WO2012001833A1

WO2012001833A1 - Moving image encoding apparatus, moving image decoding apparatus and method

Info

Publication number: WO2012001833A1
Application number: PCT/JP2010/073604
Authority: WO
Inventors: 中條　健; 山影　朋夫; 太一郎塩寺
Original assignee: 株式会社東芝
Priority date: 2010-07-02
Filing date: 2010-12-27
Publication date: 2012-01-05

Abstract

A moving image encoding apparatus according to an embodiment of the invention derives scaling information on the basis of the maximum and minimum values of a target pixel group in a local decoded image. A scaling that scales down the pixel bit length is applied to the target pixel group in accordance with the scaling information. A scaling unit limits the values of particular pixels, which are to be scaled, with respect to particular values, thereby generating a reference pixel group as scaled. A fixed bit length is used to express the description of first scaling information in a case of including the particular values or the description of second scaling information in a case of excluding the particular values and to express the reference pixel group as scaled in accordance with the corresponding scaling information. A reference image is reconstructed by use of inverse scaling and a predicted image is generated. Information indicating the difference between an input image and the predicted image is encoded by an encoding unit.

Description

Moving picture coding apparatus, moving picture decoding apparatus and method

The embodiment relates to encoding and decoding of moving images.

H. is one of the international standards for video coding. H.264 / MPEG-4 AVC was jointly established by ITU-T (International Telecommunication Union-Telecommunication Standardization Sector) and ISO (International Organization for Standardization) / IEC (International Electrotechnical Commission). H. Video coding standards such as H.264 / MPEG-4 AVC usually store local decoded images (encoding side) of already encoded images or decoded images (decoding side) in an image buffer. And has a mechanism to refer to generate a predicted image.

Since the image buffer stores a large number of reference images, a main memory having a large storage capacity is required on both the encoding side and the decoding side. In addition, since a large amount of access to the image buffer occurs to generate a predicted image, a wide memory bandwidth is required on both the encoding side and the decoding side. Such a problem of hardware requirements regarding the image buffer becomes more prominent as the pixel bit length increases.

On the other hand, various internal processes on the encoding side and the decoding side such as a motion estimation process, a predicted image generation process (prediction process), and a filter process (for example, loop filter process) usually increase as the pixel bit length increases. Easy to achieve with high accuracy. Therefore, increasing the pixel bit length helps improve the coding efficiency.

* Enhancing coding efficiency can be expected by applying a large pixel bit length to various internal processes including prediction processing. On the other hand, by applying a small pixel bit length to the image buffer, it is possible to expect a reduction in hardware requirements related to the image buffer.

Therefore, the embodiment aims to apply a larger pixel bit length to various internal processes including a prediction process while applying a smaller pixel bit length to the image buffer.

The moving image encoding apparatus according to the embodiment derives scaling information based on the maximum value and the minimum value of the target pixel group in the locally decoded image. Scaling for reducing the pixel bit length is applied to the target pixel group according to the scaling information. The scaling processing unit generates a scaled reference pixel group by limiting the value of the specific pixel to be scaled with respect to the specific value. The description of the first scaling information when the specific value is included or the description of the second scaling information when the specific value is not included and the reference pixel group scaled according to the corresponding scaling information are fixed. It is expressed with the bit length. The reference image is restored by inverse scaling, and a predicted image is generated. Information indicating the difference between the input image and the predicted image is encoded by the encoding unit.

1 is a block diagram showing a moving image encoding apparatus according to a first embodiment. The block diagram which shows the moving image decoding apparatus which concerns on 1st Embodiment. The block diagram which shows the loop filter part of FIG. Explanatory drawing of an object pixel group. FIG. 4 is a flowchart showing an operation of a filter processing / scaling processing unit in FIG. 3. FIG. 4 is a flowchart showing the operation of the scaling processing unit in FIG. 3. The block diagram which shows the estimation part of FIG. 8 is a flowchart showing the operation of the inverse scaling processing unit of FIG. The block diagram which shows the moving image encoder which concerns on 2nd Embodiment. The block diagram which shows the moving image decoding apparatus which concerns on 2nd Embodiment. The table figure which shows the example of the dynamic range Dr, EncTable [Dr], and Offset [Dr] concerning 3rd Embodiment. The table figure which shows the example of the dynamic range Dr, EncTable [Dr], and Offset [Dr] concerning 3rd Embodiment. The table figure which shows the example of the dynamic range Dr and EncTable [Dr] concerning 3rd Embodiment. The table figure which shows the example of the dynamic range Dr, EncTable [Dr], and Offset [Dr] concerning 3rd Embodiment. The table figure which shows the example of the dynamic range Dr, EncTable [Dr], and Offset [Dr] concerning 3rd Embodiment. The block diagram which shows the moving image encoder which concerns on 4th Embodiment. The block diagram which shows the moving image decoding apparatus which concerns on 4th Embodiment. The block diagram which shows the loop filter part which concerns on 4th Embodiment. The figure which shows the example of the format in the reference image buffer which concerns on 4th Embodiment. The block diagram which shows the moving image encoder which concerns on 5th Embodiment. The block diagram which shows the moving image decoding apparatus which concerns on 5th Embodiment. 10 is a flowchart showing an operation of a pixel accuracy control unit according to the fifth embodiment. The figure which shows an example of the reference image buffer part which concerns on 5th Embodiment. The figure which shows another example of the reference image buffer part which concerns on 5th Embodiment.

Hereinafter, embodiments will be described with reference to the drawings. In the following description, the term “image” can be appropriately replaced with terms such as “image signal” and “image data”.
(First embodiment)
As illustrated in FIG. 1, the moving image encoding apparatus according to the first embodiment includes an encoding unit 100 and an encoding control unit 140. The encoding unit 100 encodes the input image 11 to generate encoded data. The encoding control unit 140 controls various elements in the encoding unit 100. For example, the encoding control unit 140 controls a loop filter setting unit 106, a prediction unit 120, and the like which will be described later.

The encoding unit 100 includes a subtraction unit 101, a transform / quantization unit 102, an entropy encoding unit 103, an inverse quantization / inverse transform unit 104, an addition unit 105, a loop filter setting unit 106, a reference image buffer unit 107, scaling information. A buffer unit 108, a loop filter unit 110, a prediction unit 120, and a motion vector generation unit 130 are included.

The subtraction unit 101 subtracts the prediction image from the prediction unit 120 from the input image 11 to obtain a prediction error. The transform / quantization unit 102 performs transform (for example, discrete cosine transform (DCT)) and quantization on the prediction error from the subtraction unit 101, and quantizes information on transform coefficients (hereinafter simply referred to as quantum). (Referred to as conversion coefficient).

The entropy coding unit 103 performs entropy coding on the quantized transform coefficient from the transform / quantization unit 102, the loop filter information 13 from the loop filter setting unit 106, and the motion vector information from the motion vector generation unit 130. Do. The entropy encoding unit 103 may further entropy encode information other than these (for example, prediction mode information). The type of entropy encoding is, for example, variable length encoding or arithmetic encoding. The entropy encoding unit 103 outputs encoded data obtained by entropy encoding to the outside.

The inverse quantization / inverse transform unit 104 performs inverse quantization and inverse transform (for example, inverse discrete cosine transform (IDCT), etc.) on the quantized transform coefficient from the transform / quantization unit 102, thereby reducing a prediction error. Restore. The addition unit 105 adds the prediction error restored by the inverse quantization / inverse conversion unit 104 and the corresponding prediction image from the prediction unit 120 to generate the local decoded image 12.

The loop filter setting unit 106 sets the loop filter information 13 based on the input image 11 and the corresponding local decoded image 12 from the addition unit 105 and notifies the loop filter unit 110 and the entropy encoding unit 103. The loop filter information 13 includes at least filter coefficient information and filter switching information. The filter coefficient information includes information indicating the filter coefficient. The filter coefficient information may further include information indicating an offset coefficient described later. The filter switching information includes information indicating validity / invalidity of filter application.

The loop filter unit 110 performs a filtering process or a bypass process that does not go through the filtering process on the target pixel group in the local decoded image 12 from the adding unit 105 according to the loop filter information 13 from the loop filter setting unit 106. Then, the loop filter unit 110 reduces (or maintains) the pixel bit length by performing a scaling process described later on the filter processing result group or the bypass processing result group (that is, the target pixel group itself). The loop filter unit 110 supplies the scaling processing result group to the reference image buffer unit 107 as the scaled reference pixel group 14. Further, the loop filter unit 110 supplies the scaling information 15 regarding the scaling process to the scaling information buffer unit 108. Details of the loop filter unit 110 will be described later.

The reference image buffer unit 107 stores the scaled reference pixel group 14 from the loop filter unit 110. The scaling information buffer unit 108 accumulates scaling information 15 corresponding to the scaled reference pixel group 14 in synchronization with the reference image buffer unit 107. The reference pixel group 14 accumulated in the reference image buffer unit 107 and the scaling information 15 accumulated in the scaling information buffer unit 108 are read by the prediction unit 120 or the motion vector generation unit 130 as necessary.

The motion vector generation unit 130 reads the scaled reference pixel group and the scaling information from the reference image buffer unit 107 and the scaling information buffer unit 108 as necessary. The motion vector generation unit 130 applies inverse scaling that extends (or maintains) the pixel bit length to the scaled reference pixel group according to the scaling information to restore the reference image. The motion vector generation unit 130 generates motion vector information based on the input image 11 and the restored reference image. The motion vector generation unit 130 notifies the prediction unit 120 and the entropy encoding unit 103 of the motion vector information. Details of the inverse scaling process will be described later.

The prediction unit 120 reads the scaled reference pixel group and the scaling information from the reference image buffer unit 107 and the scaling information buffer unit 108 as necessary. The prediction unit 120 restores the reference image by applying inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information. The prediction unit 120 generates a prediction image based on the motion vector information from the motion vector generation unit 130 and the restored reference image. The prediction unit 120 supplies the predicted image to the subtraction unit 101 and the addition unit 105.

The details of the loop filter unit 110 will be described below with reference to FIG.
The loop filter unit 110 includes a switch 111, a filter processing / scaling processing unit 112, and a scaling processing unit 113. Note that FIG. 3 only illustrates the loop filter unit 110. For example, the loop filter unit 110 may include one or a plurality of filter processing / scaling processing units (not shown) different from the filter processing / scaling processing unit 112. The number of switches 111 selected can be changed to 3 or more according to the configuration of the loop filter unit 110.

The switch 111 selects the output destination of the target pixel group included in the local decoded image 12 according to the loop filter information 13. As a simple example, the switch 111 guides the target pixel group to the filter processing / scaling processing unit 112 if the loop filter information 13 indicates that the filter application of the target pixel group is valid. On the other hand, if the loop filter information 13 indicates that the filter application of the target pixel group is invalid, the switch 111 guides the target pixel group to the scaling processing unit 113. For example, as shown in FIG. 4, in the locally decoded image 12, the filter application is valid (On) / invalid (Off) for each pixel group (for example, a block) of variable (may be fixed) size. Filter switching information indicating () is set. These pixel groups are all shown as rectangles in FIG. 4, but their shapes may be changed depending on the design. The filter switching information of each pixel group is set by the filter setting unit 106 described above and can be referred to via the loop filter information 13.

