JP5702820B2 - Prediction mode determination method, video encoding apparatus, and prediction mode determination program - Google Patents

Description

The present invention relates to a prediction mode determination method, a video encoding device, and a prediction mode determination program.

  H. In a video encoding standard such as H.264 or HEVC (High Efficiency Video Coding), a picture of a moving image is encoded in units of blocks. H. In H.264, a macro block (a block of 16 pixels in the vertical direction and a block of 16 pixels in the horizontal direction) is used as a unit, and the macro block can be divided into sub-blocks for encoding. H. In H.264, a skip mode can be used as a macroblock prediction mode. H. In the H.264 skip mode, since the motion vector is calculated from the motion vectors of the surrounding encoded blocks, it is not necessary to encode the motion vector. Also, the prediction residual block when motion compensation is performed with the motion vector is not encoded.

  On the other hand, in HEVC, encoding is performed in units of LCU (Largest Coding Unit). The LCU is a block having a maximum length of 64 pixels and a width of 64 pixels (64 × 64 pixels), which can be divided into sub-blocks and encoded. H. The difference from H.264 is that there are many variations in the size of the sub-block, and this sub-block is called a CU (Coding Unit) (CU includes the LCU itself). FIG. 9 shows variations of the CU size. As shown in (a) to (d) of the figure, it is possible to recursively divide the LCU of 64 × 64 pixels into four CUs of 32 × 32 pixels, 16 × 16 pixels, and 8 × 8 pixels. It can also be seen that it can be expressed by a quadtree having four hierarchies. In encoding, CU partitioning using an arbitrary quadtree structure with the number of hierarchies within 4 can be used. An example of this is shown in FIG. As shown in the figure, the CU in the first layer is an LCU and is divided into four CUs CU1 to CU4. CU2 in the second hierarchy is further divided into CU21 to CU24, and CU4 in the second hierarchy is further divided into CU41 to CU44. Further, the CU 42 in the third hierarchy is divided into CUs 421 to 424 that are the fourth hierarchy, and the CU 43 in the third hierarchy is divided into CUs 431 to 434 that are the fourth hierarchy. In HEVC, it is possible to freely switch between using intra prediction or inter prediction in CU coding. Note that the CU can be further divided into sub-blocks called PUs (Prediction Units).

  In HEVC, when encoding CU by inter prediction, it can freely switch between encoding in skip mode or encoding in other prediction modes. That is, the skip mode can be applied to blocks of various sizes in one picture. Further, in the HEVC skip mode, H.264 is used in that the motion vector itself is not encoded. The same as H.264, but a list including motion vector / reference image index pair candidates (hereinafter referred to as “merge candidate list”) is generated. Then, it is possible to select which one to use from the merge candidate list, and the difference is that the index is encoded. The motion vector and the reference image index included in the list are generated from the motion vector used in the adjacent encoded block or another encoded picture.

  As described above, in HEVC, blocks of various sizes (CUs) in one picture can be encoded in the skip mode. However, in the following description of the prior art, for the sake of simplification, the skip mode uses a fixed block size as a unit. Can be used as

  In the skip mode, since the prediction vector is used for encoding, motion search is not performed. Therefore, before performing motion search in the encoding target block, if it is possible to determine whether or not the block is to be encoded in the skip mode (hereinafter referred to as “skip determination”), the encoding is performed in the skip mode. It becomes possible to omit the motion search in the block that has become, and it is possible to reduce the amount of calculation in encoding.

In general, the prediction mode selected in each block in a picture has a spatial correlation, and it is conceivable to perform skip determination using this. As such a system, for example, there is a conventional technique described in Non-Patent Document 1 (hereinafter referred to as “conventional technique A”).
In the prior art A, a value obtained by multiplying the average value of the cost of the skip mode in the block already coded as the skip mode by a coefficient is used as a threshold, and when the cost of the skip mode in the encoding target block is lower than the threshold, The skip mode is determined. An application example of this method is shown in FIG. As shown in the figure, when the skip mode is used in the encoded block adjacent to the encoding target block (block C) and the encoded block adjacent to the left (block D), the block A value obtained by multiplying the average of the skip mode costs of C and block D by a coefficient can be used as the threshold value. Hereinafter, such a threshold is referred to as a “skip determination threshold”.
In the prior art A, when the encoded block to be referenced is determined to be skipped by the prior art A, the skip determination threshold is determined based on the cost with low reliability, and an erroneous determination occurs and encoding is performed. Efficiency may be reduced.

  Therefore, in order to determine the skip determination threshold based on the cost with high reliability, it can be easily analogized to refer only to the encoded block to which the skip determination has not been applied. In prior art B, as shown in FIG. 11, when block C and block D were encoded in skip mode, block C and block D were decided to be encoded in skip mode by skip determination. Determine. If both or one of the blocks C and D is determined to be the skip mode by the skip determination, the skip determination for the encoding target block is not performed (in other cases, the skip determination is performed). Thereby, a highly reliable threshold value can be set in skip determination. In the following, a block that is encoded as a skip mode but is not determined to be in the skip mode in the skip determination performed before the motion search is referred to as an “optimum skip block”.

Jongho Kim, Seyoon Jeong, Sukhee Cho, and Jin Soo Choi, "Adaptive Coding Unit Early Termination Algorithm for HEVC", 2012 IEEE International Conference on Consumer Electronics (ICCE), p.261-262

  In the conventional technique B, there are few blocks to which the skip determination can be applied, and there is a problem that the effect of reducing the calculation amount is limited. FIG. 12 shows an example of a block to which skip determination can be applied in the related art B. The area consisting of 16 blocks of 4 vertical blocks and 4 horizontal blocks shown in the figure is an area in which the optimal prediction mode is set to the skip mode in the picture. Can perform skip determination and reduce the amount of calculation. The blocks to which skip determination can be applied in the prior art B are shaded blocks shown in the figure, and only four blocks can perform skip determination in 16 blocks.

