KR20200085234A - Image encoding/decoding method and apparatus using efficient motion information prediction - Google Patents

Abstract

A method and an apparatus for image encoding/decoding, according to the present invention, may generate a candidate list of a current block and perform inter-prediction of the current block by using any one of a plurality of candidates belonging to the candidate list. Here, the plurality of candidates comprise at least one from among a spatial candidate, a temporal candidate, and a restoration information-based candidate, and the restoration information-based candidate may be added from a buffer which stores motion information decoded prior to the current block.

Description

Image encoding/decoding method and apparatus using efficient motion information prediction{IMAGE ENCODING/DECODING METHOD AND APPARATUS USING EFFICIENT MOTION INFORMATION PREDICTION}

The present invention relates to an image encoding/decoding method and apparatus, and more particularly, when generating a prediction block of a current block, motion information corresponding to a predictor for determining optimal motion information of a current block among a plurality of restored motion information It relates to a method and apparatus for efficiently determining.

Recently, the demand for multimedia data such as video has been rapidly increasing on the Internet. However, the speed at which the bandwidth of the channel develops is difficult to keep up with the rapidly increasing amount of multimedia data. Accordingly, the International Standardization Organization's Video Coding Expert Group (VCEG) of the ITU-T and the Moving Picture Expert Group (MPEG) of the ISO/IEC announced the High Efficiency Video Coding (HEVC) version 1, the video compression standard, in February 2014. Instituted.

HEVC defines techniques such as intra-picture prediction (or intra prediction), inter-picture prediction (or inter prediction), transformation, quantization, entropy encoding, and in-loop filters.

The present invention is to propose a method for improving prediction efficiency by efficiently deriving motion information used when generating a MERGE/AMVP candidate list.

The video encoding/decoding method and apparatus according to the present invention may generate a candidate list of a current block, and perform inter prediction of the current block by using any one of a plurality of candidates belonging to the candidate list.

In the image encoding/decoding method and apparatus according to the present invention, the plurality of candidates include at least one of a spatial candidate, a temporal candidate, or a reconstruction information-based candidate, and the reconstruction information-based candidate is motion information decoded before the current block. It can be added from the buffer to store.

In the video encoding/decoding method and apparatus according to the present invention, motion information stored in the buffer is added to the candidate list in the order of motion information stored in the buffer later, or the candidate list in the order of motion information stored in the buffer first. Can be added to

In the image encoding/decoding method and apparatus according to the present invention, the order or number of motion information stored in the buffer to be added to the candidate list may be differently determined according to a prediction mode between screens of the current block.

In the image encoding/decoding method and apparatus according to the present invention, the candidate list is filled using motion information stored in the buffer until the maximum number of candidates of the candidate list is reached, or 1 is selected from the maximum number of candidates. It can be filled using the motion information stored in the buffer until it reaches the subtracted number.

In the image encoding/decoding method and apparatus according to the present invention, the buffer may be initialized in any one of a coding tree unit (CTU), a CTU row, a slice, or a picture.

The computer-readable recording medium according to the present invention can store a bitstream decoded by an image decoding method.

In a computer-readable recording medium according to the present invention, the image decoding method comprises generating a candidate list of a current block and using one of a plurality of candidates belonging to the candidate list, And performing prediction.

In the computer-readable recording medium according to the present invention, the plurality of candidates include at least one of a spatial candidate, a temporal candidate, or a reconstruction information-based candidate, and the reconstruction information-based candidate is a motion decoded before the current block. It can be added from a buffer that stores information.

In the computer-readable recording medium according to the present invention, the motion information stored in the buffer is added to the candidate list in the order of motion information stored in the buffer later, or in the order of motion information stored in the buffer first. Can be added.

In the computer-readable recording medium according to the present invention, the order or number of motion information stored in the buffer added to the candidate list may be differently determined according to a prediction mode between screens of the current block.

In the computer-readable recording medium according to the present invention, the candidate list is filled with motion information stored in the buffer until the maximum number of candidates in the candidate list is reached, or 1 is subtracted from the maximum number of candidates Until the number is reached, it may be filled using motion information stored in the buffer.

In the computer-readable recording medium according to the present invention, the buffer may be initialized by any one of coding tree unit (CTU), CTU row, slice, or picture.

According to the present invention, prediction information can be improved by efficiently deriving motion information used when generating a MERGE/AMVP candidate list.

1 is a flowchart briefly showing an image encoding apparatus.
2 is a view for explaining in detail the prediction unit of the video encoding apparatus.
3 is a diagram for explaining a method for deriving candidate motion information in SKIP and MERGE modes.
4 is a flowchart showing a method of deriving candidate motion information in the AMVP mode.
5 is a flowchart illustrating a method of encoding prediction information.
6 is a flowchart briefly illustrating an image decoding apparatus.
7 is a view for explaining a prediction unit of an image decoding apparatus.
8 is a flowchart illustrating a method of decoding prediction information.
9 is a flowchart illustrating a method of configuring a MERGE/AMVP candidate list according to the present embodiment.
10 is a view for explaining a method of deriving temporal candidate motion information according to the present embodiment.
11 is a diagram for explaining a first method of determining a target block in a target picture when deriving temporal candidate motion information according to the present embodiment.
12 is a diagram for describing a second method of determining a target block in a target picture when deriving temporal candidate motion information according to the present embodiment.
13 is a diagram for explaining a third method of determining a target block in a target picture when deriving temporal candidate motion information according to the present embodiment.
14 is a diagram for explaining a first method for deriving history-based candidate motion information according to the present embodiment.
15 is a view for explaining a second method for deriving the history-based candidate motion information according to the present embodiment.
16 is a diagram for describing a first method of deriving average candidate motion information according to the present embodiment.
17 is a diagram for describing a second method of deriving average candidate motion information according to the present embodiment.