The filter processing / scaling processing unit 112 performs filter processing on the target pixel group according to the loop filter information 13. Then, the filter processing / scaling processing unit 112 derives scaling information 15 based on the distribution (for example, dynamic range) of the filter processing result group, and performs scaling information for reducing the pixel bit length with respect to the filter processing result group. 15 is applied to generate a scaled reference pixel group 14. Details of the operation of the filter processing / scaling processing unit 112 will be described later.

The scaling processing unit 113 derives scaling information 15 based on the distribution (for example, dynamic range) of the target pixel group, and is scaled by applying scaling for reducing the pixel bit length to the target pixel group according to the scaling information 15. The reference pixel group 14 is generated. Details of the operation of the scaling processing unit 113 will be described later.

Details of the operation of the filter processing / scaling processing unit 112 will be described below with reference to FIG.
The filter processing / scaling processing unit 112 performs a convolution operation (filter operation) on the target pixel group in accordance with the filter coefficient information included in the loop filter information 13 (step S112-1). Specifically, when the filter coefficient is represented by F [n], the pixel value of the target pixel group is represented by P [m], and the convolution calculation result is represented by B [m], the filter processing / scaling processing unit 112 represents the following formula ( Perform the convolution operation according to 1).

In Equation (1), O represents an offset coefficient. The offset coefficient can be referred to through the filter coefficient information. The sum of the filter coefficient F [n] is assumed to be designed to be substantially equal to the 2 ^K. Further, the pixel bit length of the target pixel group is assumed to be T bits. The filter processing / scaling processing unit 112 performs such a convolution operation on each pixel of the target pixel group to obtain a convolution operation result group.

Next, the filter processing / scaling processing unit 112 searches for the maximum value Max and the minimum value Min of the convolution calculation result group of the target pixel group (step S112-2). However, the upper limit of the maximum value Max is 2 ^{K + T} −1. That is, if the value of the maximum value Max exceeds 2 ^{K + T} −1, the maximum value Max is handled as 2 ^{K + T} −1.

Next, the filter processing / scaling processing unit 112 derives the scaling information 15 of the target pixel group (step S112-3). Specifically, the filtering / scaling processing unit 112 derives the minimum reference value MinPoint by arithmetically shifting the minimum value Min by S bits to the right according to the following equation (2).

Here, S is represented by the following formula (3).

Here, L represents a pixel bit length applied to the reference image buffer unit 107. It is assumed that the pixel bit length L of the reference image buffer unit 107 is equal to or less than the pixel bit length T of the target pixel group. That is, the minimum reference value MinPoint is a value obtained by rounding the minimum value Min to L bits.

The filter processing / scaling processing unit 112 derives the scaling amount Q by executing the following calculation (4). The operation (4) is described according to the C language, but an operation having the same content can be described according to other programming languages.

The scaling amount Q can take any integer value from 0 to S. The minimum reference value MinPoint and the scaling amount Q derived as described above are supplied to the scaling information buffer unit 108 as scaling information 15. An example of a method for efficiently describing the scaling information 15 will be described. The scaling information 15 has 1-bit flag information indicating whether Q is equal to S or not. When Q and S are not equal (that is, the scaling flag information is OFF), it further has a value of scaling amount Q (1 or more and S or less) and a value of minimum reference value MinPoint. If Q is equal to S (ie, the scaling flag information is ON), the minimum reference value MinPoint is interpreted as 0.

Further, when the overhead of the scaling information amount 15 becomes a problem, there is a method for obtaining the scaling amount Q by the following calculation (5) obtained by modifying the calculation (4).

Operation (5) is a process of scaling the T-bit target pixel group to L bits when Q is equal to S. On the other hand, the operation (5) is a process of scaling the T-bit target pixel group to L−1 bits when Q and S are not equal. Since the reference pixel group is L−1 bits, it is possible to cancel the increase in the scaling information 15.

In the present embodiment, the minimum reference value MinPoint can be replaced with a maximum reference value MaxPoint based on the maximum value Max. In order to realize such replacement, various formulas and calculations in the present embodiment may be appropriately read.

In addition, in order to refer to the value of the parameter K in the inverse scaling process, filter switching information or information similar thereto (for example, the value of the parameter K itself) is required. May be included. Alternatively, since such information can be referred to via the loop filter information 13, the loop filter information 13 may be notified to each element that performs inverse scaling processing. In the following description, it is assumed that the filter switching information is included in the scaling information 15. The unit for performing the scaling process and the inverse scaling process only needs to be a common unit on the encoding side and the decoding side. In the present embodiment, the unit of scaling processing and filtering processing corresponds to the unit in which filter switching information is set. The unit of the scaling process and the filtering process may be the unit of a plurality of target pixel groups that are the same as or smaller than the unit of the target pixel group. For example, in the case of a variable block as shown in FIG. 4, the smallest block size in processing units may be used as the unit of all target pixel groups.

Next, the filter processing / scaling processing unit 112 applies scaling to each convolution calculation result according to the derived filter information 15 (step S112-4). Specifically, the filter processing / scaling processing unit 112 generates each pixel value D [m] of the scaled reference pixel group according to the following formula (6).

Clip1 (x) represents a clipping function that rounds x to a value between 0 and 2 ^L −1. The offset in the equation (6) is obtained by the following calculation (7) using the conditional operator (ternary operator) “?:”.

The scaled reference pixel group generated as described above is supplied to the reference image buffer unit 107. By the operation of the filter processing / scaling processing unit 112, the pixel bit length (= approximately (K + T) ≧ L) of the filter processing result group of the target pixel group is reduced to L bits after the filter processing.

Details of the operation of the scaling processing unit 113 will be described below with reference to FIG.
The scaling processing unit 113 searches for the maximum value Max and the minimum value Min of the target pixel group (step S113-1). Next, the scaling processing unit 113 derives the scaling information 15 of the target pixel group (step S113-2). Specifically, the scaling processing unit 113 derives the minimum reference value MinPoint by arithmetically shifting the minimum value Min by S bits to the right according to Equation (2). However, in the operation of the scaling processing unit 113, K = 0 is handled. That is, S is derived according to the following formula (8).

Also, the scaling processing unit 113 derives the scaling amount Q by executing the calculation (4) or the calculation (5). The minimum reference value MinPoint and the scaling amount Q derived as described above are supplied to the scaling information buffer unit 108 as scaling information 15.

Next, the scaling processing unit 113 applies scaling to each pixel value of the target pixel group in accordance with the derived filter information 15 (step S113-3). Specifically, the scaling processing unit 113 generates each pixel value D [m] of the scaled reference pixel group according to the following formula (9).

Offset in formula (9) is obtained by calculation (7). The scaled reference pixel group generated as described above is supplied to the reference image buffer unit 107. By the operation of the scaling processing unit 113, the pixel bit length (= T ≧ L) of the target pixel group is reduced to L bits.

Hereinafter, details of the inverse scaling process in the prediction unit 120 will be described as an example. Note that the inverse scaling process substantially the same as or similar to that of the prediction unit 120 is also performed in the motion vector generation unit 130. As illustrated in FIG. 7, the prediction unit 120 includes an inverse scaling processing unit 121 and a predicted image generation unit 122. The inverse scaling processing unit 121 restores the reference image by applying inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information. The predicted image generation unit 220 generates a predicted image based on the motion vector information and the restored reference image.

Details of the operation of the inverse scaling processing unit 121 will be described below with reference to FIG.
The inverse scaling processing unit 121 obtains a desired reference pixel group (that is, necessary for generating a predicted image) and corresponding scaling information from the reference image buffer unit 107 and the scaling information buffer unit 108, respectively (step S121). -1). Specifically, the inverse scaling processing unit 121 acquires each pixel value D [m] of the scaled reference pixel group, the minimum reference value MinPoint, the scaling amount Q, and the filter switching information. The inverse scaling processing unit 121 refers to the filter switching information, sets a predetermined value corresponding to the filter processing to the parameter K if the filter application is valid, and sets 0 to the parameter K if the filter application is invalid.

Next, the inverse scaling processing unit 121 applies inverse scaling for extending the pixel bit length to the reference pixel group according to the scaling information (step S121-2). Specifically, if QKT−U ≧ 0 is satisfied with the pixel bit length after the inverse scaling process as U bits, the inverse scaling processing unit 121 applies inverse scaling according to the following equation (10).

On the other hand, if Q−K−T + U ≧ 0 does not hold, the inverse scaling processing unit 121 applies inverse scaling according to the following equation (11).

Here, offset2 is calculated by the following calculation (12).

G [m] represents each pixel value of the restored reference pixel group. By the operation of the inverse scaling processing unit 121, the pixel bit length (= L ≦ U) of the scaled reference pixel group is expanded to U bits.

As described above, in this embodiment, adaptive scaling / inverse scaling processing based on the distribution of the target pixel group is performed. By performing such scaling / inverse scaling processing, the value B [m] or P [m] before scaling is rounded to L bits, and the value G [m] after inverse scaling is rounded to L It is guaranteed to be the same value as the value rounded to bits. Compared to the method of generating the value G [m] after inverse scaling by rounding to L bits by rounding B [m] or P [m] to L bits fixedly, the inverse obtained by the processing of this embodiment The value G [m] after scaling has high accuracy.

As described above, the video encoding apparatus according to the first embodiment performs the scaling process and the inverse scaling process before and after the reference image buffer unit, respectively, so that the pixel bit length applied to the reference image buffer unit Is smaller than the pixel bit length applied to other internal processing (prediction processing, filter processing, etc.). Therefore, according to the video encoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing a highly accurate prediction process, filter process, and the like by applying a larger pixel bit length. Can be kept small.

As shown in FIG. 2, the moving picture decoding apparatus according to the present embodiment includes a decoding unit 200 and a decoding control unit 240. The decoding unit 200 generates the output image 26 by decoding the encoded data. The decoding control unit 240 controls various elements in the decoding unit 200. For example, the decoding control unit 240 controls the prediction unit 220 described later. Note that the scaling process and the inverse scaling process in the moving picture decoding apparatus in FIG. 2 are substantially the same as or similar to the scaling process and the inverse scaling process in the moving picture encoding apparatus in FIG.

The decoding unit 200 includes an entropy decoding unit 201, an inverse quantization / inverse conversion unit 202, an addition unit 203, a loop filter unit 210, a reference image buffer unit 204, a scaling information buffer unit 205, a prediction unit 220, and a bit length normalization. Part 230.

The entropy decoding unit 201 performs entropy decoding according to syntax information on encoded data generated by, for example, the moving image encoding apparatus in FIG. The entropy decoding unit 201 supplies the decoded quantized transform coefficient to the inverse quantization / inverse transform unit 202, supplies the decoded motion vector information to the prediction unit 220, and decodes the decoded loop filter information. 23 is supplied to the loop filter unit 210.

The inverse quantization / inverse transformation unit 202 and the addition unit 203 are substantially the same or similar elements as the inverse quantization / inverse transformation unit 104 and the addition unit 105 described above. That is, the inverse quantization / inverse transform unit 202 performs inverse quantization and inverse transform (for example, inverse discrete cosine transform (IDCT), etc.) on the quantized transform coefficient from the entropy decoding unit 201 to generate a prediction error. To restore. The adding unit 203 adds the prediction error restored by the inverse quantization / inverse transform unit 202 and the corresponding prediction image from the prediction unit 220 to generate a decoded image 22.