  On the other hand, in the conventional technique B, if the block on the left side of the encoding target block is the optimal skip block, the restriction is relaxed so that the skip determination can be performed in the encoding target block using the cost of the optimal skip block. Think about the case. This is referred to as Conventional Technology B-1. In the case of the prior art B-1, skip determination can be applied to the shaded blocks shown in FIG. 13, and skip determination can be applied to half of the 16 blocks.

  However, in the case of the conventional technique B-1, even in a block where the skip mode is not optimal, it is determined as the skip mode, which may impair the coding efficiency. A conceptual diagram showing an example thereof is shown in FIG. In the figure, the subject is different between the leftmost block row and the right region. It is optimal to encode the left subject (subject α) in the skip mode, but it is not optimal to encode the right subject (subject β) in the skip mode. In terms of meaning, it is better to use other prediction modes. In such a case, in the conventional technique B-1, the skip mode may be erroneously determined in the second block row from the left.

  In view of the above circumstances, the present invention determines whether or not to encode a block to be encoded in a picture in the skip mode so that the encoding efficiency does not deteriorate, and is capable of reducing the amount of calculation in encoding. It is an object of the present invention to provide a mode determination method, a moving image encoding device, and a prediction mode determination program.

  One aspect of the present invention is a prediction mode determination method executed by a moving image encoding apparatus that determines a prediction mode in units of blocks and encodes image data of a moving image. A block feature comparison step for calculating a difference in the feature amount between an encoding target block in image data and a reference candidate block adjacent to the encoding target block and encoded in a skip mode; and the block feature The difference between the feature amounts calculated in the comparison step is a skip determination execution presence / absence determination step that determines whether or not a condition defined for the feature amount is satisfied, and the skip determination execution presence / absence determination step. Is determined to satisfy the condition defined for the feature amount, the reference candidate block is referred to as a reference block. Whether the prediction mode of the encoding target block is set to the skip mode by comparing the threshold obtained from the cost of the skip mode in the reference block with the cost when the encoding target block is set to the skip mode. And a skip determination step for determining the prediction mode determination method.

One aspect of the present invention is the prediction mode determination method described above, wherein the feature amount includes a motion vector, the condition includes a condition defined for the motion vector, and in the block feature comparison step, The calculation of the feature amount difference includes calculation of a degree of coincidence of motion vectors between the reference candidate block and the encoding target block, and in the skip determination execution presence / absence determining step, the reference candidate block and the encoding target are calculated. When the degree of coincidence of the motion vector with the block is within a predetermined range, it is determined that the condition defined for the motion vector is satisfied.

One aspect of the present invention is the prediction mode determination method described above, wherein the feature amount includes a complexity index value, the condition includes a condition defined for the complexity index value, and the block feature comparison In the step, the calculation of the feature amount difference includes calculation of a complexity index value between the reference candidate block and the encoding target block, and in the skip determination execution presence / absence determination step, the encoding target When the complexity of the block exceeds the complexity of the reference candidate block, it is determined that the condition defined for the complexity index value is satisfied.

  One aspect of the present invention is the prediction mode determination method described above, in which when there are a plurality of reference candidate blocks that are adjacent to the encoding target block and encoded in the skip mode, the block feature comparison In the step, a difference in the feature amount between the encoding target block and each of the plurality of reference candidate blocks is calculated, and in the skip determination execution presence / absence determining step, the block feature among the plurality of reference candidate blocks is calculated. A reference candidate block is a reference candidate block in which the difference between the feature amounts calculated in the comparison step satisfies a condition defined for the feature amount.

  One embodiment of the present invention is the prediction mode determination method described above, wherein, in the block feature comparison step, a predetermined block among blocks that are adjacent to the current block to be encoded and are encoded in the skip mode. A block larger than the size of the reference candidate block is used as a reference candidate block.

  One embodiment of the present invention is a moving image encoding apparatus that determines a prediction mode in units of blocks and encodes image data of a moving image, and the encoding target in the image data for one or more feature amounts The feature amount difference between the block and the reference candidate block that is adjacent to the encoding target block and is encoded in the skip mode is calculated, and the calculated feature amount difference is calculated for the feature amount. A block feature comparison unit that determines whether or not a predetermined condition is satisfied, and the reference candidate when the block feature comparison unit determines that the difference in the feature amount satisfies the condition defined for the feature amount The block is a reference block, the threshold obtained from the cost of the skip mode in the reference block, and the encoding target block is the skip mode. And the determining the cost comparison unit whether or not to skip mode prediction mode of the encoding target block by comparing the cost of a video encoding apparatus comprising: a.

  One embodiment of the present invention is a prediction mode determination program for causing a computer used as a moving image encoding apparatus to execute each step of the above-described prediction mode determination method.

  According to the present invention, it is possible to determine whether to encode a block to be encoded in a picture in the skip mode so that the encoding efficiency does not deteriorate, and to reduce the amount of calculation in encoding.

It is a block diagram which shows the function structure of the encoding apparatus by the 1st Embodiment of this invention. It is a figure which shows the processing flow of the prediction mode and optimal division | segmentation determination with respect to the process target CU by the embodiment. It is a figure which shows the example of the process target CU division | segmentation which integrated the optimal division | segmentation in the low-order CU by the embodiment. It is a block diagram which shows the detailed functional structure of the skip determination part by the embodiment. It is a figure which shows the processing flow of the skip determination by the same embodiment. It is a figure which shows the processing flow of the skip determination by 2nd Embodiment. It is a figure which shows the processing flow of the prediction mode and optimal division | segmentation determination with respect to the process target CU by 3rd Embodiment. It is a figure which shows the example of the condition where the optimal skip CU of the size different from the process target CU exists in the left or the top by the embodiment. It is a figure which shows the variation of CU size by a prior art. It is a figure which shows the CU division | segmentation example of LCU by a prior art. It is a figure which shows the block referred in the skip determination by a prior art. It is a figure which shows the block which can apply the skip determination by a prior art. It is a figure which shows the block which can apply the skip determination by another prior art. It is a figure which shows the example of the skip determination in the to-be-photographed object boundary by another prior art.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
In a prediction mode determination method according to an embodiment of the present invention, when determining a prediction mode of a block to be encoded in a moving image encoding scheme in which a picture that is image data of a moving image is encoded in units of blocks, an encoding target Prior to block skip determination, the features of the optimal skip block and the encoding target block are compared. Then, only when it is estimated that the coding efficiency is not deteriorated even if the skip determination is performed, the optimal skip block is used as a reference block and the skip determination is performed. Note that the optimum skip block is a block that is encoded as the skip mode but is not determined to be the skip mode in the skip determination.