DETAILED DESCRIPTION Embodiments of the present invention will be described in detail so that those skilled in the art to which the present invention pertains may easily practice the present invention with reference to the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and like reference numerals are assigned to similar parts throughout the specification.

Throughout this specification, when a part is said to be'connected' to another part, this includes not only the case of being directly connected, but also the case of being electrically connected with another element in between.

In addition, when it is said that a part "includes" a certain component in the whole specification, it means that other components may be further included rather than excluding other components unless otherwise specified.

Further, terms such as first and second may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from other components.

In addition, in the embodiments of the apparatus and method described herein, a part of the configuration of the apparatus or a part of the method may be omitted. Also, the order of some of the components of the device or some of the steps of the method may be changed. In addition, other configurations or other steps may be inserted in some of the configuration of the device or in some of the steps of the method.

Further, some of the components or steps of the first embodiment of the present invention may be added to the second embodiment of the present invention, or may replace some components or steps of the second embodiment.

In addition, the components shown in the embodiments of the present invention are shown independently to indicate different characteristic functions, and do not mean that each component is composed of separate hardware or one software component unit. That is, for convenience of description, each component is listed as each component, and at least two components of each component may be combined to form one component, or one component may be divided into a plurality of components to perform a function. The integrated and separated embodiments of each of these components are also included in the scope of the present invention without departing from the essence of the present invention.

In the present specification, blocks may be variously expressed in units, regions, units, partitions, and the like, and samples may be variously expressed in pixels, pels, pixels, and the like.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In describing the present invention, redundant description of the same components will be omitted.

1 is a block flow diagram briefly showing the configuration of a video encoding apparatus. The image encoding apparatus is an apparatus for encoding an image, and includes a block division unit, a prediction unit, a transformation unit, a quantization unit, an entropy encoding unit, an inverse quantization unit, an inverse transformation unit, an addition unit, an in-loop filter unit, a memory unit, and a subtraction unit. can do.

The block dividing unit 101 divides blocks from a maximum sized block (hereinafter referred to as a maximum coding block) to a minimum sized block (hereinafter referred to as a minimum coding block). There are various methods of the block division method. Quad-tree splitting (hereinafter referred to as QT (Quad-Tree) splitting) is a split that accurately divides the current coding block. Binary-tree partitioning (hereinafter, referred to as BT (Binary-Tree) partitioning) is a partition that divides an encoding block into horizontal or vertical directions exactly. A ternary-tree split is a split that divides an encoded block into either a horizontal or vertical direction. When the coding block is split in the horizontal direction, the ratio of the height of the divided block may be {1:n:1}. Alternatively, when the coding block is split in the vertical direction, the ratio of the width of the divided block may be {1:n:1}. Here, n may be 1, 2, 3, or more natural numbers. In addition, there may be various partitioning methods. In addition, it is also possible to divide by considering several division methods simultaneously.

The prediction unit 102 generates a prediction block using neighboring pixels of a block to be currently predicted (hereinafter referred to as a prediction block) from a current original block or pixels in a reference picture that has been previously encoded/decoded. The prediction block may generate one or more prediction blocks in the coding block. When there is one prediction block in the coding block, the prediction block has the same form as the coding block. The prediction technique of a video signal is largely composed of intra-screen prediction and inter-screen prediction. In-screen prediction is a method of generating a prediction block using surrounding pixels of a current block, and inter-screen prediction has been previously encoded/decoded. This is a method of finding a block most similar to the current block in the finished reference picture and generating a predictive block. Then, the optimal prediction mode of the prediction block is determined by using various techniques such as rate-distortion optimization (RDO) of the residual block minus the original block and the prediction block. The RDO cost calculation equation is as shown in Equation 1.

D, R, and J are deterioration by quantization, rate of compressed stream, and RD cost, Φ is the encoding mode, and λ is the Lagrangian multiplier, which is a scale correction factor to match the unit between the amount of error and the amount of bit. use. In order to be selected as the optimal encoding mode in the encoding process, J when the corresponding mode is applied, that is, the RD-cost value must be smaller than when other modes are applied. In the equation for obtaining the RD-cost value, bit rate and error are considered simultaneously. To calculate.

The intra prediction unit (not shown) may generate a prediction block based on reference pixel information around the current block, which is pixel information in the current picture. When the prediction mode of the neighboring block of the current block to which the intra prediction is to be performed is inter prediction prediction, the reference pixel included in the neighboring block to which the inter prediction prediction is applied is referred to as a reference pixel in another block around the intra prediction is applied. Can be replaced. That is, when the reference pixel is not available, the available reference pixel information may be replaced with at least one reference pixel among the available reference pixels.

In the intra prediction, the prediction mode may have a directional prediction mode that uses reference pixel information according to a prediction direction and a non-directional mode that does not use directional information when performing prediction. The mode for predicting the luminance information and the mode for predicting the color difference information may be different, and the prediction prediction mode information or the predicted luminance signal information used in the screen used to predict the luminance information may be used to predict the color difference information. have.

The intra prediction unit may include an adaptive intra smoothing (AIS) filter, a reference pixel interpolation unit, and a DC filter. The AIS filter is a filter that performs filtering on the reference pixel of the current block and can adaptively determine whether to apply the filter according to the prediction mode of the current prediction unit. When the prediction mode of the current block is a mode that does not perform AIS filtering, the AIS filter may not be applied.

The reference pixel interpolation unit of the on-screen prediction unit interpolates the reference pixel and the reference pixel at a fractional unit position when the intra-prediction prediction mode of the prediction unit is a prediction unit that performs intra-prediction prediction based on the pixel value that interpolated the reference pixel. Can generate If the prediction mode of the current prediction unit is a prediction mode in which a prediction block is generated without interpolating a reference pixel, the reference pixel may not be interpolated. The DC filter may generate a prediction block through filtering when the prediction mode of the current block is the DC mode.

