JP6339977B2 - Video encoding apparatus and video encoding program - Google Patents

Description

  The present invention relates to a video encoding device and a video encoding program that perform video encoding using motion prediction.

  MPEG-2, MPEG-4, and MPEG-4 / AVC are often used as video encoding technologies. Recently, the next generation video encoding standard, HEVC (High Efficiency Video Coding), has been standardized. Is expected to spread. In the video coding standard, intra-frame coding that performs coding using information closed in one picture and inter-screen coding that performs coding using a plurality of temporally continuous pictures are used. Yes. In inter-frame coding, motion prediction processing is performed to reduce the difference value between the screens, and the information amount is reduced by encoding the difference value and the motion vector information.

  In motion prediction in inter-frame coding, a motion vector is detected and motion prediction is performed to reduce the information to be encoded, so that a large amount of calculation is required for motion vector detection processing. As a technique for obtaining a highly accurate motion vector with a small amount of calculation, a hierarchical motion vector detection method or the like is used. In the hierarchical motion vector detection method, a preliminary motion prediction process that captures the rough motion of the encoding target block is performed, and a higher-precision motion prediction is performed again based on the search result obtained from the preliminary motion prediction process. The amount of calculation required for motion vector search is reduced.

  In the motion prediction process, a motion vector between pictures is detected for each coding block. In MPEG-4 / AVC and HEVC, the size of each coding block can be selected from a plurality of block size candidates. it can. In MPEG-4 / AVC, encoding processing is performed in units of 16 × 16 pixels called macroblocks, and three types of block size candidates of 16 × 16, 8 × 8, and 4 × 4 can be selected as prediction processing units. In HEVC, encoding processing is performed with four types of block size candidates of 64 × 64, 32 × 32, 16 × 16, and 8 × 8 in units of coding units (CU), and an additional CU is used for inter-screen prediction. The process is divided into different shapes and processed in units called prediction units (PU). By performing motion vector search with each block size, more efficient inter-frame coding can be performed.

  On the other hand, the process of determining the most efficient inter-frame prediction method by performing a motion vector search considering all of the plurality of block size candidates is compared with the case of the motion prediction process limited to a single block size. There is a problem that the amount of calculation processing increases significantly. On the other hand, instead of sequentially integrating costs starting from a small block size, the cost of any two block sizes is calculated and compared to determine the block size at an early stage for motion prediction. A technique for reducing the amount of calculation required has been devised (see, for example, Patent Document 1).

JP 2014-127891 A

  However, since the method described in Patent Document 1 uses an RD (Rate Distortion) cost, an encoding result of the target block is necessary for cost value acquisition, and when used in combination with the hierarchical motion vector detection method, Since the block size is not narrowed down until motion search, the amount of calculation becomes enormous. Furthermore, since the amount of calculation required for motion prediction varies depending on the encoding target block in the video encoding device described in Patent Document 1, from the viewpoint of hardware implementation, it leads to an increase in hardware scale and a decrease in hardware usage efficiency. There's a problem.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a video encoding device and a video encoding program capable of reducing the amount of calculation while suppressing a decrease in encoding efficiency. .

  One aspect of the present invention is a video encoding apparatus that performs temporal prediction of a picture to be encoded, performs motion prediction for each encoding block, and encodes a differential signal using a temporal correlation of a picture of an input video signal. Then, using the reference picture of the encoding target picture and the decoded picture signal of the motion prediction destination, a pre-motion prediction process for capturing a rough motion of the target encoding block is performed for each of all encoding block sizes, and each encoding block size is determined. A pre-motion prediction processing means for calculating a pre-motion prediction processing cost in the block, and a block size candidate cost in the first coding block size candidate and a second coding block size candidate based on the pre-motion prediction processing cost Block size candidate cost means for determining a block size candidate cost, and the encoded input video signal A cost comparison offset value is set based on parameter determination means for determining a motion statistical parameter based on the motion vector information of the picture, the motion statistical parameter, and the depth of the hierarchy in the reference structure of the encoding target picture. Based on the offset value setting means, the block size candidate cost in the first encoded block size candidate, the block size candidate cost in the second encoded block size candidate, and the cost comparison offset value, the code Block size candidate determining means for determining a coding block size candidate of the encoding target picture as one of the first block size candidate and the second block size candidate from all block size candidates, and the determined coding block Using the size candidates, the encoding target picture A video encoding apparatus and a motion search unit to perform block size determination and motion vector decision performs motion prediction processing.