  The feature comparison between the optimum skip block and the encoding target block may be performed by a plurality of evaluation methods. In this embodiment, the optimum skip block is only used when a predetermined condition is satisfied by all the evaluation methods. Is determined from the reference candidate block to the reference block, and skip determination is performed. As an example of the evaluation method, there is a comparison between the motion vector of the optimum skip block (reference block) and the motion vector of the encoding target block. When the distance between these motion vectors is smaller than a predetermined threshold, it can be estimated that these two blocks belong to the same subject, and therefore it can be estimated that there is no deterioration in coding efficiency even if skip determination is performed.

  As another example of the evaluation method, the complexity index value of the optimum skip block (reference block) is compared with the complexity index value of the encoding target block, and the skip is performed when the complexity of the encoding target block exceeds. It can be considered to make a judgment. The complexity index value can use an activity related to the original image of each block. In general, the prediction residual energy tends to increase in prediction as the block becomes more complex. Therefore, even though the complexity of the encoding target block exceeds the optimal skip block, the cost when the encoding target block is encoded in the skip mode is lower than the skip determination threshold value. Even if encoding is performed in the mode, it is considered that the encoding efficiency does not deteriorate.

  In the following, an embodiment in which the present invention is applied to an encoding device (moving image encoding device) that outputs a bit stream compliant with HEVC (High Efficiency Video Coding) will be described. As the CU (Coding Unit) size, it is assumed that a variation of the four-layer CU size in which the LCU (Largest Coding Unit) is 64 × 64 pixels and the minimum CU size is 8 × 8 pixels can be used. In order to simplify the description of the embodiment, a PU (Prediction Unit) obtained by dividing a CU into smaller blocks is not used.

[First Embodiment]
FIG. 1 is a functional block diagram showing the configuration of the encoding device 1 according to this embodiment. As shown in the figure, the encoding apparatus 1 includes a skip determination unit 11, a block division determination unit 12, an inter prediction unit 13, an intra prediction unit 14, a switching unit 15, a conversion unit 16, a quantization unit 17, and entropy coding. 18, an inverse quantization unit 19, an inverse transform unit 20, a loop filter unit 21, and a reference image memory 22. The encoding device 1 receives the encoding target LCU in the encoding target picture of the moving image. On the other hand, an encoded stream corresponding to the LCU is an output of the encoding device 1. This is repeatedly executed for each LCU in the encoding target picture in the raster scan order, whereby the encoding target picture is encoded.

  In HEVC, LCU is divided | segmented into CU, intra prediction or inter prediction is performed in each CU, and the difference is encoded. Therefore, the block division determination unit 12 performs CU division of the encoding target LCU input to the encoding device 1, and further determines whether each CU uses intra prediction or inter prediction. The inter prediction unit 13 generates a prediction image of the CU by inter prediction according to the instruction of the block division determination unit 12. The intra prediction unit 14 generates a prediction image of the CU by intra prediction according to the instruction of the block division determination unit 12. The switching unit 15 switches the output between the prediction image of the CU generated by the inter prediction unit 13 and the prediction image of the CU generated by the intra prediction unit 14 according to the determination of the block division determination unit 12, and The predicted image is output. When the difference image between the image of the encoding target LCU input to the encoding device 1 and the prediction image of intra prediction or inter prediction of the encoding target LCU output from the switching unit 15 is calculated, the difference image is converted. Input to the unit 16.

  The conversion unit 16 converts the input difference image into conversion coefficients, and the quantization unit 17 quantizes the conversion coefficients converted by the conversion unit 16. The entropy encoding unit 18 performs entropy encoding on the quantized value obtained by the quantization unit 17 and outputs the encoded value from the encoding device 1 as an encoded stream. On the other hand, the inverse quantization unit 19 inversely quantizes the quantized value of the transform coefficient obtained in the quantization unit 17. The inverse transform unit 20 inversely transforms the result obtained by the inverse quantization in the inverse quantization unit 19. The loop filter unit 21 applies a loop filter to the sum of the image obtained by the inverse transform in the inverse transform unit 20 and the intra prediction or inter prediction predicted image of the encoding target LCU output from the switching unit 15. The reference image memory 22 stores an image to which the loop filter is applied by the loop filter unit 21. Note that the image created in this way is equivalent to a decoded image obtained by a decoder compliant with HEVC, and the image in the reference image memory 22 is the inter prediction of the later encoding target LCU in the inter prediction unit 13. Used as a reference image. The sum of the image inversely transformed by the inverse transform unit 20 and the prediction image of the LCU for intra prediction or inter prediction output from the switching unit 15 is the intra prediction of the encoding target LCU later in the intra prediction unit 14. Used as a reference image.

  The skip determination unit 11 performs skip determination to determine whether or not to encode the block in the skip mode before performing motion search in the encoding target block. In this embodiment, the skip determination is performed only in the LCU, and when it is determined by the skip determination that the LCU is encoded in the skip mode, processing related to a smaller CU (intra / inter prediction in each CU size, Determination of optimal CU partition by comparison of CU partitions) is not performed, thereby realizing a significant reduction in the amount of calculation. This flow is shown in FIG.