The inter-screen prediction unit (not shown) generates a prediction block using pre-restored reference images and motion information stored in the memory 110. The motion information may include, for example, a motion vector, a reference picture index, a list 1 prediction flag, a list 0 prediction flag, and the like.

The inter-screen prediction unit may derive a prediction block based on information of at least one of a previous picture or a subsequent picture of the current picture. In addition, a prediction block of a current block may be derived based on information of some regions in which encoding in a current picture is completed. The inter-screen prediction unit according to an embodiment of the present invention may include a reference picture interpolation unit, a motion prediction unit, and a motion compensation unit.

The reference picture interpolator may receive reference picture information from the memory 110 and generate pixel information of an integer pixel or less in the reference picture. In the case of luminance pixels, a DCT-based 8-tap interpolation filter (DCT-based interpolation filter) having different filter coefficients may be used to generate pixel information of integer pixels or less in units of 1/4 pixels. In the case of a color difference signal, a DCT-based interpolation filter (DCT-based interpolation filter) having different filter coefficients may be used to generate pixel information of an integer pixel or less in units of 1/8 pixels.

The motion prediction unit may perform motion prediction based on the reference picture interpolated by the reference picture interpolation unit. As a method for calculating a motion vector, various methods such as Full Search-based Block Matching Algorithm (FBMA), Three Step Search (TSS), and New Three-Step Search Algorithm (NTS) can be used. The motion vector may have a motion vector value in units of 1/2 or 1/4 pixels based on the interpolated pixels. The motion prediction unit may predict a prediction block of a current block by differently using a motion prediction method. Various methods such as a skip mode, a merge mode, and an advanced motion vector prediction (AMVP) mode may be used as a motion prediction method.

2 is a flowchart illustrating a flow in the prediction unit of the video encoding apparatus. When performing intra prediction using the original information and the reconstructed information (201), the optimal intra prediction mode for each prediction mode is determined using an RD-cost value (202), and a prediction block is generated. When inter-screen prediction is performed using the original information and the reconstruction information (203), RD-cost values are calculated for the SKIP mode, the MERGE mode, and the AMVP mode. The MERGE candidate search unit 204 configures a SKIP mode and a set of candidate motion information for the MERGE mode. Among the set of candidate motion information, optimal motion information is determined using an RD-cost value (205). The AMVP candidate search unit 206 configures a candidate motion information set for the AMVP mode. Motion estimation is performed using the corresponding candidate motion information sets (207), and optimal motion information is determined. Motion compensation is performed using the optimal motion information determined in each mode (208) to generate a prediction block.

The above-described inter-screen prediction may be composed of three modes (SKIP mode, MERGE mode, and AMVP mode). Each prediction mode may obtain a prediction block of a current block using motion information (prediction direction information, reference picture information, motion vector), and there may be an additional prediction mode using motion information.

The SKIP mode determines optimal prediction information using motion information of an already restored region. A motion information candidate group is constructed in the reconstructed region, and among the candidate groups, a candidate having the smallest RD-cost value is used as prediction information to generate a prediction block. Here, a method of configuring the motion information candidate group is described in the MERGE mode described below. Since it is the same as the method of configuring the motion information candidate group, it is omitted in this description.

The MERGE mode is the same as the SKIP mode in that optimal prediction information is determined using motion information of an already restored region. However, the SKIP mode differs in that the motion information that causes the prediction error to be 0 is searched for in the motion information candidate group, and the MERGE mode searches the motion information candidate group for which the prediction error is not 0. As in the SKIP mode, a motion information candidate group is configured in the reconstructed region, and among the candidate groups, a candidate having a minimum RD-cost value is used as prediction information to generate a prediction block.

3, 301 shows a method of generating candidate motion information groups in the SKIP mode and the MERGE mode. The maximum number of motion information candidate groups may be determined in the same manner by the video encoding apparatus and the video decoding apparatus, and transmitted from the upper stage of the block, such as the upper header of the video encoding apparatus (the upper header column, the video parameter stage, the sequence parameter stage, and the picture parameter stage). The corresponding number information may be previously transmitted in the meaning of the parameters. In the description of steps S305 and S306, only when the spatial candidate block and the temporal candidate block are encoded in the inter prediction mode, motion information derived using the corresponding motion information is included in the motion information candidate group. In step S305, four candidates are selected from among five spatial candidate blocks around the current block in the same picture. The location of the spatial candidate is referred to 302 in FIG. 3, and the location of each candidate can be changed to any block in the reconstructed region. Spatial candidates are considered in order of A ₁ , A ₂ , A ₃ , A ₄ , and A ₅ , and first, motion information of usable spatial candidate blocks is determined as spatial candidates. However, this is only an example of priority, and the priority may be in the order of A ₂ , A ₁ , A ₃ , A ₄ , A ₅ , or A ₂ , A ₁ , A ₄ , A ₃ , A ₅ . When there is overlapping motion information, only motion information of a candidate having a high priority is considered. In step S306, one candidate among two temporal candidate blocks is selected. The location of the temporal candidate is referred to 302 of FIG. 3, and the location of each candidate is determined based on a block of the same location as the current block location of the current picture in the collocated picture. Here, the collocated picture can be set under the same conditions in the video encoding device and the video decoding device in the reconstructed picture. The temporal candidates are considered in the order of B ₁ and B ₂ blocks, and motion information of the available candidate blocks is first determined as temporal candidates. For the method of determining the motion information of the temporal candidate, refer to 303 of FIG. 3. The candidate block (B ₁ , B ₂ ) motion information in the collocated picture indicates a prediction block in the reference picture B. (However, the reference pictures of each candidate block may be different from each other. For convenience, this is referred to as reference picture B). By calculating the ratio, the motion vector of the candidate block is scaled by the corresponding ratio to determine the motion vector of the temporal candidate motion information. Equation 2 means a scaling equation.