  One aspect of the present invention is the video encoding device, wherein the parameter determination unit calculates all motion vectors in a specified encoded picture of the input video signal, and sums absolute values of the motion vectors. Are determined as motion statistics parameters of the picture to be encoded.

  One aspect of the present invention is the video encoding device, wherein the parameter determining unit calculates all motion vectors in a specified encoded picture of the input video signal, and calculates a variance value of the motion vector. It is determined as a motion statistical parameter of the picture to be encoded.

  One aspect of the present invention is the video encoding device, wherein the block size candidate determining unit calculates a sum of a cost of a motion vector and the cost comparison offset value in the second block size candidate. If the cost of the first block size candidate is determined to be larger as a result of comparing the sum and the cost of the motion vector in the first block size candidate, the block size candidate of the encoding target picture Is the second block size candidate, and if it is determined that the cost of the first block size candidate is smaller, the block size candidate of the coding target picture is set as the first block size candidate.

  One aspect of the present invention is a video encoding program for causing a computer to function as the video encoding device.

  According to the present invention, in a video encoding system having a motion prediction unit in which there are a plurality of encoding block size candidates, based on the layer depth in the reference structure of the encoding target picture and the motion vector information of the encoded picture. By appropriately determining the block size candidates, it is possible to reduce the amount of calculation while suppressing a decrease in encoding efficiency.

It is a block diagram which shows the structure of the video coding apparatus by one Embodiment of this invention. It is a block diagram which shows the structure of the inter prediction process part 102 shown in FIG. It is a flowchart which shows operation | movement of the inter prediction process part 102 shown in FIG. It is a figure which shows the reference structure between screens in random access encoding mode. It is a block diagram which shows the structure of the block size candidate determination part 202 shown in FIG. 6 is a flowchart showing an operation of a block size candidate determination unit 202 shown in FIG. 5. It is a flowchart which shows the detailed operation | movement of the candidate cost production | generation part 301 shown in FIG. It is a flowchart which shows the detailed operation | movement of the offset determination part 302 shown in FIG. It is a flowchart which shows the detailed operation | movement of the candidate cost comparison part 303 shown in FIG.

  Hereinafter, a video encoding apparatus according to an embodiment of the present invention will be described with reference to the drawings. The “encoded block” used below is MPEG-2 or H.264. The H.264 / AVC standard indicates a macro block, and HEVC indicates a coding unit (CU) or a prediction unit (PU). FIG. 1 is a block diagram showing a configuration of a video encoding apparatus according to the embodiment. In the video encoding apparatus 100 shown in FIG. 1, the inter prediction process 102 is a part different from the prior art, and the other part is H.264. This is the same as the conventional general configuration used as a video encoding device such as H.264 / AVC or HEVC.

  The video encoding apparatus 100 receives a video signal (original image) to be encoded as an input, divides a picture of the input video signal into blocks, encodes each block, and outputs the bit stream as an encoded stream. For this encoding, the prediction residual signal generation unit 103 obtains a difference between the input video signal and the prediction signal output from the intra prediction processing unit or the inter prediction processing unit, and outputs it as a prediction residual signal.

  The transform / quantization processing unit 104 receives the prediction residual signal, performs orthogonal transform such as discrete cosine transform on the input prediction residual signal, quantizes the transform coefficient, and the quantized transform coefficient Is output. The entropy encoding processing unit 105 receives the quantized transform coefficient as input, entropy-encodes the input quantized transform coefficient, and outputs it as an encoded stream.

  On the other hand, the quantized transform coefficient is also input to the inverse quantization / inverse transform processing unit 106, where inverse quantization and inverse orthogonal transform are performed, and a prediction residual decoded signal is output. The decoded signal generation unit 107 adds the prediction residual decoded signal output from the inverse quantization / inverse conversion processing unit 106 and the prediction signal output from the intra prediction processing unit 101 or the inter prediction processing unit 102 to perform encoding. The decoded signal of the encoding target block is generated. This decoded signal is input to the loop filter processing unit 108 for use as a reference image by the inter prediction processing unit 102. The loop filter processing unit 108 performs filtering processing for reducing the coding distortion on the decoded signal, and inputs the image after the filtering processing to the inter prediction processing unit 102 as a reference image.