  FIG. 2 is a diagram illustrating a processing flow of the prediction mode and the optimal division determination for the processing target CU. In this embodiment, when it is not determined that the LCU is encoded as the skip mode by the skip determination, the block division determination unit 12 obtains the RD cost for all the CUs obtained by recursively dividing the LCU into four optimal CUs. Determine the split. The RD cost is given by the following equation (1).

RD cost = D + λR (1)

  Here, D and R are respectively the coding distortion and the generated code amount when a certain prediction mode (skip mode, intra prediction, etc.) is applied to the processing target CU. The encoding distortion is the sum of the square errors of the original image and the decoded image for each pixel of the processing target CU. λ is Lagrange's undetermined multiplier. The processing flow shown in the figure shows the flow in the processing target CU. The block division determination unit 12 further sets the processing target CU (subordinate CU) obtained by dividing the processing target CU into four parts. Is recursively called. That is, by executing this flow for the LCU, processing is performed for all the CU sizes, and optimal CU partitioning is determined. Operations of the skip determination unit 11, the block division determination unit 12, the inter prediction unit 13, and the intra prediction unit 14 will be described using this flow.

  First, in step S101, the skip determination unit 11 determines whether or not the processing target CU is an LCU. When the skip determination unit 11 determines that the processing target CU is an LCU (step S101: YES), the skip determination unit 11 proceeds to the process of step S102, and the processing target CU, that is, the LCU adjacent to the encoding target LCU on the left or on the left Checks whether is an optimal skip block. The optimal skip block is a block that has been determined to be encoded in the skip mode without performing the skip determination in step S103 to be described later, that is, as a result of comparison with other prediction modes by the flow after step S105, It is a block. When the skip determination unit 11 determines that neither the LCU adjacent to the encoding target LCU nor the LCU adjacent to the left are optimal skip blocks (step S102: NO), the encoding apparatus 1 performs processing after step S105 described later. I do. When the skip determination unit 11 determines that the upper or left LCU adjacent to the encoding target LCU is the optimal skip block (step S102: YES), the encoding device 1 performs the skip determination process in step S103.

  When the process proceeds to the skip determination process in step S103, the inter prediction unit 13 encodes the processing target LCU in the skip mode and calculates an RD cost when the processing target LCU is encoded in the skip mode. At this time, the inter prediction unit 13 calculates the RD cost when the pair of each reference image index and motion vector included in the merge candidate list generated in the encoding in the skip mode is used, and the RD cost is minimized. The reference image index and the motion vector are determined to be used in the skip mode encoding. The inter prediction unit 13 outputs the motion vector and the reference image index used in the skip mode determined as described above, and the corresponding RD cost to the skip determination unit 11. The skip determination unit 11 performs skip determination based on the inputs from these inter prediction units 13. Details of the operation of the skip determination unit 11 will be described later.

In step S104, the encoding apparatus 1 performs branching based on the result of the skip determination by the skip determination unit 11. When it is not determined that the encoding target LCU is encoded in the skip mode (step S104: NO), the encoding apparatus 1 proceeds to the process of step S105, and when it is determined that the encoding is performed in the skip mode (step S104). (S104: YES), the process proceeds to step S113.
When the process moves to step S113, the block division determination unit 12 determines that the “optimal division” of the encoding target LCU is to set the LCU to be a CU (that is, does not divide), and performs processing related to the CU. Complete. The block division determination unit 12 instructs the inter prediction unit 13 to output a prediction image obtained by encoding the encoding target LCU in the skip mode, and switches the switching unit 15 to the inter prediction unit 13 side.

  Next, the processing after step S105 will be described. The processing after step S105 is processing when the skip determination is not performed or when it is not determined that the encoding is performed in the skip mode in the skip determination, and various methods can be considered because it is out of the scope of the present embodiment. An example will be described below. In step S <b> 105, the inter prediction unit 13 generates a prediction image of the processing target CU in each prediction mode of inter prediction according to an instruction from the block division determination unit 12, and calculates the RD cost of each prediction mode. Here, each prediction mode of inter prediction is a skip mode and a normal inter mode that encodes a motion vector and a prediction residual. In the normal inter mode, motion search is also performed. In step S <b> 106, the intra prediction unit 14 generates a prediction image of the processing target CU by intra prediction according to an instruction from the block division determination unit 12, and calculates an RD cost for intra prediction. In step S107, the RD costs of all prediction modes obtained in steps S105 and S106 are input to the block division determination unit 12, and these RD costs are compared. The block division determination unit 12 determines the prediction mode that minimizes the RD cost as the optimum prediction mode of the processing target CU.

In step S108, the block division determination unit 12 determines whether or not the processing target CU has a minimum CU size (8 × 8 pixels). If it is the minimum CU size, further division is impossible. Therefore, when the block division determination unit 12 determines that the processing target CU has the minimum CU size (step S108: YES), the block division determination unit 12 performs the process of step S113 and determines that the processing target CU is not divided into smaller CUs. While setting as “optimal division”, the RD cost is saved, and the processing related to this CU is completed.
When it is determined that the processing target CU is not the minimum CU size (step S108: NO), in the processing of step S109, the block division determination unit 12 sets the processing target CU into four CUs (subordinate CUs) as processing targets. Call this flow (the entire flow in FIG. 2). At this time, the four lower CUs are called in the order of upper left, upper right, lower left, and lower right. Note that this is done sequentially, and after the upper left lower CU flow is completed, the upper right lower CU flow is called (the processing of the left / upper / upper left CU necessary for encoding a certain CU is completed. To make it happen). Then, after this flow in the four subordinate CUs is completed, the block division determination unit 12 proceeds to the process of step S110.