MV is a motion vector of temporal candidate block motion information, MV _scale is a scaled motion vector, TB is a temporal distance between a collocated picture and reference picture B, and TD is a temporal distance between a current picture and a reference picture A. Also, the reference pictures A and B may be the same reference picture. The scaled motion vector is determined as the motion vector of the temporal candidate, and the reference picture information of the temporal candidate motion information is determined as the reference picture of the current picture to derive the motion information of the temporal candidate. Steps S307 are performed only when the maximum number of motion information candidate groups is not satisfied in steps S305 and S306, and a new bidirectional motion information candidate group is added with a combination of motion information candidates derived in the previous step. The bidirectional motion information candidate is obtained by combining the motion information of the past or future direction derived one by one and combining them as a new candidate. Table 304 of FIG. 3 shows the priority of the bidirectional motion information candidate combination. In addition to the combinations of this table, additional combinations may appear, and this table shows one example. If the maximum number of motion information candidate groups is not satisfied even when using the bidirectional motion information candidate, step S307 is performed. In step S308, the motion vector of the motion information candidate is fixed to the zero motion vector, and the reference picture according to the prediction direction is varied to fill the maximum number of motion information candidate groups.

The AMVP mode determines the optimal motion information through motion estimation for each reference picture according to the prediction direction. Here, the prediction direction may be unidirectional using only one direction of the past/future, or it may be bidirectional using both the past and future directions. Motion compensation is performed using the optimal motion information determined through motion estimation to generate a prediction block. Here, a candidate motion information group for motion estimation is derived for each reference picture according to the prediction direction. The candidate motion information group is used as a starting point for motion estimation. 4 for a method of deriving a candidate motion information group for motion estimation in the AMVP mode.

The maximum number of motion information candidate groups may be determined in the same manner by the video encoding apparatus and the video decoding apparatus, or the corresponding number information may be previously transmitted from the upper header of the video encoding apparatus. In the description of steps S401 and S402, only when the spatial candidate block and the temporal candidate block are encoded in the inter prediction mode, motion information derived using the corresponding motion information is included in the motion information candidate group. In step S401, unlike the description of step S305, the number (two) derived as the spatial candidate may be different, and the priority for selecting the spatial candidate may also be different. The rest of the explanation is the same as that of step S305. Step S402 is the same as the description of step S306. In step S403, if there is overlapping motion information among candidates derived so far, it is removed. Step S404 is the same as the description of step S308. Among the motion information candidates derived in this way, the motion information candidate having the smallest RD-cost value is selected as the optimum motion information candidate, and then the motion estimation process is performed based on the motion information to obtain the optimum motion information in the AMVP mode.

The transform unit 103 generates a transform block by transforming a residual block that is a difference between the original block and the predictive block. The transform block is the smallest unit used for transform and quantization processes. The transform unit converts the residual signal into the frequency domain to generate a transform block having a transform coefficient. Here, as a method of transforming the residual signal into the frequency domain, various transform techniques such as Discrete Cosine Transform (DCT)-based transform, Discrete Sine Transform (DST), and Karhunen Loeve Transform (KLT) can be used. , Using this, the residual signal is transformed into the frequency domain to generate a transform coefficient. In order to conveniently use the transformation technique, matrix operations are performed using a basis vector. Depending on the prediction mode in which the prediction block is encoded, various transformation techniques can be used in the matrix operation. For example, according to the prediction mode in the intra prediction, a discrete cosine transform may be used in the horizontal direction and a discrete sine transform may be used in the vertical direction.

The quantization unit 104 quantizes a transform block to generate a quantized transform block. That is, the quantization unit quantizes transform coefficients of a transform block generated by the transform unit 103 to generate a quantized transform block having a quantized transform coefficient. A dead zone uniform threshold quantization (DZUTQ) or a quantization weighted matrix (DZUTQ) may be used as the quantization method, but various quantization methods such as improved quantization may be used.

On the other hand, in the above, although the image encoding apparatus is shown and described as including a transform unit and a quantization unit, the transform unit and the quantization unit may be selectively included. That is, the image encoding apparatus may transform the residual block to generate a transform block and not perform a quantization process, and not only transform the residual block into frequency coefficients, but also perform a quantization process, and even transform and quantize processes You may not do all of them. In the image encoding apparatus, a block that enters an input of an entropy encoding unit is usually referred to as a'quantized transformation block' even if some of the transformation units and the quantization units are not performed or all of the processes are not performed.

The entropy encoding unit 105 encodes a quantized transform block and outputs a bitstream. That is, the entropy encoding unit encodes coefficients of the quantized transform block output from the quantization unit using various encoding techniques such as entropy encoding, and additional information necessary for decoding the block in an image decoding apparatus described later ( For example, a bitstream including information about a prediction mode (information on a prediction mode may include motion information determined by a prediction unit or information on a prediction mode in a screen, etc.) and quantization coefficients are generated and output.

The inverse quantization unit 106 restores the inverse quantized transform block by inversely performing a quantization technique used for quantization on the quantized transform block.

The inverse transform unit 107 inversely transforms the inverse quantized transform block using the same method used for transform to restore the residual block, and performs an inverse transform by performing the transform technique used by the transform unit inversely.

On the other hand, in the above, the inverse quantization unit and the inverse transformation unit may perform inverse quantization and inverse transformation using the quantization method and the transformation method used by the quantization unit and the transformation unit inversely. In addition, when only the quantization unit and the quantization unit do not perform the transformation, only the inverse quantization may be performed and the inverse transformation may not be performed. If both the transform and the quantization are not performed, the inverse quantization unit and the inverse transform unit do not perform both inverse transform and inverse quantization or may be omitted without being included in the image encoding apparatus.

The adder 108 restores the current block by adding the residual signal generated by the inverse transform unit and the prediction block generated through prediction.