The loop filter unit 210 is substantially the same as or similar to the loop filter unit 110 described above. In other words, the loop filter unit 210 performs a filtering process or a bypass process that does not go through the filtering process on the target pixel group in the decoded image 22 from the adding unit 203 according to the loop filter information 23 from the entropy decoding unit 201. . Then, the loop filter unit 210 performs the above-described scaling processing on the filter processing result group or the bypass processing result group (that is, the target pixel group) to reduce the pixel bit length. The loop filter unit 210 supplies the scaling processing result group to the reference image buffer unit 204 as the scaled reference pixel group 24. In addition, the loop filter unit 210 supplies the scaling information 25 regarding the scaling processing to the scaling information buffer unit 205. The loop filter unit 210 is substantially the same as or similar to the loop filter unit 110 described above, and a detailed description thereof will be omitted.

The reference image buffer unit 204 stores the scaled reference pixel group 24 from the loop filter unit 210. The scaling information buffer unit 205 accumulates scaling information 25 corresponding to the scaled reference pixel group 24 while synchronizing with the reference image buffer unit 204. The reference pixel group 24 accumulated in the reference image buffer unit 204 and the scaling information 25 accumulated in the scaling information buffer unit 205 are read by the prediction unit 220 or the bit length normalization unit 230 as necessary. For example, in order to generate the output image 26, the bit length normalization unit 230 reads out a desired reference pixel group (that is, necessary for generating the output image 26) and corresponding scaling information according to the display order.

The prediction unit 220 is substantially the same or similar element as the prediction unit 120 described above. That is, the prediction unit 220 reads the scaled reference pixel group and scaling information from the reference image buffer unit 204 and the scaling information buffer unit 205, respectively, as necessary. The prediction unit 220 restores the reference image by applying inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information. The prediction unit 220 generates a prediction image based on the motion vector information from the entropy decoding unit 201 and the restored reference image. The prediction unit 220 supplies the predicted image to the addition unit 203.

The bit length normalization unit 230 reads the scaled reference pixel group and scaling information from the reference image buffer unit 204 and the scaling information buffer unit 205, respectively, as necessary. The bit length normalization unit 230 applies inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information, and obtains a desired pixel bit length U (where the bit length normalization unit 230 The pixel bit length U related to the operation of (2) does not necessarily match the pixel bit length U related to the operation of the prediction unit 220). The bit length normalization unit 230 supplies the output image 26 to the outside. Note that the pixel bit length of the output image 26 may be, for example, the same as or different from the pixel bit length of the input image 11 in the moving image apparatus of FIG. When the decoded image 22 is used as it is as the output image 26 (without passing through the loop filter unit 210 and without changing the pixel bit length), the bit length normalization unit 230 can be removed. It is.

As described above, the moving picture decoding apparatus according to the first embodiment performs the scaling process and the inverse scaling process before and after the reference image buffer unit, so that the pixel bit length applied to the reference image buffer unit Is smaller than the pixel bit length applied to other internal processing (prediction processing, filter processing, etc.). Therefore, according to the video decoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing a highly accurate prediction process, filter process, and the like by applying a larger pixel bit length. Can be kept small.

It should be noted that a plurality of scaling processes in the loop filter unit 110 can be realized in a common scaling processing unit. The common scaling processing unit sets a parameter K according to filter switching information regarding each target pixel group, and applies scaling to the target pixel group or the filter processing result group.

It is also possible to implement the inverse scaling processing in the prediction unit 120 and the motion vector generation unit 130 in a common inverse scaling processing unit. Similarly, the inverse scaling processing in the prediction unit 220 and the bit length normalization unit 230 can be realized in a common inverse scaling processing unit. These common inverse scaling processing units set a pixel bit length U according to the output destination, and apply inverse scaling to the scaled reference pixel group.

(Second Embodiment)
As illustrated in FIG. 9, the moving image encoding apparatus according to the second embodiment includes an encoding unit 300 and an encoding control unit 140. 9, parts that are the same as those in FIG. 1 are given the same reference numerals, and in the following description, different parts between FIG. 9 and FIG. 1 will be mainly described. For example, the scaling process and the inverse scaling process in the moving picture decoding apparatus in FIG. 9 are substantially the same as or similar to the scaling process and the inverse scaling process in the moving picture encoding apparatus in FIG.

The encoding unit 300 encodes the input image 11 to generate encoded data. The encoding unit 300 includes a bit length extension unit 309, a subtraction unit 101, a transform / quantization unit 102, an entropy encoding unit 303, an inverse quantization / inverse transform unit 104, an addition unit 105, a loop filter setting unit 106, a loop filter Unit 110, reference image buffer unit 107, scaling information buffer unit 108, prediction unit 120, and motion vector generation unit 130.

The bit length extension unit 309 extends the pixel bit length of the input image 11 and supplies it to the subtraction unit 101, the loop filter setting unit 106, and the motion vector generation unit 130. As a result of the operation of the bit length extension unit 309, for example, the pixel bit length applied to the internal processing by the loop filter unit 110, the prediction unit 120, and the like is larger than the original pixel bit length of the input image 11. The bit length extension unit 309 notifies the internal bit length information 37 to the entropy encoding unit 303. The internal bit length information 37 may be information indicating the extension amount of the pixel bit length by the bit length extension unit 309, for example, or the extension amount of the pixel bit length is determined in advance between the encoding side and the decoding side. If it is, information (for example, a flag) indicating whether the pixel bit length extension is valid / invalid may be used. On the other hand, if the extension of the pixel bit length on the encoding side is performed at a timing known to the decoding side (for example, always or when a predetermined condition is satisfied) and the extension amount is also known to the decoding side, the bit length The extension unit 309 may not notify the entropy encoding unit 303 of the internal bit length information 37.

The entropy encoding unit 303 further performs entropy encoding of the internal bit length information 37 as necessary, in addition to the operation of the entropy encoding unit 103 described above. When the entropy encoding unit 303 performs entropy encoding of the internal bit length information 37, the encoded internal bit length information is also included in the encoded data.

As described above, the moving image encoding apparatus according to the second embodiment employs the same scaling process / inverse scaling process as the first embodiment while extending the internal pixel bit length compared to the input image. is doing. Therefore, according to the video encoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing further highly accurate prediction processing, filter processing, and the like by extending the internal pixel bit length. Can be kept small.

As illustrated in FIG. 10, the moving picture decoding apparatus according to the second embodiment includes a decoding unit 400 and a decoding control unit 240. 10, parts that are the same as those in FIG. 2 are given the same reference numerals, and in the following description, different parts between FIG. 10 and FIG. 2 will be mainly described. For example, the scaling process and the inverse scaling process in the moving picture decoding apparatus in FIG. 10 are substantially the same as or similar to the scaling process and the inverse scaling process in the moving picture encoding apparatus in FIG.

The decoding unit 400 generates the output image 26 by decoding the encoded data. The decoding unit 400 includes an entropy decoding unit 401, an inverse quantization / inverse conversion unit 202, an addition unit 203, a loop filter unit 210, a reference image buffer unit 204, a scaling information buffer unit 205, a prediction unit 220, and a bit length normalization. Part 430.

The entropy decoding unit 401 further performs entropy decoding of the encoded internal bit length information as necessary in addition to the operation of the entropy decoding unit 201 described above. When performing entropy decoding of the encoded internal bit length information, the entropy decoding unit 401 notifies the bit length normalization unit 430 of the decoded internal bit length information 47.

When the internal bit length information 47 is notified, the bit length normalization unit 430 performs the same operation as the above-described bit length normalization unit 230 while considering the internal bit length information 47 as necessary. Thus, the output image 26 normalized to a desired pixel bit length is generated. As an example, the bit length normalization unit 430 checks the pixel bit length of the input image on the encoding side by referring to the internal bit length information 47, and outputs the output image 26 normalized to the pixel bit length of the input image. Is generated.

As described above, the moving picture decoding apparatus according to the second embodiment expands the internal pixel bit length as compared with the input image on the encoding side, and performs the same scaling process / inverse as in the first embodiment. Scaling processing is adopted. Therefore, according to the video decoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing further highly accurate prediction processing, filter processing, and the like by extending the internal pixel bit length. Can be kept small.

(Third embodiment)
The video encoding apparatus according to the third embodiment has substantially similar elements to the video encoding apparatus according to the first embodiment described above, but differs in the details of the scaling process / inverse scaling process. In the following description, details of the scaling process / inverse scaling process according to the present embodiment will be described with reference to FIGS. 5, 6, and 8.

Details of the operation of the filter processing / scaling processing unit 112 will be described below with reference to FIG.
As in the first embodiment, the filter processing / scaling processing unit 112 performs a convolution operation (filter operation) on the target pixel group according to the filter coefficient information included in the loop filter information 13 (step S112-1). ). Next, the filter processing / scaling processing unit 112 searches for the maximum value Max and the minimum value Min of the convolution calculation result group of the target pixel group as in the first embodiment (step S112-2).

Next, the filter processing / scaling processing unit 112 derives the scaling information 15 of the target pixel group (step S112-3). Specifically, the filter processing / scaling processing unit 112 derives the minimum reference value MinPoint according to Equation (2) and Equation (3). Further, the filter processing / scaling processing unit 112 derives the maximum reference value MaxPoint according to the following calculation (13) using the conditional operator (ternary operator) “?:”.

However, the upper limit of the value of MaxPoint is 2 ^L −1.

The minimum reference value MinPoint and the maximum reference value MaxPoint derived as described above are supplied to the scaling information buffer unit 108 as scaling information 15. In addition, in order to refer to the value of the parameter K in the inverse scaling process, filter switching information or information similar thereto is required, and such information may also be included in the scaling information 15. Alternatively, since such information can be referred to via the loop filter information 13, the loop filter information 13 may be notified to an element that performs inverse scaling processing. In the following description, it is assumed that the filter switching information is included in the scaling information 15.

Next, the filter processing / scaling processing unit 112 applies scaling to each convolution calculation result according to the derived filter information 15 (step S112-4). Specifically, the filter processing / scaling processing unit 112 generates each pixel value D [m] of the scaled reference pixel group according to the following formula (14).

EncTable [Dr] is obtained by the following equation (15).

Dr represents the dynamic range of the scaled reference pixel group as shown in the following equation (16). Offset [Dr] represents a rounding offset value. N represents the bit length of EncTable [Dr]. In addition, since Formula (14) includes division, it is assumed that the value of EncTable [Dr] is calculated in advance for each value of Dr and stored in a table format, for example.

The values of EncTable [Dr] and Offset [Dr] corresponding to the dynamic range Dr fall within the same number of bits X (X <(K + T) and X <U) between B [m] and G [m] described later. When doing so, the values of B [m] and G [m] after rounding may be set to be equal.