In step S110, the block division determination unit 12 compares the RD cost of the optimum prediction mode of the processing target CU obtained in step S107 with the sum of the RD costs corresponding to the optimum division obtained in the flows of the four lower CUs. To do. Note that the RD cost corresponding to the optimal division of the lower CU is stored in the process of step S112 or step S113 in the flow recursively executed with the lower CU as a processing target. In step S111, the block division determination unit 12 branches based on the comparison result. When the RD cost of the optimal prediction mode of the processing target CU is smaller (step S111: YES), the block division determination unit 12 proceeds to the process of step S113, and the optimal division is determined not to divide the processing target CU into smaller CUs. And the RD cost of the optimal prediction mode of the processing target CU is stored as the RD cost corresponding to the optimal division. In other cases (step S111: NO), the block division determination unit 12 proceeds to the process of step S112, and the division (an example shown in FIG. 3 described later) obtained by integrating the optimum divisions in the four lower CUs is processed. And the corresponding RD cost (the sum of the RD costs of the optimal division in the four subordinate CUs) is stored.
The block division determination unit 12 instructs the inter prediction unit 13 and the intra prediction unit 14 to output a prediction image in the optimal prediction mode for each optimal division CU, and switches the switching unit 15 based on the optimal prediction mode.

  FIG. 3 is a diagram illustrating an example of processing target CU partitioning that integrates optimal partitioning in four subordinate CUs. In the example shown in the figure, for the lower CU1 and the lower CU4 constituting the processing target CU, the division into smaller CUs is the optimum division, and for the lower CU2 and the lower CU3, it is determined that the lower CU is the optimum division. Has been. The optimum division of the processing target CU is obtained by integrating these optimum divisions of the lower CU1 to CU4.

Next, details of the configuration of the skip determination unit 11 in the encoding device 1 shown in FIG. 1 and details of the skip determination processing in step S103 in the processing flow shown in FIG. 2 will be described.
FIG. 4 is a block diagram illustrating a detailed functional configuration of the skip determination unit 11. As shown in the figure, the skip determination unit 11 includes a block feature comparison unit 111 and a cost comparison unit 112. The block feature comparison unit 111 compares the feature amount of the encoding target LCU (input image) with the feature amount of the upper or left optimal skip block adjacent to the encoding target LCU, and performs skip determination for the encoding target LCU. Determine whether to do it. If the block feature comparison unit 111 determines that the skip determination is to be performed, the cost comparison unit 112 compares the RD cost when the encoding target LCU is encoded in the skip mode and the threshold value, and performs encoding in the skip mode. A skip decision is made whether or not.

  FIG. 5 is a diagram showing a detailed processing flow of the skip determination processing in step S103 in the processing flow shown in FIG. The processing of the block feature comparison unit 111 and the cost comparison unit 112 of the skip determination unit 11 illustrated in FIG. 4 will be described according to the skip determination processing flow illustrated in FIG. When this skip determination process is performed, the upper or left LCU adjacent to the encoding target LCU is an optimal skip block (hereinafter referred to as “optimal skip LCU”).

  In step S201, the block feature comparison unit 111 calculates the feature amount of the encoding target LCU and the feature amount of the optimum skip LCU. In step S202, the block feature comparison unit 111 compares the feature amount of the encoding target LCU with the feature amount of the optimal skip LCU, and determines whether or not to perform skip determination for the encoding target LCU. If both the upper and left LCUs to be encoded are optimal skip LCUs, here, the feature amounts of the left optimal skip LCU are compared. Variations in processing when both the upper and left LCUs are optimal skip LCUs will be described in a second embodiment to be described later. In this embodiment, a comparison using two features, such as a comparison using an activity and a comparison using a motion vector, is performed, and when a predetermined condition (skip determination execution subcondition) is satisfied in each, ”Is satisfied and skip determination processing after step S204 is performed. In this embodiment, two feature amounts are compared. However, one feature amount may be compared, or an arbitrary number of two or more feature amounts may be compared. In this embodiment, (Comparison 1) and (Comparison 2) are performed according to the following two characteristics.

(Comparison 1) Comparison using activities (determination of whether or not the skip determination execution subcondition 1 is satisfied):
The activity is an index indicating the texture complexity of the target block. Specifically, the target block is divided into a plurality of sub-blocks, and the variance of the pixel values of the original image regarding each sub-block is obtained. Then, either the average / minimum value or the maximum value of the variance of the sub-blocks in the target block is set as the activity. When the target block itself is considered as a sub-block (that is, when the target block is composed of one sub-block), the distribution itself of the target block is an activity. In the present embodiment, for example, the average value of the variance of each sub-block when the LCU is equally divided into four sub-blocks can be used as the activity. Therefore, in step S201, the block feature comparison unit 111 uses the input encoding target LCU and the original image of the optimal skip LCU stored in the reference image memory 22, and performs the encoding target LCU by the method described above. Activity ACT_enc and optimum skip LCU activity ACT_ref. The block feature comparison unit 111 determines that the skip determination execution subcondition 1 is satisfied when the condition shown in the following expression (2) is satisfied.

ACT_enc> ACT_ref (2)

(Comparison 2) Comparison using motion vectors (determination of whether or not the skip determination execution subcondition 2 is satisfied):
In the skip determination execution subcondition 2, the motion vector of the encoding target LCU is compared with the motion vector of the optimum skip LCU. However, it is also checked whether or not the reference images referenced by the respective motion vectors are the same (whether or not the reference image indexes are the same). This check is also performed by the block feature comparison unit 111, and the motion vector and the reference image index input from the inter prediction unit 13 are used. As a motion vector and a reference image index to be used, it is conceivable to use a motion vector and a reference image index used for encoding in the skip mode in each LCU. Alternatively, for the optimal skip LCU, it is possible to use a motion vector and a reference image index in a prediction mode other than the skip mode. That is, any motion vector or reference image index obtained as long as it represents motion in the LCU may be used. Here, the motion vector and reference image index of the encoding target LCU are MV_enc and REFIDX_enc, respectively, and the motion vector and reference image index of the optimal skip LCU stored in the reference image memory 22 are MV_ref and REFIDX_ref, respectively. Each of these is a two-dimensional vector having x and y components as elements. As the comparison of motion vectors, determination of the degree of coincidence of motion vectors by comparing whether or not MV_enc and MV_ref are the same vector, determination by Euclidean distance between MV_enc and MV_ref, and the like can be considered. In this embodiment, the Euclidean distance is used, and when the following two conditions 1 and 2 are satisfied, the block feature comparison unit 111 determines that the skip determination execution subcondition 2 is satisfied.