  In the present embodiment, the inter prediction processing unit 102 shown in FIG. 1 performs motion search processing by appropriately narrowing down block size candidates, thereby reducing the processing amount required for the inter prediction processing without reducing the coding efficiency. Is.

  The following description will be described as an embodiment based on the HEVC standard. The following embodiment is a method of narrowing down from four types of block size candidates of 64 × 64, 32 × 32, 16 × 16, and 8 × 8 in HEVC to three types of block size candidates. The processing in the embodiment is executed in units of 64 × 64 areas. The cost value used in the following embodiments refers to the sum of absolute differences (SAD) between the original image and the reference image indicated by the motion vector, or the sum of absolute differences obtained by performing two-dimensional Hadamard transformation on the difference values. (SATD) and a value represented by the sum of motion vector cost values obtained by simply estimating the amount of code generated when the motion vector is encoded.

  Next, the configuration and operation of the inter prediction processing unit 102 illustrated in FIG. 1 will be described. FIG. 2 is a block diagram illustrating a configuration of the inter prediction processing unit 102 illustrated in FIG. 1. The inter prediction processing unit 102 includes a pre-motion prediction processing unit 201 that predicts pre-motion, a block size candidate determination unit 202 that determines block size candidates, and a motion search processing unit 203 that performs motion search.

  Next, the operation of the inter prediction processing unit 102 shown in FIG. 2 will be described with reference to FIG. FIG. 3 is a flowchart showing the operation of the inter prediction processing unit 102 shown in FIG. First, the prior motion prediction processing unit 201 receives an input video signal (original image) and a decoded signal input from the loop filter processing unit 108 as input, and performs a prior motion search for capturing a rough motion of the input video signal (step S1). ).

  The prior motion prediction processing unit 201 performs a motion search for each block size in a target 64 × 64 region in the input video signal, calculates a motion vector for each block size, and a cost value associated therewith, and outputs the motion vector. The cost value of the 64 × 64 block size is a cost value when the target 64 × 64 region is processed as 64 × 64 blocks without being divided. The cost value of the 32 × 32 block size is a cost value in four 32 × 32 blocks when the target 64 × 64 region is divided into two parts in the vertical and horizontal directions. The cost value of the 16 × 16 block size is each cost value in 16 16 × 16 blocks when the target 64 × 64 region is divided into four in each of the vertical and horizontal directions. The cost value of the 8 × 8 block size is each cost value in 64 8 × 8 blocks when the target 64 × 64 region is divided into 8 parts in the vertical and horizontal directions. Further, the prior motion prediction processing unit 201 calculates TemporalID from the frame number of the input video signal (original image) as encoded information, and outputs the encoded information to the block size candidate determining unit 202.

  Next, the block size candidate determination unit 202 encodes the cost value of each block size calculated by the prior motion prediction processing unit 201 and the encoded past frame calculated by the motion search processing unit 203. The motion vector statistical value in a specific case in the picture and the coding information of the picture are input, and the block size candidate narrowing process is performed (step S2). The narrowing-down process is a process for reducing the number of candidates. For example, as will be described later, this refers to making a plurality of block size candidates into a group of two candidates and determining one of the candidates. In the description to be described later, an example in which two candidates are grouped and determined as one of the candidates will be described. However, the processing is not limited to this as long as the number of candidates is reduced. The motion vector for each block size and the block size candidates narrowed down by the block size candidate determination unit 202, which are the results of searching by the prior motion prediction processing unit 201, are output to the motion search processing unit 203, respectively.

  Next, the motion search processing unit 203 performs a motion search process with the size within the candidates narrowed down by the block size candidate determination unit 202 based on the result of the search by the prior motion prediction processing unit 201 (step S3). The finally determined motion vector information and the prediction difference image are output to the prediction residual signal generation unit 103. Further, only when the picture to be encoded is specific, the motion search processing unit 203 calculates a statistical value of the determined motion vector, and outputs it to the block size candidate determination unit 202.