FIG. 11 shows, as an example, values of EncTable [Dr] and Offset [Dr] corresponding to Dr when K + T = 12, L = 8, N = 8, and X = 8. Note that ExpandFlag in FIGS. 11 and 12 to 15 indicates whether or not the scaling process / inverse scaling process described in this embodiment is performed. When ExpandFlag is 0, it indicates that the scaling process / inverse scaling process described in the first or second embodiment is performed. In this case, the scaling amount Q is derived using the calculation (4), and each pixel value D [m] of the reference pixel group scaled using the equation (6) is generated. On the other hand, when ExpandFlag is 1, it indicates that the scaling process / inverse scaling process described in this embodiment is performed. As another example, when K + T = 14, L = 8, N = 8, and X = 8, the values of EncTable [Dr] corresponding to Dr are shown in FIG. 12, K + T = 10, L = 8, When N = 8 and X = 8, the values of EncTable [Dr] corresponding to Dr correspond to Dr when K + T = 12, L = 8, N = 12, and X = 8 in FIG. The values of EncTable [Dr] and Offset [Dr] are shown in FIG.

Furthermore, MaxPoint and MinPoint may be expressed by (K + T) bits. That is, the upper limit of the values of MaxPoint and MinPoint is 2 ^{K + T} −1. Therefore, Dr calculated by Expression (16) is expressed by (K + T) bits. FIG. 15 shows the values of EncTable [Dr] and Offset [Dr] corresponding to Dr represented by (K + T) bits when K + T = 12, L = 8, N = 12, and X = 8. Has been.

Note that the value of Offset [Dr] may be fixed to 1 << (K + T-1). In this case, in FIG. 11 to FIG. 15, EncTable [Dr] where Offset [Dr] is different from 1 << (K + T−1) is not used. In this case, Dr that is not used is integrated with the contents one line below. As an example, in FIG. 11, when Dr = 23 is not used, it is integrated with Dr = 24, which is one row below. That is, when Dr = 23, 24, EncTable [Dr] is set to 2557.

Details of the operation of the scaling processing unit 113 will be described below with reference to FIG.
The scaling processing unit 113 searches for the maximum value Max and the minimum value Min of the target pixel group as in the first embodiment (step S113-1). Next, the scaling processing unit 113 derives the scaling information 15 of the target pixel group (step S113-2). Specifically, the filter processing / scaling processing unit 112 derives the minimum reference value MinPoint according to Equation (2). Further, the filter processing / scaling processing unit 112 derives the maximum reference value MaxPoint according to Equation (13). However, in the operation of the scaling processing unit 113, K = 0 is handled. That is, S is derived according to Equation (8). The minimum reference value MinPoint and the maximum reference value MaxPoint derived as described above are supplied to the scaling information buffer unit 108 as scaling information 15.

Next, the scaling processing unit 113 applies scaling to each pixel value of the target pixel group in accordance with the derived filter information 15 (step S113-3). Specifically, the scaling processing unit 113 generates each pixel value D [m] of the scaled reference pixel group according to the following formula (17).

By the operation of the scaling processing unit 113, the pixel bit length (= T ≧ L) of the target pixel group is reduced to L bits.
Note that the value of EncTable [Dr] corresponding to the dynamic range Dr is the same number of bits X (X <T and X as P [m] and G [m] described later, as in the filter processing / scaling processing unit 112. <U), the values of B [m] and G [m] after rounding are set to be equal. FIG. 11 shows, as an example, the value of EncTable [Dr] corresponding to Dr when K (= 0) + T = 12, L = 8, N = 8, and X = 8. As another example, the values of EncTable [Dr] corresponding to Dr when T = 14, L = 8, N = 8, and X = 8 are shown in FIG. 12, T = 10, L = 8, N = 8, the value of EncTable [Dr] corresponding to Dr when X = 8 is shown in FIG. 13, and EncTable [Dr corresponding to Dr when T = 12, L = 8, N = 12, and X = 8. ] Are shown in FIG.

Furthermore, MaxPoint and MinPoint may be expressed by T bits. That is, the upper limit of the values of MaxPoint and MinPoint is 2 ^T −1. Therefore, Dr calculated by Expression (16) is expressed by T bits. The values of EncTable [Dr] and Offset [Dr] corresponding to Dr represented by T bits when T = 12, L = 8, N = 8, and X = 8 are shown as an example in FIG. Yes.

Details of the operation of the inverse scaling processing unit 121 will be described below with reference to FIG.
The inverse scaling processing unit 121 obtains a desired reference pixel group (that is, necessary for generating a predicted image) and corresponding scaling information from the reference image buffer unit 107 and the scaling information buffer unit 108, respectively (step S121). -1). Specifically, the inverse scaling processing unit 121 acquires each pixel value D [m] of the scaled reference pixel group, the minimum reference value MinPoint, the maximum reference value MaxPoint, and filter switching information. The inverse scaling processing unit 121 refers to the filter switching information, sets a predetermined value corresponding to the filter processing to the parameter K if the filter application is valid, and sets 0 to the parameter K if the filter application is invalid.

Next, the inverse scaling processing unit 121 applies inverse scaling for extending the pixel bit length to the reference pixel group according to the scaling information (step S121-2). Specifically, if the relationship of K + T + L ≧ U is established, the inverse scaling processing unit 121 applies inverse scaling according to the following equation (18).

The pixel bit length (= L ≦ U) of the scaled reference pixel group is expanded to U bits by the operation of the inverse scaling processing unit 121.

Here, the value of DecTable [Dr] corresponding to the dynamic range Dr is set to correspond to Dr used in the filter processing / scaling processing unit 112 or the scaling processing unit 113. FIG. 11 shows the value of DecTable [Dr] corresponding to Dr when K + T = 12, L = 8, N = 8, and X = 8 as an example. As another example, the values of DecTable [Dr] corresponding to Dr when K + T = 14, L = 8, N = 8, and X = 8 are shown in FIG. 12, and K + T = 10, L = 8, N = 8, the value of DecTable [Dr] corresponding to Dr when X = 8 is shown in FIG. 13, and DecTable [Dr corresponding to Dr when K + T = 12, L = 8, N = 12, and X = 8. ] Are shown in FIG.

Furthermore, MaxPoint and MinPoint may be expressed by (K + T) bits. That is, the upper limit of the values of MaxPoint and MinPoint is 2 ^{K + T} −1. Therefore, Dr calculated by Expression (16) is expressed by (K + T) bits. FIG. 15 shows the values of EncTable [Dr] and Offset [Dr] corresponding to Dr represented by (K + T) bits when K + T = 12, L = 8, N = 8, and X = 8. Has been.

As described above, in this embodiment, adaptive scaling / inverse scaling processing based on the distribution of the target pixel group is performed. By performing such scaling / inverse scaling processing, the value B [m] or P [m] before scaling is rounded to L bits, and the value G [m] after inverse scaling is rounded to L Guaranteed to be the same value rounded to bits. Compared to the method of generating a value G [m] after inverse scaling by rounding B [m] or P [m] to L bits by fixed rounding, the value after inverse scaling obtained by the processing of this embodiment is used. The value G [m] has high accuracy.

As described above, the moving picture encoding apparatus according to the third embodiment performs the scaling process and the inverse scaling process before and after the reference image buffer unit, so that the pixel bit length applied to the reference image buffer unit Is smaller than the pixel bit length applied to other internal processing (prediction processing, filter processing, etc.). Therefore, according to the video encoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing a highly accurate prediction process, filter process, and the like by applying a larger pixel bit length. Can be kept small.

Furthermore, the moving picture decoding apparatus according to the present embodiment can perform the same or similar scaling process / inverse scaling process as the moving picture encoding apparatus according to the present embodiment, and obtain the same effect. it can.

(Fourth embodiment)
In the fourth embodiment, a method is described in which scaling information and scaled reference pixel group data can be combined and expressed in a fixed length in bytes. As a result, the scaling information and scaled reference pixel group can be stored in the reference image buffer unit at the same time without having the scaling information buffer unit, and can be managed with a fixed length in bytes. Easy access.

As illustrated in FIG. 16, the moving image encoding apparatus according to the fourth embodiment includes an encoding unit 500 and an encoding control unit 140. The encoding unit 500 encodes the input image 11 to generate encoded data. The encoding control unit 140 controls various elements in the encoding unit 500. For example, the encoding control unit 140 controls a loop filter setting unit 106, a prediction unit 520, and the like which will be described later.

The encoding unit 500 includes a bit length extension unit 309, a subtraction unit 101, a transform / quantization unit 102, an entropy encoding unit 303, an inverse quantization / inverse transform unit 104, an addition unit 105, a loop filter setting unit 106, and a reference image A buffer unit 507, a loop filter unit 510, a prediction unit 520, and a motion vector generation unit 130 are included.

The bit length extension unit 309 is the same as that already described in the second embodiment.

However, the part in which the pixel bit length in the fourth embodiment is extended does not have to be the entire encoding unit and decoding unit as shown in FIGS. 16 and 17. It is sufficient that the pixel bit length is extended at least in the loop filter unit 510 (610: decoding device) and the prediction unit 520 (620: decoding device) described in this embodiment.

The subtraction unit 101 subtracts the prediction image from the prediction unit 520 from the input image 11 to obtain a prediction error. The transform / quantization unit 102 performs transform (for example, discrete cosine transform (DCT)) and quantization on the prediction error from the subtraction unit 101, and quantizes information on transform coefficients (hereinafter simply referred to as quantum). (Referred to as conversion coefficient).

The entropy encoding unit 303 is the same as that already described in the second embodiment.

The inverse quantization / inverse transform unit 104 performs inverse quantization and inverse transform (for example, inverse discrete cosine transform (IDCT), etc.) on the quantized transform coefficient from the transform / quantization unit 102, thereby reducing a prediction error. Restore. The adding unit 105 adds the prediction error restored by the inverse quantization / inverse transform unit 104 and the corresponding prediction image from the prediction unit 520 to generate the local decoded image 12.

The loop filter setting unit 106 sets the loop filter information 13 based on the input image 11 and the corresponding local decoded image 12 from the addition unit 105 and notifies the loop filter unit 510 and the entropy encoding unit 303 of the loop filter information 13. The loop filter information 13 includes at least filter coefficient information and filter switching information. The filter coefficient information includes information indicating the filter coefficient. The filter coefficient information may further include information indicating an offset coefficient described later. The filter switching information includes information indicating validity / invalidity of filter application.

The loop filter unit 510 performs filter processing or bypass processing that does not pass the filter processing on the target pixel group in the local decoded image 12 from the addition unit 105 according to the loop filter information 13 from the loop filter setting unit 106. Do. Then, the loop filter unit 510 performs scaling processing described later on the filter processing result group or the bypass processing result group (that is, the target pixel group itself) to reduce (or maintain) the pixel bit length. The loop filter unit 510 supplies the scaled reference pixel group 14 as the scaling processing result group and the scaling information 15 regarding the scaling processing to the reference image buffer unit 507. Details of the loop filter unit 510 will be described later.

The reference image buffer unit 507 stores the scaled reference pixel group 14 from the loop filter unit 510 and the scaling information 15 related to the scaling process. The reference pixel group 14 and the scaling information 15 accumulated in the reference image buffer unit 507 are read by the prediction unit 520 or the motion vector generation unit 130 as necessary. Details of the reference image buffer unit 507 will be described later.

The motion vector generation unit 130 reads the scaled reference pixel group and the scaling information from the reference image buffer unit 507 as necessary. The motion vector generation unit 130 applies inverse scaling that extends (or maintains) the pixel bit length to the scaled reference pixel group according to the scaling information to restore the reference image. The motion vector generation unit 130 generates motion vector information based on the input image 11 and the restored reference image. The motion vector generation unit 130 notifies the prediction unit 520 and the entropy encoding unit 303 of the motion vector information. Details of the inverse scaling process will be described later.