The filter unit 109 is a process of additionally filtering the entire picture after all blocks in the current picture are restored, and includes deblocking filtering, sample adaptive offset (SAO), and adaptive loop filter (ALF). Deblocking filtering refers to the task of reducing block distortion that occurs while coding an image in block units. Sample adaptive offset (SAO) is a method of minimizing the difference between the reconstructed image and the original image by subtracting or adding a specific value to the reconstructed pixel. Says the job. ALF (Adaptive Loop Filtering) may be performed based on a value obtained by comparing a filtered reconstructed image with an original image. After dividing the pixels included in the image into a predetermined group, one filter to be applied to the group may be determined to perform filtering differently for each group. Information related to whether to apply the ALF may be transmitted for each coding unit (CU), and the shape and/or filter coefficient of the ALF filter to be applied may be changed according to each block. Also, the ALF filter of the same form (fixed form) may be applied regardless of the characteristics of the block to be applied.

The memory 110 adds the residual signal generated by the inverse transform unit and the prediction block generated through prediction, and then stores the restored current block through additional filtering in the in-loop filter unit, such as the next block or the next picture. It can be used to predict.

The subtraction unit 111 subtracts the prediction block from the current original block to generate a residual block.

5 is a flowchart illustrating an encoding flow of coding information in an entropy encoding unit in an image encoding apparatus. In step S501, operation information of the SKIP mode is encoded. In step S502, the operation of the SKIP mode is determined. If the SKIP mode is operated in step S502, after the MERGE candidate index information for the SKIP mode is encoded in step S507, this flow chart ends. If the SKIP mode does not operate in step S502, the prediction mode is encoded in step S503. In step S503, it is determined whether the prediction mode is inter-screen prediction or intra-screen prediction mode. If the prediction mode was the inter prediction mode in step S504, the operation information of the MERGE mode is encoded in step S505. In step S506, it is determined whether the MERGE mode is activated. If the MERGE mode is operated in step S506, the process proceeds to step S507, after encoding the MERGE candidate index information for the MERGE mode, the flowchart ends. If the MERGE mode does not operate in step S506, the prediction direction is encoded in step S508. Here, the prediction direction may be one of a past direction, a future direction, and two directions. In step S509, it is determined whether the prediction direction is the future direction or not. If the prediction direction is not the future direction in step S509, the reference picture index information of the past direction is encoded in step S510. In step S511, motion vector difference (MVD) information of a past direction is encoded. In step S512, the motion vector predictor (MVP) information of the past direction is encoded. In step S509, it is determined whether the prediction direction is in the future direction or in both directions, or in step S513, when the prediction direction is in the past direction. If the prediction direction is not the past direction in step S513, the future direction reference picture index information is encoded in step S514. In step S515, MVD information of a future direction is encoded. In step S516, after encoding the MVP information in the future direction, this flow chart ends. If the prediction mode was the intra prediction mode in step S504, after encoding the intra prediction mode information in step S517, the flowchart ends.

6 is a block flow diagram briefly showing the configuration of the video decoding apparatus 600.

The image decoding apparatus 600 is an apparatus for decoding an image, and may largely include a block entropy decoding unit, an inverse-quantization unit, an inverse-transformation unit, a prediction unit, an adder unit, an in-loop filter unit, and a memory. The encoding block in the image encoding apparatus is referred to as a decoding block in the image decoding apparatus.

The entropy decoding unit 601 interprets the bitstream transmitted from the image encoding apparatus and reads various information and quantized transform coefficients necessary for decoding the corresponding block.

The inverse quantization unit 602 reconstructs an inverse quantized block having an inverse quantized coefficient by inversely performing a quantization technique used for quantization on the quantization coefficient decoded by the entropy decoding unit.

The inverse transform unit 603 inversely transforms the inverse-quantized transform block using the same method used for transform to restore a residual block having a differential signal. Inverse transform is performed by performing the inverse transform technique used by the transform unit do.

The prediction unit 604 generates a prediction block using prediction mode information decoded by the entropy decoding unit, which uses the same method as the prediction method performed by the prediction unit of the image encoding apparatus.

The adder 605 restores the current block by adding the residual signal restored by the inverse transform unit and the prediction block generated through prediction.

After restoring all blocks in the current picture, the filter unit 606 performs de-blocking filtering, sample adaptive offset (SAO), and ALF as a process of performing additional filtering throughout the picture. This is the same as described in the in-loop filter section of the device.

The memory 607 adds the residual signal generated by the inverse transform unit and the prediction block generated through prediction, and then stores the restored current block that has undergone additional filtering in the in-loop filter unit, such as the next block or the next picture. It can be used to predict.

7 is a flowchart illustrating a flow in the prediction unit of the video decoding apparatus. If the prediction mode is intra prediction, the optimal intra prediction mode information is determined (701), and intra prediction is performed (702) to generate a prediction block. If the prediction mode is inter-screen prediction, an optimal prediction mode among SKIP, MERGE, and AMVP modes is determined (703). When decoding in the SKIP mode or the MERGE mode, the MERGE candidate search unit 704 configures a SKIP mode and a set of candidate motion information for the MERGE mode. Among the set of candidate motion information, optimal motion information is determined using the transmitted candidate index (e.g., merge index) (705). When decoding in the AMVP mode, the AMVP candidate search unit 706 configures a candidate motion information set for the AMVP mode. Among the candidate motion information candidates, optimal motion information is determined using the transmitted candidate index (e.g., MVP information) (707). Thereafter, motion compensation is performed using the optimal motion information determined in each mode (708) to generate a prediction block.