(Condition 1): REFIDX_enc and REFIDX_ref are equal.
(Condition 2): The Euclidean distance between MV_enc and MV_ref is less than or equal to a predetermined threshold.

  In step S202, the block feature comparison unit 111 branches conditionally depending on whether or not the skip determination execution condition is satisfied. When one or both of the above-described skip determination execution subconditions 1 and 2 are not satisfied (step S202: NO), the block feature comparison unit 111 proceeds to the process of step S203 because the skip determination execution condition is not satisfied, and skips. The cost comparison unit 112 is notified of skip determination presence / absence information indicating that the determination is not performed. Since the cost comparison unit 112 indicates that the skip determination presence / absence information indicates that skip determination is not performed, the cost comparison unit 112 notifies the block division determination unit 12 as skip determination information that the encoding target LCU is not determined to be encoded in the skip mode. , The flow is completed. On the other hand, when both of the skip determination execution subconditions 1 and 2 are satisfied (step S202: YES), the block feature comparison unit 111 determines that the skip determination execution condition is satisfied and indicates that the skip determination is performed. The cost comparison unit 112 is notified of skip determination presence / absence information. If the skip determination presence / absence information indicates that the skip determination is performed, the cost comparison unit 112 proceeds to the process of step S204 and performs the skip determination.

  In step S204, first, the cost comparison unit 112 sets a skip determination threshold. It is assumed that the skip determination threshold is obtained by multiplying the RD cost when the upper or left LCU adjacent to the encoding target LCU is encoded in the skip mode by comparing the feature amount as described above by a coefficient. That is, when the RD cost is RD_ref and the coefficient is scale, the skip determination threshold is expressed by the following equation (3).

Skip determination threshold = RD_ref × scale (3)

  In step S204, the cost comparison unit 112 compares the RD cost when the encoding target LCU is encoded in the skip mode with the skip determination threshold. In step S205, if the cost comparison unit 112 determines that the RD cost is below the skip determination threshold (step S205: YES), the process proceeds to step S206. In step S206, the cost comparison unit 112 determines to encode the encoding target LCU in the skip mode, notifies the block division determination unit 12 of the determination result as skip determination information, and completes the flow. On the other hand, when the cost comparison unit 112 determines that the RD cost is equal to or higher than the skip determination threshold (step S205: NO), the cost comparison unit 112 proceeds to the process of step S203. In step S203, the cost comparison unit 112 notifies the block division determination unit 12 that the encoding target LCU is not determined to be encoded in the skip mode, as skip determination information, and completes the flow.

[Second Embodiment]
In the first embodiment described above, when both the left and right LCUs adjacent to the encoding target LCU are the optimum skip LCUs in step S102 of the flow of FIG. 2, the left is unconditionally left in step S103. Skip determination was made with reference to the LCU. This embodiment shows another embodiment when both the upper and left adjacent LCUs are optimal skip LCUs. In addition, this embodiment can be easily applied to a case where a plurality of blocks are referred to, such as a block to be referred to is adjacent and includes not only the left block but also an adjacent upper left block. Hereinafter, the difference from the first embodiment will be mainly described.

FIG. 6 shows a processing flow of skip determination according to this embodiment. In the present embodiment, instead of the processing shown in the processing flow for skip determination according to the first embodiment shown in FIG. 5, processing shown in the processing flow shown in FIG. This embodiment will be described with reference to the flowchart shown in FIG.
In step S301, the block feature comparison unit 111 compares the feature amount of the encoding target LCU with the feature amount of the adjacent optimal and left optimal skip LCUs, and determines whether or not to perform skip determination for the encoding target LCU. Determine. The feature amounts and skip determination execution subconditions used in this embodiment are the same as those in the first embodiment. That is, the block feature comparison unit 111 evaluates the activity difference for the left optimal skip LCU adjacent to the encoding target LCU to determine whether or not the corresponding skip determination execution sub-condition is satisfied, and the motion vector and The difference in the reference image index is evaluated to determine whether or not the corresponding skip determination execution sub-condition is satisfied. Furthermore, the block feature comparison unit 111 similarly performs the same determination for the upper optimal skip LCU adjacent to the encoding target LCU. In step S302, the block feature comparison unit 111 determines whether or not the upper and left optimum skip LCUs satisfy the skip determination execution condition.

  If there is no optimal LCU block that satisfies the skip determination execution condition (step S302: NO), the block feature comparison unit 111 proceeds to the process of step S303. In step S303, the block feature comparison unit 111 notifies the cost comparison unit 112 of skip determination presence / absence information indicating that skip determination is not performed. Since the cost comparison unit 112 indicates that the skip determination presence / absence information indicates that skip determination is not performed, the cost comparison unit 112 notifies the block division determination unit 12 as skip determination information that the encoding target LCU is not determined to be encoded in the skip mode. , The flow is completed.

  On the other hand, when one or both of the upper optimal skip LCU and the left optimal skip LCU satisfy the skip determination execution condition (step S302: YES), the block feature comparison unit 111 indicates skip determination indicating that skip determination is performed. The cost comparison unit 112 is notified of the presence / absence information and the optimal skip LCU that satisfies the skip determination execution condition. When the skip determination presence / absence information indicates that the skip determination is performed, the cost comparison unit 112 proceeds to the process of step S304.

  In step S304, the cost comparison unit 112 determines whether there is one optimal skip LCU that satisfies the skip determination execution condition. If the block feature comparison unit 111 determines that there is one optimal skip LCU that satisfies the skip determination execution condition (step S304: YES), the process proceeds to step S305. In step S <b> 305, the cost comparison unit 112 calculates a skip determination threshold using the optimal skip LCU that satisfies the skip determination execution condition as a reference LCU. Since this process is the same as that of the first embodiment, a description thereof will be omitted. After calculating the skip determination threshold, the cost comparison unit 112 proceeds to the process of step S307.

