unit that encodes information about the filter pattern selected by the filter pattern selection unit;
A frame image encoding unit that generates a filter based on information on the filter pattern selected by the filter pattern selection unit, creates a reference image with fractional pixel accuracy, performs motion compensation, and encodes a frame image;
Equipped with a,
The filter pattern is a binary pattern of a combination of a significant coefficient and an insignificant coefficient.
A moving picture coding apparatus characterized by the above. The filter pattern selection unit selects a filter pattern in units of moving image sequences from a plurality of filter patterns stored in the filter pattern information storage unit,
The moving image encoding apparatus according to claim 1, wherein the filter pattern information encoding unit encodes information about the filter pattern selected by the filter pattern selection unit in units of moving image sequences. The frame image encoding unit generates a filter for each frame image to be encoded based on the filter pattern selected by the filter pattern selection unit, creates a reference image with fractional pixel accuracy, performs motion compensation, The video encoding apparatus according to claim 1, wherein the video is encoded. The filter pattern selection unit selects a filter pattern for each image sequence to be encoded from a plurality of filter patterns stored in the filter pattern information storage unit,
The filter pattern information encoding unit encodes information about the filter pattern selected by the filter pattern selection unit in units of moving image sequences,
The frame image encoding unit generates a filter for each frame image to be encoded based on the filter pattern selected by the filter pattern selection unit, creates a reference image with fractional pixel accuracy, performs motion compensation, The video encoding apparatus according to claim 1, wherein the video is encoded. The filter pattern information storage unit stores each of the plurality of filter patterns in association with an identifier,
The filter pattern information encoding unit, said as the information relates to a filter pattern, video encoding according to any one of claim 1 to 4, the identifier associated with the filter pattern, characterized in that encode Device. The filter pattern information storage unit stores the plurality of filter patterns in association with identifiers for each fractional pixel position between integer pixels,
The filter pattern selection unit selects one filter pattern from the plurality of filter patterns for each fractional pixel position between the integer pixels,
The filter pattern information encoding unit encodes the identifier associated with the filter pattern selected by the filter pattern selection unit for each fractional pixel position between the integer pixels as information on the filter pattern,
The frame image encoding unit generates a filter based on a filter pattern associated with the identifier for each fractional pixel position between the integer pixels, generates a fractional pixel precision reference image to perform motion compensation, The moving image encoding apparatus according to claim 5 , wherein the image is encoded. The moving image encoding device is:
The code output unit further comprising: a code output unit that outputs the filter pattern code encoded by the filter pattern information encoding unit and the frame image code encoded by the frame image encoding unit. The moving image encoding device according to any one of 1 to 6 . A video decoding device that creates a reference image with fractional pixel accuracy using a filter, performs motion compensation, and decodes a video sequence composed of a time sequence of frame images,
A filter pattern information decoding unit that decodes a filter pattern code obtained by encoding information on a filter pattern that is a basic pattern of a filter, and obtains information on the filter pattern;
A frame image decoding unit that generates a filter based on information about the filter pattern decoded by the filter pattern information decoding unit, creates a fractional pixel precision reference image, performs motion compensation, and decodes the frame image;
Equipped with a,
The filter pattern is a binary pattern of a combination of a significant coefficient and an insignificant coefficient.