  Hereinafter, a specific case where the statistical value of the motion vector determined by the motion search processing unit 203 is calculated and output to the block size candidate determination unit 202 will be described. In the random access coding mode in HEVC, it is possible to obtain high inter-frame coding efficiency by making prediction possible not only from past pictures on the time axis but also from future pictures. FIG. 4 is a diagram showing an inter-screen reference structure in the random access encoding mode. In FIG. 4, each picture is classified into four layers according to the reference pattern, and a value called TemporalID (time identifier) is assigned to each layer. A picture from the picture with TemporalID = 0 to the next picture with TemporalID = 0 is called GOP (Group of Picture).

  FIG. 4 shows an example of the random access encoding mode when the GOP size is 8. The motion search processing unit 203 calculates the variance of all motion vectors in the current encoding target picture only when the encoding order of the current encoding target picture is one immediately before the picture with TemporalID = 0, and calculates motion vector statistics Is output to the block size candidate determination unit 202.

  Here, the variance of the motion vector is calculated as the motion vector statistical value. However, a method of outputting the arithmetic mean, square sum, or absolute value sum of the motion vector as the motion vector statistical value may be used. When motion vector dispersion is used as a motion vector statistic value, an effect of facilitating selection of a specific block size candidate can be expected depending on whether the motion between screens is uniformly distributed within one picture. When the sum of absolute values of motion vectors is used as the motion vector statistic value, an effect of easily selecting a specific block size candidate can be expected depending on whether the motion between screens is large or small as a whole.

  Next, the configuration of the block size candidate determination unit 202 shown in FIG. 2 will be described with reference to FIG. FIG. 5 is a block diagram showing a configuration of the block size candidate determination unit 202 shown in FIG. The block size candidate determination unit 202 includes a candidate cost generation unit 301, an offset determination unit 302, and a candidate cost comparison unit 303.

  Next, the operation of the block size candidate determination unit 202 shown in FIG. 5 will be described with reference to FIG. FIG. 6 is a flowchart showing the operation of the block size candidate determination unit 202 shown in FIG. First, the candidate cost generation unit 301 receives the cost of each block size as input, and calculates the cost of the block size candidate (step S11). The candidate cost generation unit 301 outputs the calculated costs (candidate cost 1 and candidate cost 2) to the candidate cost comparison unit 303.

  On the other hand, the offset determination unit 302 receives the motion vector statistical value of the encoded picture calculated by the motion search processing unit 203 and the encoding information of the encoding target picture, and receives the motion vector statistical value of the encoded picture. Then, an offset value for cost comparison is calculated based on the encoding information of the encoding target picture (step S12). The offset determination unit 302 outputs the calculated cost comparison offset value to the candidate cost comparison unit 303.

  Next, the candidate cost comparison unit 303 receives two costs (candidate cost 1 and candidate cost 2) and a cost comparison offset value as inputs, and performs comparison based on the two costs to determine a block size candidate (step S13). ). Candidate cost comparison unit 303 outputs the determined block size candidates to motion search processing unit 203.

  Next, detailed operation of the candidate cost generation unit 301 shown in FIG. 5 will be described with reference to FIG. FIG. 7 is a flowchart showing a detailed operation of the candidate cost generation unit 301 shown in FIG. First, the candidate cost generation unit 301 inputs the cost value of each block size calculated by the prior motion prediction processing unit 201 (step S21). Then, the candidate cost generation unit 301 calculates the cost value (candidate cost 1) of the block size candidate 1 (step S22). The block size candidate 1 is only a block size of 64 × 64. The cost value of the block size candidate 1 is a cost value of 64 × 64 block size calculated by the prior motion prediction processing unit 201.

  Subsequently, the candidate cost generation unit 301 calculates the cost value (candidate cost 2) of the block size candidate 2 (step S23). Block size candidate 2 has three block sizes of 32 × 32, 16 × 16, and 8 × 8. Compare the cost value of the 16x16 block size and the sum of the cost values of the four 8x8 block sizes obtained by dividing the 16x16 area into 2 parts vertically and horizontally in the 16x16 area when the 64x64 area is divided into 4 parts. When the cost value of the 16 × 16 block size is small, the cost value of the 16 × 16 area is set as the cost value of the 16 × 16 block size, and when the cost value of the 16 × 16 block size is large, the cost of the 16 × 16 area is Let the value be the sum of the cost values of four 8 × 8 block sizes. The above 16 × 16 area cost calculation processing is performed on 16 16 × 16 areas.