The prediction unit 520 reads the scaled reference pixel group and scaling information from the reference image buffer unit 507 as necessary. The prediction unit 520 restores the reference image by applying inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information. The prediction unit 520 generates a prediction image based on the motion vector information from the motion vector generation unit 130 and the restored reference image. The prediction unit 520 supplies the predicted image to the subtraction unit 101 and the addition unit 105.

The loop filter unit 510 will be described on the assumption that it has the same configuration as that of FIG. 3 already described in the first embodiment.

Note that the loop filter unit 510 does not have to have the same configuration as that of the first embodiment. For example, as shown in FIG. 18, according to the loop filter information 13 from the loop filter setting unit 106, a filter process is performed on the target pixel group in the local decoded image 12 from the addition unit 105, or bypass that does not pass through the filter process Process. A configuration may be adopted in which scaling processing is performed by the scaling processing unit 113 described later on the target pixel group of the filter processing result group or the bypass processing result group. The filter processing unit 1001 is not limited to a filter for which application or non-application of a filter is determined on a block basis as shown in FIG. For example, the filter processing unit 1001 may include a deblocking filter that applies filter processing to block boundaries.

In the fourth embodiment, the operation of the filter processing / scaling processing unit 112 and the operation of the scaling processing unit 113 are different from those in the first embodiment. Details of the operation of the filter processing / scaling processing unit 112 will be described below with reference to FIG.

The filter processing / scaling processing unit 112 performs a convolution operation (filter operation) on the target pixel group in accordance with the filter coefficient information included in the loop filter information 13 (step S112-1). Specifically, when the filter coefficient is represented by F [n], the pixel value of the target pixel group is represented by P [m], and the convolution calculation result is represented by B [m], the filter processing / scaling processing unit 112 represents the following formula ( Perform the convolution operation according to 1).

In Equation (19), O represents an offset coefficient. The offset coefficient can be referred to through the filter coefficient information.

In the fourth embodiment, the unit of the scaling processing is set to 16 pixels, the sum of the filter coefficient F [n] is described as being designed to be substantially equal to the 2 ^K. The 16-pixel unit is, for example, a 4 × 4 pixel block unit, but there is no restriction on the shape. In the following description, the pixel bit length of the target pixel group is 8 bits, and the number of extension bits at the time of input to the filter / scaling processing unit is 4 bits. Note that the pixel bit length and the number of extension bits of the target pixel group are not limited to these. The filter processing / scaling processing unit 112 performs such a convolution operation on each pixel of the target pixel group to obtain a convolution operation result group.

Next, the filter processing / scaling processing unit 112 searches for the maximum value Max and the minimum value Min of the convolution calculation result group of the target pixel group (step S112-2). However, the upper limit of the maximum value Max is, for example, 2 ^{K + 12} −1. That is, if the value of the maximum value Max is larger than 2 ^{K + 12} −1, the value of the maximum value Max is treated as 2 ^{K + 12} −1.

As another method, there is a method in which 0 ≦ MinP ≦ MaxP ≦ 255, and the lower limit of the minimum value Min is (MinP × 2 ^{K + 4} ) and the upper limit of the maximum value Max is (MaxP × 2 ^{K + 4} ). . In this case, if the value of the minimum value Min is smaller than (MinP × 2 ^{K + 4} ), the value of the minimum value Min is assumed to be (MinP × 2 ^{K + 4} ), and the value of the maximum value Max is larger than (MaxP × 2 ^{K + 4} ). Then, the value of the maximum value Max is handled as (MaxP × 2K ^{+ 4} ).

Next, the filter processing / scaling processing unit 112 derives the scaling information 15 of the target pixel group (step S112-3). Specifically, the filter processing / scaling processing unit 112 derives the minimum reference value MinPoint by arithmetically shifting the minimum value Min by (K + 6) bits according to the following formula (2).

Here, MinPoint is a 6-bit value.

Also, the filter processing / scaling processing unit 112 derives the scaling amount Q by executing the following calculation (21).

The scaling amount Q can take any integer value from 0 to 4. The minimum reference value MinPoint and the scaling amount Q derived as described above are supplied to the reference image buffer unit 507 as scaling information 15.

Next, the filter processing / scaling processing unit 112 applies scaling to each convolution operation result according to the derived scaling information 15 (step S112-4). Specifically, the filter processing / scaling processing unit 112 generates each pixel value D [m] of the scaled reference pixel group according to the following calculation (22) and calculation (23).

If the scaling amount Q is 4, the scaled reference pixel group is generated by executing the calculation (22). In this case, the first pixel B [0] of the target pixel group is set to 1 when B [0] is smaller than 2K ^{+ 3} so that the scaled pixel value D [0] does not become 0. . In other cases, rounding is performed by normal rounding.

Note that D [0] is limited to a value from 1 to 255, but D [m] in other cases is limited to a value from 0 to 255. In this example, 0 does not appear in D [0], but this value may be other than 0. For example, there is a method of using 255. Further, regarding the limitation of the value of D [m], for example, a method may be used in which the luminance signal is limited to a range of 16 to 235 and the color difference signal is limited to a value of 16 to 240. In this case, the value of D [0] does not need to be specifically limited.

The scaled reference pixel group generated as described above is supplied to the reference image buffer unit 507. By the operation of the filter processing / scaling processing unit 112, (K + 12) bits that are the pixel bit length of the filter processing result group of the target pixel group are reduced to 8 bits after the filter processing.

When the scaling amount Q is less than 4, according to the following calculation (23), the minimum reference value MinPoint is arithmetically left shifted by K + 6 bits to return to the original bit length, and this value is subtracted from the value of the target pixel group. Thus, each pixel value D [m] of the scaled reference pixel group is generated by shifting left by Q + K bits.

The scaled reference pixel group generated as described above is supplied to the reference image buffer unit 507. By the operation of the filter processing / scaling processing unit 112, (K + 12) bits that are the pixel bit length of the filter processing result group of the target pixel group are reduced to 7 bits after the filter processing.

Hereinafter, the operation of the scaling processing unit 113 will be described in detail with reference to FIG.

In the fourth embodiment, the unit of scaling processing is 16 pixel units, and the 16 pixel unit is, for example, a 4 × 4 pixel block unit, but there is no restriction on the shape. In the following description, the pixel bit length of the target pixel group is 8 bits, and the number of extension bits at the time of input to the scaling processing unit is 4 bits. The scaling processing unit 113 generates scaling information for the target pixel group and a scaled reference pixel group.

The scaling processing unit 113 searches for the maximum value Max and the minimum value Min of the target pixel group (step S113-1). However, the upper limit of the maximum value Max is set to 2 ¹² −1, for example. That is, if the value of the maximum value Max is larger than 2 ¹² −1, the value of the maximum value Max is handled as 2 ¹² −1.

As another method, there is a method in which 0 ≦ MinP ≦ MaxP ≦ 255, the lower limit of the minimum value Min is (MinP × 2 ⁴ ), and the upper limit of the maximum value Max is (MaxP × 2 ⁴ ). . In this case, if the value of the minimum value Min is smaller than (MinP × 2 ⁴ ), the value of the minimum value Min is treated as (MinP × 2 ⁴ ), and the value of the maximum value Max is from (MaxP × 2 ⁴ ). If it is larger, the maximum value Max is treated as (MaxP × 2 ⁴ ).

Next, the scaling processing unit 113 derives the scaling information 15 of the target pixel group (step S113-2). Specifically, the scaling processing unit 113 derives the minimum reference value MinPoint by performing an arithmetic right shift of the minimum value Min by about 6 bits according to Expression (24). However, in the operation of the scaling processing unit 113, K = 0 is handled.

Also, the scaling processing unit 113 derives the scaling amount Q by executing the calculation (25). The minimum reference value MinPoint and the scaling amount Q derived as described above are supplied to the reference image buffer unit 507 as scaling information 15.

Next, the scaling processing unit 113 applies scaling to each pixel value of the target pixel group in accordance with the derived scaling information 15 (step S113-3). Specifically, the scaling processing unit 113 generates each pixel value D [m] of the scaled reference pixel group according to the following formula (26).

If the scaling amount Q is 4, a scaled reference pixel group is generated by executing the calculation (26). In this case, the first pixel B [0] of the target pixel group is set to 1 when B [0] is a value smaller than 8 so that the scaled pixel value D [0] does not become 0. In other cases, rounding is performed by normal rounding.

D [0] is limited to a value from 1 to 255, but D [m] in other cases is limited to a value from 0 to 255. In this example, 0 does not appear in D [0], but this value may be other than 0. For example, there is a method of using 255. Further, regarding the limitation of the value of D [m], for example, a method may be used in which the luminance signal is limited to a range of 16 to 235 and the color difference signal is limited to a value of 16 to 240. In this case, the value of D [0] does not need to be specifically limited.

The scaled reference pixel group generated as described above is supplied to the reference image buffer unit 507. By the operation of the scaling processing unit 113, the pixel bit length of 12 bits of the target pixel group is reduced to 8 bits after the filter processing.

If the scaling amount Q is less than 4, according to the following calculation (27), the minimum reference value MinPoint is arithmetically shifted left by about 6 bits to return to the original bit length, and this value is subtracted from the value of the target pixel group. In addition, the pixel values D [m] of the scaled reference pixel group are generated by shifting left by Q bits.

The scaled reference pixel group generated as described above is supplied to the reference image buffer unit 507. By the operation of the

scaling processing unit

113, 12 bits, which is the pixel bit length of the target pixel group, is reduced to 7 bits after the filter processing.

An example of a method for efficiently describing the scaling information 15 and the scaled pixel group will be described with reference to FIG. In the syntax shown in FIG. 19, the descriptor in the right column represents the description bit length of each element of the scaling block. A scaling block (or adaptive scaling block) is a representation that includes scaling information and a group of scaled pixels. Here, description elements for expressing a scaling block with a fixed length will be described, but the order of writing the description elements of the scaling block to the reference image buffer or the order of reading the description elements of the scaling block from the reference image buffer is arbitrary. .

When the scaling amount Q is 4, 16 scaled pixel groups D [m] are described with a fixed length of 8 bits. That is, after D ₄ [0] that is not 0 is described in 8 bits, each of D ₄ [m] (m = 1 to 15) is described in 8 bits. Therefore, 16 scaled pixel groups D [m] when Q is 4 can be described by 128 bits (ie, 1 × 8 + 16 × 8).

When the scaling amount Q is less than 4, describe the fixed bit of the 8-bit value 0, then describe the scaling amount Q of the value from 0 to 3 with a fixed length of 2 bits, and MinPoint with a fixed length of 6 bits After that, 16 scaled pixel groups D [m] are described with a fixed length of 7 bits. Therefore, 16 scaled pixel groups D [m] when Q <4 can be described by 128 bits (ie, 8 + 2 + 6 + 16 × 7).

Thus, it becomes possible to describe a scaling block including scaling information and a group of 16 pixels of 4 × 4 blocks with a fixed length of 128 bits. If the 8-bit value from the beginning of the scaling block is 0 (ie, D [0] = 0), the scaling amount Q is found to be 4, and the 8-bit value from the beginning of the scaling block is not 0 (ie, If D [0] ≠ 0 (for example, D [0] is limited to 1), the scaling amount Q is determined from the next 2 bits, and MinPoint is determined from the subsequent 6 bits. As described above, in the embodiment, the description of the scaling information can be switched by limiting the scaled leading pixel value so as not to be a specific value.