  On the other hand, when the block feature comparison unit 111 determines that there are a plurality of optimal skip LCUs that satisfy the skip determination execution condition (in this embodiment, both the upper and left optimal skip LCUs) (step S304: NO), The process proceeds to step S306. In step S306, the cost comparison unit 112 calculates a skip determination threshold from a statistic based on the skip mode RD cost of the optimal skip LCU. As specific statistics, the average value, median value, minimum value, etc. of the RD cost in the skip mode of the optimal skip LCU that satisfies the skip determination execution condition can be considered. For example, when the minimum value RD_min is used, the skip determination threshold is given by the following equation (4).

Skip determination threshold = RD_min × scale (4)

After calculating the skip determination threshold, the cost comparison unit 112 proceeds to the process of step S307. The processing after step S307 is the same as the processing after setting the skip determination threshold in step S204 of FIG.
That is, in step S307, the cost comparison unit 112 compares the RD cost when the encoding target LCU is encoded in the skip mode with the skip determination threshold calculated in step S305 or step S306. If it is determined in step S308 that the RD cost is lower, the cost comparison unit 112 proceeds to the process of step S309. The cost comparison unit 112 determines to encode the encoding target LCU in the skip mode, and notifies the block division determination unit 12 of the determination result as skip determination information. On the other hand, if it is determined in step S308 that the RD cost is greater than or equal to the skip determination threshold, the cost comparison unit 112 proceeds to the process in step S303. The cost comparison unit 112 notifies the block division determination unit 12 as skip determination information that the encoding target LCU is not determined to be encoded in the skip mode.

[Third Embodiment]
In HEVC, encoding can be performed in skip mode in each CU size. In the first and second embodiments, skip determination is performed only for the LCU. In these embodiments, the optimum skip block to be referred to is only the LCU. In the present embodiment, skip determination is performed for each CU size, and various sizes of the optimum skip block to be referred to are allowed. Hereinafter, the difference from the first and second embodiments will be mainly described.

The processing flow of the prediction mode and optimal division determination for the processing target CU in the first and second embodiments is as shown in FIG. 2, but in this embodiment, skip determination may be executed for each CU size. Therefore, the flow is slightly different.
FIG. 7 is a diagram illustrating a processing flow of the prediction mode and the optimal division determination for the processing target CU according to the present embodiment. The processing flow shown in FIG. 2 differs from the processing flow of the first embodiment shown in FIG. 2 in that, regardless of the size of the processing target CU, in step S401 shown in FIG. In other words, it is determined whether or not there is an optimal skip block. Here, the optimum skip block is a CU (hereinafter referred to as “optimum skip CU”), and the CU size is arbitrary.

  FIG. 8 shows an example of a situation where there is an optimal skip CU having a different size from the processing target CU on the left or above. In the example shown in the figure, the 8 × 8 pixel CU adjacent to the left of the 16 × 16 pixel processing target CU and in the lower hierarchy than the processing target CU is the optimal skip CU.

  If it is determined in step S401 in FIG. 7 that the optimum skip CU exists on the left or above adjacent (step S401: YES), the process proceeds to step S402 described later, and skip determination is performed. On the other hand, when it is determined that there is no optimal skip CU on the left or above adjacent (step S401: NO), the process proceeds to step S404 and subsequent steps. The flow after step S404 is the same as the flow after step S105 shown in FIG.

In the skip determination of step S402, the skip determination unit 11 reads the LCU as a CU, and executes the skip determination processing flow of the second embodiment shown in FIG. 6 except for the following points.
In step S301, the block feature comparison unit 111 compares the feature amounts of the encoding target CU and the optimum skip CU. However, the CU sizes are different, and the feature value value depends on the CU size. In the case of using something that changes, the comparison is performed after normalizing or the like (scaling to a value independent of size) for the feature amount. Also in steps S305 and S306, the RD cost needs to be normalized. In the case of the RD cost, it is necessary to normalize only the encoding distortion D. Therefore, when calculating the normalized RD cost RD_s of a certain CU, the cost comparison unit 112 uses the number of pixels pnum_CU of the CU to scale the encoding distortion D as shown in the following equation (5). Then, the normalized encoding distortion D_s is calculated.

D_s = D / pnum_CU (5)

  The normalized RD cost RD_s is given by the following equation (6).

RD_s = R + D_s (6)

  The processing after step S403 after skip determination in step S402 in FIG. 7 is the same as the processing after step S104 shown in FIG. That is, when it is not determined that the encoding target CU is to be encoded in the skip mode (step S403: NO), the encoding apparatus 1 proceeds to the process of step S404, and when it is determined that the encoding is performed in the skip mode. (Step S403: YES), the process proceeds to Step S412. The process of step S412 is the same as the process of step S113 shown in FIG.

  In step S401 in FIG. 7, the optimum skip block whose presence is to be determined has an arbitrary size, but the optimum skip CU whose CU size is larger than a predetermined size exists on the left or above adjacent to the processing target CU. It may be determined whether or not. For example, when the predetermined size is 8 × 8 pixels, in the example shown in FIG. 8, since the optimal skip CU larger than 8 × 8 pixels does not exist on the left or above adjacent to the processing target CU, it is determined as NO, and step S404 Move on to processing. The reason for providing such a size restriction is as follows. That is, there is a high possibility that a region encoded with a small CU cannot be predicted with high accuracy in both inter prediction and intra prediction. In such a case, the RD cost in all prediction modes including the skip mode is significantly high. Thus, in order not to refer to the optimum skip CU having a remarkably large RD cost, it is effective not to refer to a small size CU.