  In the 32 × 32 area when the 64 × 64 area is divided into 2 parts vertically and horizontally, the cost value of the 32 × 32 block size is compared with the sum of the four 16 × 16 area costs obtained by dividing the 32 × 32 area into 2 parts vertically and horizontally. When the cost value of the × 32 block size is small, the cost value of the 32 × 32 region is set as the cost value of the 32 × 32 block size, and when the cost value of the 32 × 32 block size is large, the cost value of the 32 × 32 region is set to 4. The sum of the two 16 × 16 area costs. The above 32 × 32 region cost calculation procedure is performed for all four 32 × 32 regions obtained by dividing the target 64 × 64 region into two vertically and horizontally, and the sum of the four 32 × 32 region cost values is used as the cost value of block size candidate 2. To do.

  Finally, the candidate cost generation unit 301 outputs the calculated cost values of the block size candidates 1 and 2 (step S24).

Next, the detailed operation of the offset determination unit 302 shown in FIG. 5 will be described with reference to FIG. FIG. 8 is a flowchart showing a detailed operation of the offset determination unit 302 shown in FIG. First, the latest value σ _MV ² of the motion vector variance for each picture calculated by the motion search processing unit 203 is compared with the threshold value TH, and it is determined whether or not the latest value σ _MV ² of the motion vector variance is smaller than the threshold value TH. (Step S31). As a result of this determination, if the motion vector variance σ _MV ² is smaller than the threshold value TH, that is, if the motion of the object is uniform within one picture, the cost comparison offset value λ _offset is set to 0 (step S32).

On the other hand, when the motion vector variance σ _MV ² is larger than the threshold value TH, that is, when the motion of the object varies within one picture, the cost comparison offset value λ _offset is set to a non-zero real number (for example, α · TemporalID). (Step S33). In this case, a negative real number may be used for λ _offset . This makes it easy to select block size candidate 1, that is, a block size of 64 × 64 for a picture with scattered motion, and as a result, the block size is prevented from becoming too small due to motion variation, and the code amount can be reduced. I can hope.

Also, as a method for determining λ _offset in this case, it is preferable to obtain TemporalID of the picture to be encoded from the encoding information and set the cost comparison offset value λ _offset to a constant multiple α · TemporalID of TemporalID. α is preferably a negative real constant. Although the accuracy of motion prediction tends to be high for a picture with a large Temporal ID, the cost comparison offset value λ _offset is set to a constant multiple α · Temporal ID of the Temporal ID. The block size can be easily selected, and the amount of code can be reduced by performing motion prediction on a larger block.

Here, the cost comparison offset value λ _offset when the motion vector statistical value calculated by the motion search processing unit 203 is larger than the threshold value TH is set to a constant multiple α · TemporalID of TemporalID, but an appropriate λ _offset is fixed for each TemporalID. A value may be prepared in advance, and λ _offset may be switched according to the Temporal ID of the encoding target picture.

Also, here, the calculation formula of the cost comparison offset value λ _offset is switched by comparing the motion vector statistical value with the threshold TH, but the motion vector statistical value is divided by the GOP size in order to consider the GOP structure. The calculation formula of λ _offset may be switched by comparing the value with the threshold value TH. As a result, the motion vector statistical value becomes a scaled value compared with the GOP size, and there is no need to change the threshold TH depending on the size of the GOP size.

  The two candidate cost values calculated by the candidate cost generation unit 301 and the cost comparison offset value calculated by the offset determination unit 302 are input to the candidate cost comparison unit 303, and the encoded block size candidates are set as two block size candidates. Determine one of the following.

  Next, detailed operation of the candidate cost comparison unit 303 shown in FIG. 5 will be described with reference to FIG. FIG. 9 is a flowchart showing a detailed operation of the candidate cost comparison unit 303 shown in FIG. First, the candidate cost comparison unit 303 receives the cost value and the cost comparison offset value of the block size candidates 1 and 2 (step S41).