8 is a flowchart illustrating a decoding flow of coding information in an image decoding apparatus. In step S801, the operation information of the SKIP mode is decoded. In step S802, it is determined whether the SKIP mode operates. If the SKIP mode is operated in step S802, after decoding the MERGE candidate index information for the SKIP mode in step S807, the flow diagram ends. If the SKIP mode does not operate in step S802, the prediction mode is decoded in step S803. In step S803, it is determined whether the prediction mode is inter-screen prediction or intra-screen prediction mode. If the prediction mode was the inter-screen prediction mode in step S804, the operation information of the MERGE mode is decoded in step S805. In step S806, it is determined whether the MERGE mode is activated. If the MERGE mode is operated in step S806, the process proceeds to step S807, after decoding the MERGE candidate index information for the MERGE mode, the flowchart ends. If the MERGE mode does not operate in step S806, the prediction direction is decoded in step S808. Here, the prediction direction may be one of a past direction, a future direction, and two directions. In step S809, it is determined whether the prediction direction is the future direction or not. If the prediction direction is not the future direction in step S809, the reference picture index information of the past direction is decoded in step S810. In step S811, MVD (Motion Vector Difference) information of a past direction is decoded. In step S812, the past direction MVP (Motion Vector Predictor) information is decoded. In step S809, it is determined whether the prediction direction is in the future direction or in both directions, or in step S813, when the prediction direction is in the past direction. If the prediction direction is not the past direction in step S813, the future direction reference picture index information is decoded in step S814. In step S815, the MVD information in the future direction is decoded. In step S816, after decoding the MVP information in the future direction, this flow chart ends. If the prediction mode was the intra prediction mode in step S804, after decoding the intra prediction mode information in step S1317, the flow diagram ends.

(Common Example)

In the following embodiment, a method for deriving candidate motion information for inter-frame prediction of the current block by the MERGE candidate search units 204 and 704 and the AMVP candidate search units 206 and 706 in the prediction units of the video encoding device and the video decoding device I will explain. The candidate motion information is directly determined as motion information of the current block in the MERGE candidate search unit, and used as a predictor for transmitting optimal motion information of the current block in the AMVP candidate search unit.

9 is a flowchart illustrating a method of deriving candidate motion information for a MERGE/AMVP mode. Although the method for deriving the candidate motion information of the Merge mode and the AMVP mode in this flow chart is shown in the same flow chart, some candidates for each mode may not be used. Accordingly, candidate motion information derived for each mode may be different, and the number of derived candidate motion information may also be different. For example, in the Merge mode, 4 (B) candidates among 5 (A) spatial candidates can be selected, and in the AMVP mode, only 2 (B) candidates among 4 (A) spatial candidates can be selected. In steps S901 and S902, A, B, C, and D (A, B, C, and D are integers of 1 or more) are the number of spatial candidates, the number of spatial candidates, the number of temporal candidates, and the number of temporal candidates, respectively. Means

The description of step S901 is the same as the description of steps S305 and S401 described above. However, the positions of neighboring blocks for spatial candidates may be different. In addition, the neighboring block for the spatial candidate may belong to at least one of the first group, the second group, or the third group. Here, the first group includes at least one of the left block (A ₁ ) or the lower left block (A ₄ ) of the current block, and the second group is the upper block (A ₂ ) or the upper right block (A ₃ ) of the current block. ), and the third group may include at least one of an upper left block A ₅ of the current block, a block adjacent to the bottom of the upper left block, or a block adjacent to the left of the upper left block.

The description of step S902 is the same as the description of steps S306 and S402 described above. Similarly, the positions of blocks for temporal candidates may be different.

In step S903, temporal candidates for each sub-block are added. However, when a temporal candidate for each sub-block is added from the list of AMVP candidates, only one candidate motion information of one arbitrary sub-block should be used as a predictor based on the above-described AMVP mode motion vector derivation method. It can also be used as a predictor by using block candidate motion information. The contents of this step will be described in detail in Example 1 below.

In step S904, a history-based candidate is added. The contents of this step will be described in detail in Example 2 below.

In step S905, an average candidate between candidate motion information of the Merge/AMVP list is added. The contents of this step will be described in detail in Example 3 below.

After going through step S905, if the candidate motion information of the Merge/AMVP list is not filled up to the maximum number, after adding zero motion information in step S906 to fill up to the maximum number, the flowchart is terminated, and the candidate motion for each mode Complete the information list configuration. The candidate motion information described in this embodiment may be utilized in various prediction modes in addition to the Merge/AMVP mode. 9, the candidate list does not limit the order of added candidates. For example, temporal candidates in sub-block units may be added to the candidate list in preference to spatial candidates. Alternatively, the average candidate may be added to the candidate list in preference to the History-based candidate. In this specification, the candidate motion information list, the candidate motion information set, the motion information candidate group, and the candidate list may be understood to have the same meaning.

(Example 1)

In this embodiment, a method of deriving temporal candidates and sub-block unit temporal candidates in steps S902 and S903 of FIG. 9 will be described in detail. The temporal candidate means a block-unit temporal candidate and can be distinguished from a sub-block-unit temporal candidate. Here, the sub-block refers to a basic block unit for deriving motion information of a current block by dividing a block to be encoded or decoded (hereinafter, a current block) into a block having an arbitrary NxM (N, M ≥ 0) size. . The sub-block may have a pre-set size in the encoder and/or decoder. For example, the sub-block may be a square having a fixed size, such as 4x4 or 8x8. However, the present invention is not limited thereto, and the shape of the sub-block may be non-square, and at least one of the width and height of the sub-block may be greater than 8. The temporal candidate for each sub-block may be limited to be added to the candidate list only when the current block is larger than NxM. For example, when N and M are 8 each, only a case in which the width and height of the current block is greater than 8, a temporal candidate for each sub-block may be added to the candidate list.

10 is a basic conceptual diagram for explaining the present embodiment. FIG. 10 shows one current (sub) block of the current picture. The target (sub) block for the corresponding (sub) block is searched for in the target picture. Here, the target picture and the target (sub) block information may be transmitted in an upper header or a current block unit, respectively, and the target picture and the target (sub) block may be designated under the same conditions in the video encoding device and the video decoding device. After the target picture and the target (sub) block of the current (sub) block are determined, motion information of the current (sub) block is derived using the motion information of the target (sub) block.