[Other variations]
In the embodiment described above, skip determination is performed using the RD cost, but the same effect can be obtained even if the skip determination is performed by replacing the coding distortion D with the RD cost. More generally, the same effect can be obtained if the cost is for evaluating the coding efficiency, such as the RD cost or the coding distortion.
In the above-described embodiment, it is checked in step S102 of FIG. 2 whether or not the LCU adjacent to the encoding target LCU or the left side is the optimal skip block. It may be determined whether the LCU adjacent to the left is a block encoded in the skip mode. That is, the LCU adjacent to the top or left of the encoding target LCU is a block that has been determined to be the skip mode in the skip determination in step S103, or the result of comparison with another prediction mode by the flow after step S105, It is determined whether or not the block is determined to be in the skip mode. Similarly, in step S401 in FIG. 7, it may be determined whether there is a block encoded in the skip mode on the left or above adjacent to the processing target CU. Although it is more reliable to determine whether or not it is an optimal skip block, whether or not it is a block coded in the skip mode including the block determined to be the skip mode by the skip determination. Even in the determination, the reliability of the skip determination is improved as compared with the prior art.

[effect]
According to this embodiment, unlike the prior art B, the features of the reference block and the encoding target block are compared prior to skip determination. As a result, only when it is estimated that the coding efficiency is not deteriorated even if the skip determination is performed whether or not the encoding is performed in the skip mode that enables the motion search to be omitted, the skip determination for the encoding target block is performed. Do. For this reason, even if the calculation amount is reduced by using one reference block, the deterioration of the coding efficiency can be reduced as compared with the conventional technique B.

  You may make it implement | achieve the encoding apparatus 1 in embodiment mentioned above with a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be read into a computer system and executed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory inside a computer system serving as a server or a client in that case may be included and a program held for a certain period of time. Further, the program may be a program for realizing a part of the above-described functions, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system. You may implement | achieve using programmable logic devices, such as FPGA (Field Programmable Gate Array).

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

  The present invention can be used for an encoder that encodes a moving picture picture in units of blocks.

DESCRIPTION OF SYMBOLS 1 ... Coding apparatus, 11 ... Skip determination part, 12 ... Block division | segmentation determination part, 13 ... Inter prediction part, 14 ... Intra prediction part, 15 ... Switching part, 16 ... Conversion part, 17 ... Quantization part, 18 ... Entropy Encoding unit, 19 ... Inverse quantization unit, 20 ... Inverse transformation unit, 21 ... Loop filter unit, 22 ... Reference image memory, 111 ... Block feature comparison unit, 112 ... Cost comparison unit

Claims (7)

A prediction mode determination method executed by a moving image encoding apparatus that determines a prediction mode in block units and encodes image data of a moving image,
For one or more feature values, the difference between the feature values of the encoding target block in the image data and a reference candidate block adjacent to the encoding target block and encoded in the skip mode is calculated. A block feature comparison step;
The skip determination execution presence / absence determination step for determining whether or not the difference between the feature amounts calculated in the block feature comparison step satisfies a condition defined for the feature amount, and the feature in the skip determination execution presence / absence determination step When it is determined that a difference in amount satisfies the condition defined for the feature amount, the reference candidate block is set as a reference block, a threshold obtained from the cost of the skip mode in the reference block, and the encoding target block A skip determination step of determining whether or not to set the prediction mode of the encoding target block to the skip mode by comparing the cost when the skip mode is set;
A prediction mode determination method characterized by comprising: The feature amount includes a motion vector,
The condition includes a condition defined for a motion vector,
In the block feature comparison step, the calculation of the feature amount difference includes calculation of a degree of coincidence of motion vectors between the reference candidate block and the encoding target block,
In the skip determination execution presence / absence determination step, when the degree of coincidence between the reference candidate block and the encoding target block is within a predetermined range, it is determined that the condition defined for the motion vector is satisfied.
The prediction mode determination method according to claim 1. The feature amount includes a complexity index value,
The conditions include conditions defined for the complexity index value,
In the block feature comparison step, the calculation of the difference in the feature amount includes calculation of a magnitude of a complexity index value between the reference candidate block and the encoding target block,
In the skip determination execution presence / absence determination step, when the complexity of the encoding target block exceeds the complexity of the reference candidate block, it is determined that the condition defined for the complexity index value is satisfied.
The prediction mode determination method according to claim 1. When there are a plurality of reference candidate blocks that are adjacent to the encoding target block and encoded in the skip mode,
In the block feature comparison step, a difference in the feature amount between the encoding target block and each of the plurality of reference candidate blocks is calculated,
In the skip determination execution presence / absence determining step, among the plurality of reference candidate blocks, a reference candidate block in which a difference in the feature amount calculated in the block feature comparison step satisfies a condition defined for the feature amount is referred to as a reference block And
The prediction mode determination method according to any one of claims 1 to 3, wherein the prediction mode is determined. In the block feature comparison step, among the blocks that are adjacent to the current block to be encoded and are encoded in the skip mode, a block larger than a predetermined size is set as a reference candidate block.
The prediction mode determination method according to any one of claims 1 to 4, wherein the prediction mode is determined. A moving image encoding apparatus that determines a prediction mode in block units and encodes image data of a moving image,
For one or more feature quantities, the difference between the feature quantities of the encoding target block in the image data and a reference candidate block that is adjacent to the encoding target block and is encoded in the skip mode is calculated. A block feature comparison unit that determines whether the calculated difference between the feature values satisfies a condition defined for the feature values;
When the block feature comparison unit determines that the feature amount difference satisfies the condition defined for the feature amount, the reference candidate block is set as a reference block, and a threshold value obtained from the cost of the skip mode in the reference block And a cost comparison unit that determines whether the prediction mode of the encoding target block is set to the skip mode by comparing the cost when the encoding target block is set to the skip mode;
A moving picture encoding apparatus comprising: In a computer used as a video encoding device,
The prediction mode determination program for performing each step of the prediction mode determination method of any one of Claims 1-5.

Images