  Next, the candidate cost comparison unit 303 compares the sum of the candidate cost value of the block size candidate 1 and the cost comparison offset value determined by the offset determination unit 302 with the candidate cost value of the block size candidate 2 to determine the block size. It is determined whether or not candidate 1 cost value + cost comparison offset value <block size candidate 2 cost value (step S42). As a result of this determination, if the candidate cost value of block size candidate 2 is large, the block size candidate is determined as candidate 1 (step S43), and if the candidate cost value of block size candidate 2 is small, the block size candidate is set as candidate 2. Determine (step S44).

  As described above, in the video encoding system having a motion prediction unit in which there are a plurality of encoding block size candidates, the depth of the hierarchy in the reference structure of the encoding target picture, that is, TemporalID, and the encoded picture By determining an offset value in cost comparison of block size candidates based on statistical value information of motion vectors and appropriately narrowing down the block size candidates by cost comparison, it is possible to reduce the amount of calculation required for motion search.

  You may make it implement | achieve all or one part of the video encoding apparatus 100 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. It may be realized using hardware such as PLD (Programmable Logic Device) or FPGA (Field Programmable Gate Array).

  As mentioned above, although embodiment of this invention has been described with reference to drawings, the said embodiment is only the illustration of this invention, and it is clear that this invention is not limited to the said embodiment. is there. Therefore, additions, omissions, substitutions, and other modifications of the components may be made without departing from the technical idea and scope of the present invention.

  The present invention can be applied to a video encoding device and a video encoding program in which the amount of computation required for motion search processing is limited.

  DESCRIPTION OF SYMBOLS 102 ... Inter prediction process part, 201 ... Prior motion prediction process part, 202 ... Block size candidate determination part, 203 ... Motion search process part, 301 ... Candidate cost generation part, 302 ... Offset determination unit, 303 ... Candidate cost comparison unit

Claims (5)

A video encoding device that performs a temporal prediction of a picture of an input video signal, performs a motion prediction for each coding block for a coding target picture, and performs a coding process of a differential signal,
Using the encoding target picture and a reference picture of the decoded video signal of the motion prediction destination, a pre-motion prediction process for capturing a rough motion of the target encoding block is performed for each of all the encoding block sizes. A prior motion prediction processing means for calculating a prior motion prediction processing cost;
Block size candidate cost means for determining a block size candidate cost in the first encoded block size candidate and a block size candidate cost in the second encoded block size candidate based on the prior motion prediction processing cost;
Parameter determining means for determining a motion statistical parameter based on motion vector information of a coded picture of the input video signal;
Offset value setting means for setting a cost comparison offset value based on the motion statistic parameter and the depth of the hierarchy in the reference structure of the encoding target picture;
Based on the block size candidate cost in the first encoded block size candidate, the block size candidate cost in the second encoded block size candidate, and the cost comparison offset value, encoding of the encoding target picture A block size candidate determining means for determining a block size candidate as a first block size candidate or a second block size candidate from among all block size candidates;
A video encoding device comprising: motion search means for performing motion prediction processing of the encoding target picture and determining a block size and a motion vector by using the determined encoding block size candidate. The parameter determination means includes
The video encoding according to claim 1, wherein all motion vectors in a specified encoded picture of the input video signal are calculated, and an absolute value sum of the motion vectors is determined as a motion statistical parameter of the encoding target picture. apparatus. The parameter determination means includes
2. The video encoding device according to claim 1, wherein all motion vectors in a specified encoded picture of the input video signal are calculated, and a variance value of the motion vector is determined as a motion statistical parameter of the encoding target picture. . The block size candidate determination means includes
As a result of calculating the sum of the cost of the motion vector in the second block size candidate and the cost comparison offset value, and comparing the calculated sum with the cost of the motion vector in the first block size candidate, When it is determined that the cost of one block size candidate is higher, the block size candidate of the encoding target picture is set as a second block size candidate, and it is determined that the cost of the first block size candidate is lower 4. The video encoding device according to claim 1, wherein if the block size is selected, the block size candidate of the encoding target picture is set as a first block size candidate. 5.   A video encoding program for causing a computer to function as the video encoding device according to any one of claims 1 to 4.

Images