Specifically, each sub-block of the current block has a corresponding relationship with each sub-block of the target block. The temporal candidate for each sub-block has motion information for each sub-block in the current block, and motion information of each sub-block can be derived using motion information of a sub-block having a corresponding relationship in the target block. However, there may be a case where motion information of a sub-block having the corresponding relationship is not available. In this case, motion information of the corresponding sub-block may be set as default motion information. Here, the default motion information may mean motion information of neighboring sub-blocks adjacent to the corresponding sub-block in a horizontal direction or a vertical direction. Alternatively, the default motion information may mean motion information of a sub-block including a central sample of the target block. However, the present invention is not limited thereto, and the default motion information may mean motion information of a sub-block including any one of n corner samples of the target block. n can be 1, 2, 3, or 4. Alternatively, among sub blocks including a central sample and/or sub blocks including n corner samples, a sub block having available motion information is searched according to a predetermined priority order, and motion information of the first searched sub block is defaulted. It can also be set as motion information.

On the other hand, it is possible to preemptively determine whether the above-described default motion information is available. As a result of the determination, if the default motion information is not available, the process of deriving the motion information of the temporal candidate for each sub-block and adding it to the candidate list may be omitted. That is, as long as the default motion information is available, temporal candidates for each sub-block can be derived and added to the candidate list.

Meanwhile, the motion vector of the motion information may mean a motion vector that has been scaled. The temporal distance of the reference picture of the target picture and the reference picture of the target (sub) block is TD, and the temporal distance of the reference picture of the current picture and the current (sub) block is TB, using Equation 2, the motion vector of the target (sub) block ( MV). The scaled motion vector (MV _scale ) is used to point the prediction (sub) block of the current (sub) block within the reference picture, or may be used as a motion vector of a temporal candidate or a sub-block temporal candidate of the current (sub) block. Can. However, when deriving the scaled motion vector, the variable MV used in Equation 2 means the motion vector of the target (sub) block, and the MV _scale means the scaled motion vector of the current (sub) block.

In addition, the reference picture information of the current (sub) block may be designated under the same conditions as the video encoding device and the video decoding device, and it is also possible to transmit the reference picture information of the current (sub) block in units of the current (sub) block. .

Hereinafter, a method of determining the target (sub) block of the current (sub) block under the same conditions in the video encoding apparatus and the video decoding apparatus will be described in more detail. The target (sub) block of the current (sub) block can be indicated using one of the candidate motion information in the Merge/AMVP candidate list. More specifically, after determining the prediction mode of the candidate motion information in the candidate list, priority may be given to the prediction mode. For example, by prioritizing in the order of AMVP mode, MERGE mode, and SKIP mode, one of the motion information in the candidate list can be selected to indicate the target (sub) block.

Also, simply, the first candidate motion information in the candidate list may be unconditionally selected to indicate a target (sub) block. In the case of candidate motion information coded in the same prediction mode, various priority conditions can be used, such as selecting in the order of highest priority on the candidate list. However, when the reference picture and the target picture of the candidate motion information are different, the candidate motion information may be excluded. Alternatively, the target (sub) block in the target picture corresponding to the same position as the current (sub) block may be determined.

Specifically, the target (sub) block may be determined as a block at a position shifted by a predetermined temporal motion vector (temporal MV) from the position of the current (sub) block. Here, the temporal motion vector may be set as a motion vector of neighboring blocks spatially adjacent to the current block. The neighboring block may be any one of the left, top, bottom left, top right, or top left blocks of the current block. Alternatively, the temporal motion vector may be derived by using only a peripheral block at a fixed position pre-promised in the encoding/decoding device. For example, the peripheral block in the fixed position may be the left block A1 of the current block. Alternatively, the peripheral block in the fixed position may be the upper block A2 of the current block. Alternatively, the peripheral block in the fixed position may be the lower left block A3 of the current block. Alternatively, the peripheral block in the fixed position may be the upper right block A4 of the current block. Alternatively, the peripheral block in the fixed position may be the upper left block A5 of the current block.

The setting may be performed only when the reference picture and the target picture of the neighboring block are the same (for example, when the POC difference between the reference picture and the target picture is 0). When the reference picture and the target picture of the neighboring block are not the same, the temporal motion vector may be set to (0,0).

The set temporal motion vector may be rounded based on at least one of a predetermined offset or shift value. Here, the offset is derived based on the shift value, and the shift value may include at least one of a right shift value (rightShift) or a left shift value (leftShift). The shift value may be an integer preset in the encoding/decoding device. For example, rightShift may be set to 4 and leftShift may be set to 0, respectively. For example, rounding of a temporal motion vector may be performed as shown in Equation 3 below.

[Equation 3]

offset = (rightShift = = 0)? 0: (1 << (rightShift-1))

mvX _R [0] = ((mvX[ 0] + offset-(mvX[ 0] >= 0)) >> rightShift) << leftShift

mvX _R [1] = ((mvX[ 1] + offset-(mvX[ 1] >= 0)) >> rightShift) << leftShift

Two conditions are assumed for a more detailed explanation. The first is to store and store motion information in 4x4 sub-block units within a picture that has already been coded (here, the 4x4 sub-block unit boundaries for storing motion information and the target (sub) block boundaries in the target picture coincide) ) The second is that the size of the sub-block in the current block is 4x4. The size of the aforementioned blocks can be variously determined. At this time, the target point indicated in the target picture using the motion information in the target (sub) block position in the current (sub) block or the target (sub) block in the Merge/AMVP candidate list of the current (sub) block ( When determining the position of the sub) block, the reference coordinates of each sub-block and the reference coordinates of the current (sub) block may not correspond to each 4x4 sub-block unit in which motion information is stored in the target picture. For example, the coordinates of the top left pixel of the current (sub) block are (12, 12), but the top left coordinate of the target (sub) block may be (8, 8), such a mismatch may occur. This is an unavoidable phenomenon that occurs because the block division structure of the target picture is different from the current picture.