Hereinafter, details of the inverse scaling processing in the prediction unit 520 will be described as an example. Note that the inverse scaling process substantially the same as or similar to the prediction unit 520 is also performed in the motion vector generation unit 130. As illustrated in FIG. 7, the prediction unit 520 includes an inverse scaling processing unit 121 and a predicted image generation unit 122.

The inverse scaling processing unit 121 restores the reference image by applying inverse scaling for extending the pixel bit length to the scaled reference pixel group according to the scaling information. The predicted image generation unit 220 generates a predicted image based on the motion vector information and the restored reference image.

Hereinafter, the operation of the inverse scaling processing unit 121 will be described in detail with reference to FIG. The inverse scaling processing unit 121 obtains a desired scaled reference pixel group (that is, necessary for generating a predicted image) and corresponding scaling information from the reference image buffer unit 507 (step S121-1). Specifically, the inverse scaling processing unit 121 acquires each pixel value D [m] of the scaled reference pixel group, the minimum reference value MinPoint, the scaling amount Q, and the filter switching information. In this case, if the first pixel value D [0] of the scaled reference pixel group is other than 0, the scaling amount Q is set to 4, and if D [0] is 0, the minimum reference value MinPoint and the scaling value Q are described. Loaded data.

Next, the inverse scaling processing unit 121 applies inverse scaling for extending the pixel bit length to the scaled reference pixel group according to the scaling information (step S121-2).

If the scaling value Q is 4, the pixel value of the reference pixel group scaled using the operation (28) is arithmetically shifted to the left by about 4 bits, and inverse scaling processing is performed.

If the scaling value Q is less than 4, the reference pixel scaled using the operation (29) is arithmetically shifted to the left by Q bits, and the minimum reference value MinPoint is arithmetically shifted to the left by about 6 bits and restored. The inverse scaling process is performed by adding the objects and adding the offset value ((Q! = 0)? (1 << (Q-1)): 0) according to the value of the scaling value Q.

G [m] represents each pixel value of the restored reference pixel group. By the operation of the inverse scaling processing unit 121, the pixel bit length (here, 8 bits) of the scaled reference pixel group is expanded to 12 bits. Note that the value of G [m] is limited to a range that can be expressed by 12 bits.

As described above, in this embodiment, adaptive scaling / inverse scaling processing based on the difference between the maximum value and the minimum value of the target pixel group is performed. By performing such scaling / inverse scaling processing, the value D [m] obtained by rounding the value B [m] before scaling to 8 bits by rounding off becomes D [0] in the case of 4-bit scaling. Except where the power is set to 1, it is guaranteed that the value G [m] after inverse scaling is rounded to 8 bits by rounding off. Compared to the method of generating a value G [m] after inverse scaling by rounding B [m] to 8 bits by fixed rounding to 8 bits, the value G after inverse scaling obtained by the processing of this embodiment is used. [M] has high accuracy.

Normally, in the case of an 8-bit image signal, the pixel value 0 and the pixel value 255 may not be used as a timing reference in various standards, and therefore, D [0] is limited within encoding and decoding. There is no substantial loss.

In the above description, the occurrence of 0 in the value of D [0] in the case of 4-bit scaling is prohibited and distinguished from the case of a scaling value other than 4 bits. Similar implementations are possible with other values such as.

In the above embodiment, the pixel value of the 8-bit input image is expanded by 4 bits, the target pixel group for performing scaling and inverse scaling processing is 16 pixels, the maximum value of the scaling value Q is 4, and the minimum reference value Although the shift amount at the time of generating MinPoint is 6, the pixel value scaled by the maximum scaling value is 8 bits, and the scaled pixel value of other scaling values is 7 bits, the same applies to some other values. Is possible.

For example, specifically, the pixel value of the 10-bit input image is not expanded, the target pixel group to be subjected to scaling and inverse scaling processing is set to 16 pixel units, the maximum value of the scaling value Q is set to 2, and the minimum reference value MinPoint is set. It is also possible to adopt a configuration in which the shift amount at the time of generation is set to 7, the bit length of the pixel value scaled with the maximum scaling value is set to 8 bits, and the bit length of the pixel value scaled with other scaling values is set to 7 bits. .

As described above, the video encoding apparatus according to the fourth embodiment performs the scaling process and the inverse scaling process before and after the reference image buffer unit, respectively, so that the pixel bit length applied to the reference image buffer unit Is smaller than the pixel bit length applied to other internal processing (prediction processing, filter processing, etc.). Therefore, according to the video encoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing a highly accurate prediction process, filter process, and the like by applying a larger pixel bit length. Can be kept small. Further, since the scaling information and the scaled reference pixel group data can be managed in a fixed length in units of bytes, data access in the reference image buffer is facilitated.

As illustrated in FIG. 17, the moving picture decoding apparatus according to the present embodiment includes a decoding unit 600 and a decoding control unit 240. The decoding unit 600 generates the output image 26 by decoding the encoded data. The decoding control unit 240 controls various elements in the decoding unit 600. For example, the decoding control unit 240 controls the prediction unit 620 and the like described later. Note that the scaling process and the inverse scaling process in the moving picture decoding apparatus in FIG. 17 are substantially the same as or similar to the scaling process and the inverse scaling process in the moving picture encoding apparatus in FIG.

The decoding unit 600 includes an entropy decoding unit 401, an inverse quantization / inverse transformation unit 202, an addition unit 203, a loop filter unit 610, a reference image buffer unit 604, a prediction unit 620, and a bit length normalization unit 430.

The entropy decoding unit 401 is the same as that already described in the second embodiment.

The inverse quantization / inverse transformation unit 202 and the addition unit 203 are substantially the same or similar elements as the inverse quantization / inverse transformation unit 104 and the addition unit 105 described above. That is, the inverse quantization / inverse transform unit 202 performs inverse quantization and inverse transform (for example, inverse discrete cosine transform (IDCT), etc.) on the quantized transform coefficient from the entropy decoding unit 401 to generate a prediction error. To restore. The adding unit 203 adds the prediction error restored by the inverse quantization / inverse transform unit 202 and the corresponding prediction image from the prediction unit 620 to generate a decoded image 22.

The loop filter unit 610 is an element that is substantially the same as or similar to the loop filter unit 510 described above. That is, the loop filter unit 610 performs a filtering process or a bypass process that does not pass the filtering process on the target pixel group in the decoded image 22 from the adding unit 203 according to the loop filter information 23 from the entropy decoding unit 401. . Further, the loop filter unit 610 reduces the pixel bit length by performing the above-described scaling processing on the filter processing result group or the bypass processing result group (that is, the target pixel group). The loop filter unit 610 supplies the scaled reference pixel group 24 and the scaling information 25 related to the scaling process to the reference image buffer unit 604 as the scaling process result group.

The reference image buffer unit 604 stores the scaled reference pixel group 24 and the scaling information 25 from the loop filter unit 610. The reference pixel group 24 and the scaling information 25 accumulated in the reference image buffer unit 604 are read by the prediction unit 620 or the bit length normalization unit 430 as necessary. For example, in order to generate the output image 26, the bit length normalization unit 430 reads out a desired reference pixel group (ie, necessary for generating the output image 26) and corresponding scaling information according to the display order.

The prediction unit 620 is an element that is substantially the same as or similar to the prediction unit 520 described above. That is, the prediction unit 620 reads the scaled reference pixel group and scaling information from the reference image buffer unit 604 as necessary. The prediction unit 620 restores the reference image by applying inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information. The prediction unit 620 generates a prediction image based on the motion vector information from the entropy decoding unit 401 and the restored reference image. The prediction unit 620 supplies the predicted image to the addition unit 203.

The bit length normalization unit 430 reads the scaled reference pixel group and the scaling information from the reference image buffer unit 604 as necessary. The bit length normalization unit 430 applies inverse scaling that extends the pixel bit length to the scaled reference pixel group according to the scaling information, and obtains a desired pixel bit length U (here, the bit length normalization unit 430). The pixel bit length U related to the operation of the prediction unit 620 does not necessarily match the pixel bit length U related to the operation of the prediction unit 620). The bit length normalization unit 430 supplies the output image 26 to the outside. Note that the pixel bit length of the output image 26 may be the same as or different from the pixel bit length of the input image 11 in the moving image encoding device, for example. When the decoded image 22 is used as it is as the output image 26 (without passing through the loop filter unit 610 and without changing the pixel bit length), the bit length normalization unit 430 can be omitted. It is. The bit length normalization unit 430 may perform bit length normalization of the output image 26 according to the internal pixel bit length information 47 from the entropy decoding unit 401.

As described above, the moving picture decoding apparatus according to the fourth embodiment uses the pixel bit length applied to the reference image buffer unit as the pixel bit applied to other internal processing (prediction processing, filter processing, etc.). The reference pixel group and the scaling information can be simultaneously stored in the reference image buffer unit. Therefore, according to the video decoding device according to the present embodiment, the pixel bit length applied to the reference image buffer unit while realizing a highly accurate prediction process, filter process, and the like by applying a larger pixel bit length. Can be kept small. Further, since the scaling information and the scaled reference pixel group data can be managed in a fixed length in units of bytes, data access in the reference image buffer is facilitated.

(Fifth embodiment)
In the fifth embodiment, instead of accumulating scaling information and a scaled reference image group in a reference image buffer, a method for controlling the degree of deterioration of pixel accuracy caused by scaling will be described. According to the fifth embodiment, it is possible to match the encoding result and the decoding result regardless of the presence or absence of scaling processing and inverse scaling before and after the reference image buffer.

As illustrated in FIG. 20, the moving image encoding apparatus according to the fifth embodiment includes an encoding unit 700 and an encoding control unit 140. The encoding unit 700 encodes the input image 11 to generate encoded data. The encoding control unit 140 controls various elements in the encoding unit 700. For example, the encoding control unit 140 controls the prediction unit 720 and the like.

The encoding unit 700 includes a bit length extension unit 309, a subtraction unit 101, a transform / quantization unit 102, an entropy encoding unit 703, an inverse quantization / inverse transform unit 104, an addition unit 105, a reference image buffer unit 707, a loop filter. A unit 710, a pixel accuracy control unit 721, a prediction unit 720, and a motion vector generation unit 730.

The bit length extension unit 309 is the same as that already described in the second embodiment. However, the part in which the pixel bit length is extended in the fifth embodiment does not have to be the entire encoding unit and decoding unit as shown in FIGS. 20 and 21, and in this embodiment, This can be realized if the pixel bit length is extended up to the output of the

loop filter units

710 and 810 and the input of the

prediction units

720 and 820 described.

The subtraction unit 101 subtracts the prediction image from the prediction unit 720 from the input image 11 to obtain a prediction error. The transform / quantization unit 102 performs transform (for example, discrete cosine transform (DCT)) and quantization on the prediction error from the subtraction unit 101, and quantizes information on transform coefficients (hereinafter simply referred to as quantum). (Referred to as conversion coefficient).