A moving picture decoding apparatus characterized by the above. The moving picture decoding apparatus according to claim 8 , wherein the filter pattern information decoding unit decodes information on the filter pattern in units of moving picture sequences. The frame image decoding unit generates a filter for each frame image to be decoded based on information about the filter pattern decoded by the filter pattern information decoding unit, creates a reference image with fractional pixel accuracy, and moves 9. The moving picture decoding apparatus according to claim 8 , wherein compensation is performed and the frame picture is decoded. The filter pattern information decoding unit decodes information about the filter pattern in units of moving image sequences,
The frame image decoding unit generates a filter for each frame image to be decoded based on information about the filter pattern decoded by the filter pattern information decoding unit, creates a reference image with fractional pixel accuracy, and moves 9. The moving picture decoding apparatus according to claim 8 , wherein compensation is performed and the frame picture is decoded. The moving image decoding apparatus further includes a filter pattern information storage unit that stores a plurality of filter patterns in association with identifiers of the plurality of filter patterns,
The moving picture decoding according to any one of claims 8 to 11 , wherein the filter pattern information decoding unit decodes the identifier associated with the filter pattern as information on the filter pattern. Device. The filter pattern information storage unit stores the plurality of filter patterns in association with identifiers for each fractional pixel position between integer pixels,
The filter pattern information decoding unit decodes the identifier associated with the filter pattern for each fractional pixel position between the integer pixels as information on the filter pattern,
The frame image decoding unit generates a filter based on a filter pattern associated with the identifier for each fractional pixel position between the integer pixels, creates a fractional pixel precision reference image to perform motion compensation, 13. The moving picture decoding apparatus according to claim 12 , wherein the moving picture decoding apparatus decodes an image. The moving picture decoding apparatus comprises:
A code separation unit that separates the input code into a filter pattern code obtained by coding information on the filter pattern and a frame image code obtained by coding the frame image; The moving picture decoding apparatus according to claim 8 , wherein the moving picture decoding apparatus is characterized. A moving image encoding method executed by a moving image encoding apparatus that generates a fractional pixel precision reference image using a filter and performs motion compensation on a moving image sequence composed of a time sequence of frame images. ,
A filter pattern information storage step for storing a plurality of filter patterns that are basic patterns of the filter;
A filter pattern selection step of selecting a filter pattern from a plurality of filter patterns stored in the filter pattern information storage step;
A filter pattern information encoding step for encoding information on the filter pattern selected in the filter pattern selection step;
A frame image encoding step of generating a filter based on information on the filter pattern selected in the filter pattern selection step, creating a reference image with fractional pixel accuracy to compensate for motion, and encoding a frame image;
Equipped with a,
The filter pattern is a binary pattern of a combination of a significant coefficient and an insignificant coefficient.
A video encoding method characterized by the above. In the filter pattern selection step, from the plurality of filter patterns stored in the filter pattern information storage step, select a filter pattern in units of moving image sequences,
16. The moving picture coding method according to claim 15 , wherein in the filter pattern information coding step, information on the filter pattern selected in the filter pattern selection step is coded in a moving picture sequence unit. In the frame image encoding step, based on the filter pattern selected in the filter pattern selection step, a filter is generated for each frame image to be encoded, and a reference image with fractional pixel accuracy is generated to compensate for motion, 16. The moving image encoding method according to claim 15 , wherein the image is encoded. In the filter pattern selection step, from the plurality of filter patterns stored in the filter pattern information storage step, a filter pattern is selected for each image sequence to be encoded,
In the filter pattern information encoding step, information on the filter pattern selected in the filter pattern selection step is encoded in units of moving image sequences,
In the frame image encoding step, based on the filter pattern selected in the filter pattern selection step, a filter is generated for each frame image to be encoded, and a reference image with fractional pixel accuracy is generated to compensate for motion, 16. The moving image encoding method according to claim 15 , wherein the image is encoded. In the filter pattern information storage step, each of the plurality of filter patterns is stored in association with an identifier,
The moving image code according to any one of claims 15 to 18 , wherein, in the filter pattern information encoding step, the identifier associated with the filter pattern is encoded as information on the filter pattern. Method. In the filter pattern information storage step, each of the plurality of filter patterns is stored in association with an identifier for each fractional pixel position between integer pixels,
In the filter pattern selection step, one filter pattern is selected from the plurality of filter patterns for each fractional pixel position between the integer pixels,
In the filter pattern information encoding step, as the information about the filter pattern, the identifier associated with the filter pattern selected in the filter pattern selection step is encoded for each fractional pixel position between the integer pixels,
In the frame image encoding step, a filter is generated based on a filter pattern associated with the identifier for each fractional pixel position between the integer pixels, a reference image with fractional pixel accuracy is created to perform motion compensation, and a frame The moving image encoding method according to claim 19 , wherein the image is encoded. The moving image encoding method includes:
The code output step of outputting the filter pattern code encoded in the filter pattern information encoding step and the frame image code encoded in the frame image encoding step together. The moving image encoding method according to any one of 15 to 20 . This is a moving picture decoding method executed by a moving picture decoding apparatus that generates a reference picture with fractional pixel accuracy using a filter, performs motion compensation, and decodes a moving picture sequence composed of a time series of frame images. And
A filter pattern information decoding step for decoding a filter pattern code obtained by encoding information on a filter pattern, which is a basic pattern of a filter, and obtaining information on the filter pattern;
A frame image decoding step of generating a filter based on the information about the filter pattern decoded in the filter pattern information decoding step, creating a reference image with fractional pixel accuracy to perform motion compensation, and decoding the frame image;
Equipped with a,
The filter pattern is a binary pattern of a combination of a significant coefficient and an insignificant coefficient.
A moving picture decoding method characterized by the above. 23. The moving picture decoding method according to claim 22 , wherein in the filter pattern information decoding step, information on the filter pattern is decoded in a moving picture sequence unit. In the frame image decoding step, based on the information about the filter pattern decoded in the filter pattern information decoding step, a filter is generated for each frame image to be decoded, and a reference image with fractional pixel accuracy is generated and moved. 23. The moving picture decoding method according to claim 22 , wherein compensation is performed and the frame picture is decoded. In the filter pattern information decoding step, information on the filter pattern is decoded in units of video sequences,
In the frame image decoding step, based on the information about the filter pattern decoded in the filter pattern information decoding step, a filter is generated for each frame image to be decoded, and a reference image with fractional pixel accuracy is generated and moved. 23. The moving picture decoding method according to claim 22 , wherein compensation is performed and the frame picture is decoded. The moving picture decoding method further includes a filter pattern information storage step of storing a plurality of filter patterns in association with identifiers of the plurality of filter patterns,
26. The moving picture decoding according to claim 22 , wherein, in the filter pattern information decoding step, the identifier associated with the filter pattern is decoded as information on the filter pattern. Method. In the filter pattern information storage step, each of the plurality of filter patterns is stored in association with an identifier for each fractional pixel position between integer pixels,
In the filter pattern information decoding step, as the information about the filter pattern, the identifier associated with the filter pattern is decoded for each fractional pixel position between the integer pixels,
In the frame image decoding step, a filter is generated based on a filter pattern associated with the identifier for each fractional pixel position between the integer pixels, a reference image with fractional pixel accuracy is created and motion compensated, 27. The moving picture decoding method according to claim 26 , wherein the picture is decoded. The moving picture decoding method includes:
A code separation step for separating the input code into a filter pattern code obtained by coding information on the filter pattern and a frame image code obtained by coding the frame image; 28. The moving picture decoding method according to any one of claims 22 to 27 , characterized in that: A moving image encoding program for causing a computer to function to compensate for motion by creating a reference image with fractional pixel accuracy using a filter for a moving image sequence composed of a time sequence of frame images,
Computer
A filter pattern information storage unit that stores a plurality of filter patterns that are basic patterns of filters;
A filter pattern selection unit that selects a filter pattern from a plurality of filter patterns stored in the filter pattern information storage unit;
A filter pattern information encoding unit that encodes information about the filter pattern selected by the filter pattern selection unit;
A frame image encoding unit that generates a filter based on information about the filter pattern selected by the filter pattern selection unit, creates a reference image with fractional pixel accuracy, performs motion compensation, and encodes a frame image;
Function as
The filter pattern is a binary pattern of a combination of a significant coefficient and an insignificant coefficient.
A moving picture encoding program characterized by the above. A moving picture decoding program for causing a computer to create a reference picture with fractional pixel accuracy using a filter to perform motion compensation and to decode a moving picture composed of a time series of frame images,
Computer
A filter pattern information decoding unit that decodes a filter pattern code obtained by encoding information on a filter pattern that is a basic pattern of a filter, and obtains information on the filter pattern;
A frame image decoding unit that generates a filter based on information about the filter pattern decoded by the filter pattern information decoding unit, creates a fractional pixel precision reference image, performs motion compensation, and decodes the frame image;
Function as
The filter pattern is a binary pattern of a combination of a significant coefficient and an insignificant coefficient.
A moving picture decoding program characterized by the above.

Images