The entropy encoding unit 703 performs entropy encoding on the quantized transform coefficient from the transform / quantization unit 102, the motion vector information from the motion vector generation unit 730, and the pixel bit length extension information. Note that the entropy encoding unit 703 may further entropy encode information other than these (for example, prediction mode information). The type of entropy encoding is, for example, variable length encoding or arithmetic encoding. The entropy encoding unit 703 outputs encoded data obtained by entropy encoding to the outside.

The inverse quantization / inverse transform unit 104 performs inverse quantization and inverse transform (for example, inverse discrete cosine transform (IDCT), etc.) on the quantized transform coefficient from the transform / quantization unit 102, thereby reducing a prediction error. Restore. The adding unit 105 adds the prediction error restored by the inverse quantization / inverse transform unit 104 and the corresponding prediction image from the prediction unit 720 to generate the local decoded image 12.

The loop filter unit 710 receives the local decoded image 12 from the adding unit 106, performs a loop filter process such as a deblocking process or an image restoration process on the local decoded image 12, and generates a decoded image signal.

The reference image buffer unit 707 stores the decoded image signal from the loop filter unit 710. The detailed operation of the image buffer unit 707 will be described later with reference to FIGS.

The pixel accuracy control unit 721 includes a decoded image signal (hereinafter referred to as a reference image signal) used for prediction processing among the decoded image signals stored in the reference image buffer unit 707, and pixels from the bit length extension unit 309. Bit length extension information is received. The pixel accuracy control unit 721 performs processing for controlling the degree of deterioration in pixel accuracy caused by scaling. Detailed operation of the pixel accuracy control unit 721 will be described later with reference to FIG.

The motion vector generation unit 730 receives the image signal from the bit length extension unit 309 and the reference image signal from the pixel accuracy control unit 721, calculates a motion vector, and generates motion vector information. The motion vector generation unit 730 notifies the motion vector information to the prediction unit 720 and the entropy encoding unit 703.

Hereinafter, details of the operation of the pixel accuracy control unit 721 will be described with reference to FIG.

In the fifth embodiment, description will be made assuming that the unit of pixel accuracy control processing is 16 pixels. The 16-pixel unit is, for example, a 4 × 4 pixel block unit, but there is no restriction on the shape. In the following description, the pixel bit length of the target pixel group is 8 bits, and the number of pixel extension bits is 4 bits. Note that the pixel bit length and the pixel extension bit number of the target pixel group are not limited to these specific bit numbers.

The pixel accuracy control unit 721 searches for the maximum value Max and the minimum value Min of the pixel values of the target pixel group (step S712-1). However, the upper limit of the maximum value Max is set to 2 ¹² −1, for example. That is, if the value of the maximum value Max exceeds 2 ¹² −1, the value of the maximum value Max is handled as 2 ¹² −1. There is also a method in which the upper limit of the maximum value Max is, for example, (255 × 2 ⁴ ). In this case, if the maximum value Max exceeds (255 × 2 ⁴ ), the maximum value Max is handled as (255 × 2 ⁴ ).

Next, the pixel accuracy control unit 721 derives the scaling amount Q of the target pixel group (step S712-2). Specifically, the scaling amount Q is derived by executing the following calculation (30).

The scaling amount Q can take any integer value from 0 to 4.

Next, the pixel accuracy control unit 721 applies scaling to each pixel value according to the derived scaling amount Q (step S712-3). Specifically, the pixel accuracy control unit 721 generates each pixel value G [m] of the reference pixel group whose pixel accuracy is controlled according to the following calculations (31), (32), and (33).

If the scaling amount Q is 4, the scaled reference pixel group is generated by executing the calculation (31). At this time, for the first pixel P [0] of the target pixel group, when P [0] is a value smaller than 8 so that the pixel value G [0] whose pixel accuracy is controlled is not 0, 16 To do. Otherwise, after adding 8, zero padding is performed on the lower 4 bits.

When the scaling amount Q is 0, the input pixel value is used as the output value as it is as in the following calculation (32).

When the scaling amount Q is a value other than the above, the lower Q bits are zero-padded according to the following calculation (33), and the offset value (1 << (Q-1)) is added.

Incidentally, G [0] if the scaling amount Q is 4, but is limited to 16 to a value of 255 × ^{2 4,} otherwise, is limited to a value of from 0 to 255 × ^{2 4} Shall. In this example, when P [0] is smaller than 8, 0 does not appear in G [0]. However, this value may be other than 0. For example, 255, G [0] there is also a method of limiting the value of 254 × ^{2 4.}

In addition, since the pixel value of the input image signal may be limited, for example, the range from 16 × 2 ^{4 to} 235 × 2 ⁴ for the luminance signal and 16 × 2 for the color difference signal. ⁴ may be a technique that limits a value of up to 240 × ^{2 4.} In this case, it is not necessary to limit the value of G [0] when the scaling amount Q is 4.

In the present embodiment, the pixel accuracy control method corresponding to the scaling process and the inverse scaling process of the fourth embodiment has been described. However, when another scaling method is employed, for example, 12 bits are fixed. In the case of rounding to 8 bits, the processing of Expression (34) may be performed on the entire screen.

Next, details of the operation of the reference image buffer 707 will be described with reference to FIGS.

FIG. 23 shows an example in which the scaling processing unit 113 and the inverse scaling processing unit 121 exist before and after the reference frame unit. The scaling processing unit 113 and the inverse scaling processing unit 121 are the same as those shown in the fourth embodiment. Based on the pixel bit length extension information, the scaling processing of FIG. 6 and the inverse scaling processing of FIG. I do. On the other hand, the reference image buffer unit in FIG. 24 shows an example in which the scaling processing unit 113 and the inverse scaling processing unit 121 do not exist before and after the reference frame unit.

In the fifth embodiment, by introducing the pixel accuracy control unit 721, the configuration of the reference image buffer unit 707 is the configuration shown in FIG. 23 or the configuration shown in FIG. The output results from the pixel accuracy control unit 721 are the same. That is, the same result can be obtained regardless of the presence or absence of the scaling process and the inverse scaling in the reference image buffer unit, and the introduction of the scaling process and the inverse scaling process can be selected depending on the mounting method.

Therefore, in the description so far, the position of the pixel accuracy control unit 721 is positioned after the reference image buffer 707, but the pixel accuracy control unit 721 may be positioned before the reference image buffer 707. In the case of the configuration of the reference image buffer 707 in FIG. 23, the same encoding and decoding results can be obtained even if the pixel accuracy control unit 721 does not exist.

Further, in the above fifth embodiment, the pixel value of the 8-bit input image is expanded by 4 bits, the target pixel group for performing the pixel accuracy control processing is set to 16 pixels, and the maximum value of the scaling value Q However, the same configuration is possible with some other values.

Specifically, for example, the pixel value of the 10-bit input image is not expanded, the unit of the target pixel group for performing the scaling and inverse scaling processing is 16 pixels, and the maximum value of the scaling value Q is 2. Is also possible.

As described above, the moving picture coding apparatus according to the fifth embodiment does not store scaling information and the scaled reference picture group in the reference picture buffer, but rather determines the degree of deterioration in pixel accuracy caused by scaling. Shown how to control. Regardless of the presence or absence of the scaling process and the inverse scaling, the encoding result and the decoding result can be matched, and the introduction of the scaling process and the inverse scaling process can be selected according to the mounting method.

As shown in FIG. 21, the moving picture decoding apparatus according to this embodiment includes a decoding unit 800 and a decoding control unit 240. The decoding unit 800 generates the output image 26 by decoding the encoded data. The decoding control unit 240 controls various elements in the decoding unit 600. For example, the decoding control unit 240 controls the prediction unit 820 and the like. 21 is the same as the pixel accuracy control unit 721 in the moving image encoding device in FIG.

The decoding unit 800 includes an entropy decoding unit 801, an inverse quantization / inverse transform unit 202, an addition unit 203, a loop filter unit 810, a reference image buffer unit 804, and a prediction unit 820.

The entropy decoding unit 801 performs entropy decoding according to the syntax information, for example, on the encoded data generated by the moving image encoding device in FIG. The entropy decoding unit 801 supplies the decoded quantized transform coefficient to the inverse quantization / inverse transform unit 202, supplies the decoded motion vector information to the prediction unit 820, and decodes the decoded pixel bit extension Information is supplied to the pixel accuracy control unit 806.

The inverse quantization / inverse transformation unit 202 and the addition unit 203 are substantially the same or similar elements as the inverse quantization / inverse transformation unit 104 and the addition unit 105 described above. That is, the inverse quantization / inverse transform unit 202 performs inverse quantization and inverse transform (for example, inverse discrete cosine transform (IDCT), etc.) on the quantized transform coefficient from the entropy decoding unit 201 to generate a prediction error. To restore. The adding unit 203 adds the prediction error restored by the inverse quantization / inverse transform unit 202 and the corresponding prediction image from the prediction unit 620 to generate a decoded image 22.

Similarly to the loop filter unit 710, the loop filter unit 810 receives the decoded image signal from the adding unit 203 and performs loop filter processing on the decoded image signal.

The reference image buffer unit 804 is the same as the reference image buffer unit 707, and can adopt either the configuration of FIG. 23 or FIG. The reference image buffer unit 804 stores the decoded image signal after the filter processing from the loop filter unit 810. Also, in response to an external request, the decoded image signal is extracted as a reference image signal from the reference image buffer unit 804, and the reference image signal is output in accordance with the display order.

The pixel accuracy control unit 806 has the same configuration as the pixel accuracy control unit 721 described above, receives a reference image signal from the reference image buffer unit 804, and receives pixel bit length extension information from the entropy decoding unit 801. Detailed processing contents are the same as those of the pixel accuracy control unit 721.

The position of the pixel accuracy control unit 806 in FIG. 21 is positioned after the reference image buffer unit 804, but may be positioned before the reference image buffer unit 804. Further, when the reference image buffer unit 804 has the same configuration as the reference image buffer unit of FIG. 23, the encoding result and the decoding result are the same even if the pixel accuracy control unit 806 does not exist.

Like the prediction unit 720, the prediction unit 820 receives a reference image signal from the pixel accuracy control unit 806, receives motion vector information from the entropy decoding unit 801, performs prediction processing, and generates a predicted image signal. The prediction unit 820 supplies the predicted image to the addition unit 203.

As described above, the moving picture decoding apparatus according to the fifth embodiment does not store scaling information and the scaled reference picture group in the reference picture buffer, but controls the degree of deterioration in pixel accuracy caused by scaling. Showed how to do. Regardless of the presence or absence of scaling processing and inverse scaling, the results of encoding and decoding can be matched, and introduction of scaling processing and inverse scaling processing can be selected depending on the implementation method.

The various processes described in the first to fifth embodiments may be realized by executing a program (software). A general-purpose computer system reads a program from a storage medium storing a program for realizing the processing according to each embodiment, and executes the program by a CPU or the like. It operates as a decoding device and brings about the same effect.

Programs include magnetic disks (flexible disks, hard disks, etc.), optical disks (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD ± R, DVD ± RW, etc.), semiconductor memory, or the like. It may be stored in other storage media. These storage media may be of any type as long as they can be read by a computer or an embedded system that reads the program. The computer or the embedded system may acquire or read the program via a communication medium such as a network. That is, a medium that downloads a program via a communication medium such as a LAN (local area network) or the Internet and stores (including temporary storage) the program is also included in the category of “storage medium”. When the program is distributed and stored in a plurality of storage media, the term “storage medium” can also refer to a plurality of storage media comprehensively.

Further, the computer or the embedded system may be a single device such as a personal computer or a microcontroller, or may be a system in which a plurality of devices are connected to a network. Further, the term “computer” is not limited to a so-called personal computer, and can comprehensively refer to an apparatus capable of executing a program, including an arithmetic processing unit, a microcontroller, and the like included in an information processing apparatus.

A part of the processing according to each embodiment may be executed using a function such as an OS (operating system) operating on a computer, database management software, MW (middleware) such as a network.

Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

DESCRIPTION OF SYMBOLS 11 ... Input image 12 ... Local decoded

image

13, 23 ...

Loop filter information

14, 24 ... Scaled

reference pixel group

15, 25 ... Scaling information 22 ... Decoded image 26. ..

Output image

37, 47... Internal

bit length information

100, 300, 500, 700... Encoding unit 101... Subtraction unit 102 ... Transform /

quantization unit

103, 303, 703. Entropy encoding unit 104 ... inverse quantization / inverse conversion unit 105 ... addition unit 106 ... loop

filter setting unit

107, 507 ... reference image buffer unit 108 ... scaling

information buffer unit

110, 510 ... Loop filter unit 111 ... Switch 112 ... Filter processing / scaling processing unit 113 ...

Scaling processing unit

120, 52 ... Prediction unit 121 ... Inverse scaling processing unit 122 ... Prediction

image generation unit

130, 730 ... Motion vector generation unit 140 ...

Coding control unit

200, 400, 600, 800 ...

Decoding Encoding unit

201, 401, 801 ... Entropy decoding unit 202 ... Inverse quantization / inverse transformation unit 203 ...

Addition unit

204, 604, 707, 804 ... Reference image buffer unit 205 ... Scaling

Information buffer unit

210, 610, 710, 810 ...

Loop filter unit

220, 620, 720, 820 ...

Prediction unit

230, 430 ... Bit length normalization unit 240 ... Decoding control unit 309 ...・ Bit length extension part 721,806 ... Pixel accuracy control part

Claims

Deriving scaling information based on the maximum value and the minimum value of the target pixel group in the locally decoded image, applying scaling according to the scaling information to reduce the pixel bit length for the target pixel group, and the specific pixel to be scaled By restricting the value of the specific value to a specific value, a scaled reference pixel group is generated, and the description of the first scaling information when the specific value is included or the first when the specific value is not included A scaling processing unit that generates a scaling block in which the description of the scaling information of 2 and the reference pixel group scaled according to the corresponding scaling information are expressed by a fixed bit length;
Applying an inverse scaling that extends a pixel bit length to the scaled reference pixel group according to the scaling information to restore a reference image, and generating a predicted image based on the reference image;
A moving image encoding apparatus comprising: an encoding unit that encodes information indicating a difference between an input image and the predicted image.
The scaling processing unit
A scaling amount is determined by the maximum value and the minimum value of the pixel values in the target pixel group,
When scaling by the maximum scaling amount, the scaling result of the first pixel of the target pixel group is limited to a value other than the specific value, and scaling is performed by rounding the L-bit output,
The scaling of the target pixel group to M (M <L) bits according to scaling information represented by a scaling value and a representative value of a minimum value when scaling by a scaling amount other than the maximum scaling amount. The moving image encoding apparatus according to 1.
The inverse scaling processing unit
If the first pixel value of the scaled reference pixel group is the specific value, the reference pixel group is inversely scaled according to scaling information following the first pixel;
The moving image encoding apparatus according to claim 1, wherein if the first pixel value is other than the specific value, the reference pixel group is inversely scaled with respect to a maximum scaling amount.
The target pixel group has 16 pixels as a unit,
The scaling processing unit
A scaling amount is determined by the maximum value and the minimum value of the target pixel group,
When the scaling amount is 4 bits, scaling is performed by rounding the 8-bit output under the restriction that the scaled value of the first pixel is a value other than 0.
2. The moving picture encoding apparatus according to claim 1, wherein when the scaling amount is less than 4 bits, the representative value of the scaling value and the minimum value is used as scaling information, and the target pixel group is scaled to 7 bits.
The target pixel group has 16 pixels as a unit,
The inverse scaling processing unit
If the value of the first 8 bits of the scaled reference pixel group is 0, the reference according to scheduling information including a scaling amount represented by the following 2 bits and a representative value of the minimum value represented by the following 6 bits Inverse scaling the pixel group,
2. If the first 8-bit value of the scaled reference pixel group is non-zero, the reference pixel group is inversely scaled assuming that the reference pixel group is scaled by a 4-bit scaling amount. The moving image encoding apparatus described in 1.
Scaling information is derived based on the maximum value and the minimum value of the target pixel group in the decoded image, and scaling for reducing the pixel bit length is applied to the target pixel group according to the scaling information. A scaled reference pixel group is generated by limiting the value with respect to a specific value, and a description of first scaling information when the specific value is included or a second when the specific value is not included A scaling processing unit that generates a scaling block in which the description of the scaling information and the reference pixel group scaled according to the corresponding scaling information is expressed by a fixed bit length;
Applying an inverse scaling that extends a pixel bit length to the scaled reference pixel group according to the scaling information to restore a reference image, and generating a predicted image based on the reference image;
A moving picture decoding apparatus comprising: a decoding unit that decodes information indicating a difference between an input image and the predicted image.
The scaling processing unit
A scaling amount is determined by the maximum value and the minimum value of the pixel values in the target pixel group,
When scaling by the maximum scaling amount, the scaling result of the first pixel of the target pixel group is limited to a value other than the specific value, and scaling is performed by rounding the L-bit output,
The scaling of the target pixel group to M (M <L) bits according to scaling information represented by a scaling value and a representative value of a minimum value when scaling by a scaling amount other than the maximum scaling amount. 6. The moving picture decoding apparatus according to 6.
The inverse scaling processing unit
If the first pixel value of the scaled reference pixel group is the specific value, the reference pixel group is inversely scaled according to scaling information following the first pixel;
The video decoding device according to claim 6, wherein if the first pixel value is other than the specific value, the reference pixel group is inversely scaled with respect to a maximum scaling amount.
The target pixel group has 16 pixels as a unit,
The scaling processing unit
A scaling amount is determined by the maximum value and the minimum value of the target pixel group,
When the scaling amount is 4 bits, scaling is performed by rounding the 8-bit output under the restriction that the scaled value of the first pixel is a value other than 0.
The moving picture decoding apparatus according to claim 6, wherein when the scaling amount is less than 4 bits, the representative value of the scaling value and the minimum value is used as scaling information, and the target pixel group is scaled to 7 bits.
The target pixel group has 16 pixels as a unit,
The inverse scaling processing unit
If the value of the first 8 bits of the scaled reference pixel group is 0, the reference according to scheduling information including a scaling amount represented by the following 2 bits and a representative value of the minimum value represented by the following 6 bits Inverse scaling the pixel group,
7. If the first 8-bit value of the scaled reference pixel group is non-zero, the reference pixel group is inversely scaled assuming that the reference pixel group has been scaled by a 4-bit scaling amount. The moving picture decoding apparatus described in 1.
A reference pixel whose pixel accuracy is controlled by deriving a scaling amount based on the maximum value and the minimum value of the target pixel group in the locally decoded image, and changing the value of the lower bits of the target pixel group according to the scaling amount A pixel accuracy control unit for generating a group;
A prediction unit that generates a predicted image based on a reference pixel group in which the pixel accuracy is controlled;
A moving image encoding apparatus comprising: an encoding unit that encodes information indicating a difference between an input image and the predicted image.
The pixel accuracy control unit includes:
When the scaling amount Q is the maximum value, the first pixel value of the target pixel group is limited to a specific value, 2 Q-1 is added to the pixel value, and the lower Q bits are padded with zeros. Done
When the scaling amount is 0, the target pixel group is output as it is,
12. The moving image code according to claim 11, wherein when the scaling amount Q is neither the maximum value nor 0, the lower Q bits of the target pixel group are zero-padded and a value of 2Q -1 is added. Device.
The pixel accuracy control unit includes:
When the scaling amount Q is the maximum value 4, the first pixel value of the target pixel group is limited to a value other than the specific value, the value 8 is added to the pixel value, and the lower 4 bits are zero-padded.
When the scaling amount is 0, the target pixel group is output as it is,
12. The value of 2 Q−1 is added after performing zero padding on the lower Q bits of the target pixel group when the scaling amount Q is neither the maximum value 4 nor 0. Video encoding device.
A reference pixel group in which pixel accuracy is controlled by deriving a scaling amount based on the maximum value and the minimum value of the target pixel group in the decoded image, and changing the value of the lower bits of the target pixel group according to the scaling amount A pixel accuracy control unit for generating
A prediction unit that generates a predicted image based on a reference pixel group in which the pixel accuracy is controlled;
A moving picture decoding apparatus comprising: a decoding unit that decodes information indicating a difference between an input image and the predicted image.
The pixel accuracy control unit includes:
When the scaling amount Q is the maximum value, the first pixel value of the target pixel group is limited to a specific value, 2 Q-1 is added to the pixel value, and the lower Q bits are padded with zeros. Done
When the scaling amount is 0, the target pixel group is output as it is,
15. The moving picture decoding according to claim 14, wherein when the scaling amount Q is neither the maximum value nor 0, the lower Q bits of the target pixel group are zero-padded and a value of 2Q -1 is added. Device.
The pixel accuracy control unit
When the scaling amount Q is the maximum value 4, the first pixel value of the target pixel group is limited to a value other than the specific value, the value 8 is added to the pixel value, and the lower 4 bits are zero-padded.
When the scaling amount is 0, the target pixel group is output as it is,
The value of 2 Q-1 is added after performing zero padding on the lower Q bits of the target pixel group when the scaling amount Q is neither the maximum value 4 nor 0. Video decoding apparatus.
Deriving scaling information based on the maximum and minimum values of the target pixel group in the local decoded image,
Applying a scaling to reduce the pixel bit length for the target pixel group according to the scaling information, and generating a scaled reference pixel group by limiting a value of a specific pixel to be scaled with respect to a specific value;
The description of the first scaling information when the specific value is included or the description of the second scaling information when the specific value is not included and the reference pixel group scaled according to the corresponding scaling information are fixed. Generate a scaling block expressed in bit length of
Applying inverse scaling to extend the pixel bit length to the scaled reference pixel group according to the scaling information to restore the reference image;
Generating a predicted image based on the reference image;
A moving image encoding method for encoding information indicating a difference between an input image and the predicted image.
Deriving scaling information based on the maximum and minimum values of the target pixel group in the decoded image,
Applying a scaling to reduce the pixel bit length for the target pixel group according to the scaling information, and generating a scaled reference pixel group by limiting a value of a specific pixel to be scaled with respect to a specific value;
The description of the first scaling information when the specific value is included or the description of the second scaling information when the specific value is not included and the reference pixel group scaled according to the corresponding scaling information are fixed. Get the scaling block expressed in bit length of
Applying inverse scaling that extends a pixel bit length to the scaled reference pixel group according to the scaling information to restore a reference image, and generate a predicted image based on the reference image;
A moving picture decoding method for decoding information indicating a difference between an input picture and the predicted picture.