11 is an exemplary diagram for explaining a method of determining a target block when deriving temporal candidate motion information in a current block unit rather than a sub-block unit. After the motion vector indicating the target position in the reference coordinates of the current block (the same position (ie, zero motion) or the motion vector derived from the candidate list) is determined, the point to which the motion vector points in the target picture is found. The motion information of the 4x4 target sub-block including the target point in the target picture can be used to derive the scaled motion vector of the current block. Or, if a plurality of reference coordinates are located in the current block, and target points indicated by each reference coordinate are searched, and if the target points all point to the same 4x4 target sub-block, the motion information of the target sub-block is derived from the scaled motion vector of the current block. It can be used to do this, but if it refers to a plurality of 4x4 target sub-blocks, it can also be used to derive the scaled motion vector of the current block by using the average motion information of each target sub-block. The target point may use two target positions of the current block center area, as in the example of FIG. 11, but it can be used anywhere other than the pixel position within the current block. Of course, there may be two or more motion information for calculating the average motion information. Alternatively, after deriving a plurality of scaled motion vectors using each of the plurality of target sub-blocks, a plurality of prediction blocks may be generated to weight the corresponding prediction blocks to generate a final prediction block.

12 and 13 are exemplary views for explaining a method of determining a target block when deriving temporal candidate motion information in units of sub-blocks in a current block. 12 is an exemplary diagram in a case where one reference position is in sub-block units, and FIG. 13 is an exemplary diagram when a plurality of reference positions are used in sub-block units.

In FIG. 12, the reference position in units of sub blocks is illustrated as the position of the upper left pixel of the sub block. After determining the target block of the same size as the current block based on the 4x4 target sub-block of the target picture, which is based on the reference coordinate of the bottom-right sub-block of the current block, targets between the current block and the same-position sub-blocks of the target block The scaled motion vector of the current sub-block may be derived using the motion information of the sub-blocks, respectively. Alternatively, when the target position of the target picture is determined through the motion vector indicating the target position in the reference coordinates of the sub-block, and when the motion information of the 4x4 target sub-block including the target position is derived to derive the scaled motion vector of the current sub-block. Can be used.

13 is an exemplary view for explaining a method of deriving a scaled motion vector using a plurality of target sub-blocks for each current sub-block. In the case of deriving the scaled motion vector of sub-block D as illustrated in FIG. 13, the target position in the target picture is determined based on a plurality of reference coordinates in sub-block D. Subsequently, when the target location points to the target sub-block having the same target, the scaled motion vector of the current sub-block can be derived using the motion information of the target sub-block, but as illustrated in FIG. 13, target locations have different target sub-blocks. In the case, the average motion information of each target sub-block may be calculated to derive a scaled motion vector of the current sub-block. In addition, after deriving different scaled motion vectors separately using motion information of each target sub-block, a prediction sub-block may be respectively generated, and a final prediction block may be generated by weighting each prediction sub-block. Other sub-blocks in the current block (sub-blocks A, B, and C) can also generate a predictive sub-block in the manner described above.

(Example 2)

In this embodiment, step 904 of FIG. 9 will be described in detail. The history-based candidate (hereinafter referred to as a'recovery information-based candidate') is a sequence, picture, or slice unit, and a motion information-based motion candidate storage buffer (hereinafter,'for storing motion information encoded/decoded before the current block). H buffer'). The buffer manages the coded motion information by updating the buffer in a FIFO (First-in First-out) method. The H buffer may be initialized in units of CTU, CTU rows, slices, or pictures, and when the current block is predicted and coded by motion information, corresponding motion information is updated to the H buffer. The motion information stored in the H buffer may be used as a Merge/AMVP candidate in step S904 of FIG. 9. When the candidate motion information of the H buffer is added to the candidate list, the most recently updated motion information may be added to the H buffer, and vice versa. Alternatively, according to the inter prediction mode, it may be determined in which order the motion information of the H buffer is extracted and added to the candidate list.

Specifically, for example, through redundancy check between the H buffer and the candidate list, motion information of the H buffer may be added to the candidate list. In the merge mode, redundancy check may be performed on a part of the merge candidate of the candidate list and a part of the motion information of the H buffer. A part of the candidate list may include a left block and an upper block among spatial merge candidates. However, the present invention is not limited thereto, and may be limited to any one of the spatial merge candidates, and may further include at least one of a lower left block, an upper right block, an upper left block, or a temporal merge candidate. Meanwhile, a part of the H buffer may mean m pieces of motion information most recently added to the H buffer. Here, m is 1, 2, 3 or more, and may be a fixed value pre-promised in the encoding/decoding device. It is assumed that 5 motion information is stored in the H buffer, and 1 to 5 indexes are allocated to each motion information. The larger the index, the later the stored motion information. At this time, redundancy between motion information having indexes 5, 4, and 3 and merge candidates of the candidate list may be checked. Alternatively, redundancy between motion information having indexes 5 and 4 and merge candidates of the candidate list may be checked. Alternatively, the redundancy between the motion information having the indexes 4 and 3 and the merge candidate of the candidate list may be checked, except for the motion information of the index 5 added last. As a result of the redundancy check, if there is even the same motion information, the motion information of the H buffer may not be added to the candidate list. On the other hand, if the same motion information does not exist, the motion information of the H buffer can be added to the last position of the candidate list. In this case, the motion information stored in the H buffer may be added to the candidate list in the order of the recently stored motion information (that is, from the largest index to the smallest index). However, the motion information (motion information having the largest index) stored last in the H buffer may be restricted so that it is not added to the candidate list.

On the other hand, in the case of the AMVP mode, the motion information (especially motion vector) stored first in the H buffer may be added to the candidate list in order. That is, among the motion information stored in the H buffer, motion information having a small index may be added to a candidate list before motion information having a large index.

Meanwhile, the motion vector stored in the H buffer may be added to the candidate list in the same way, or the motion vector to which the above-described rounding process is applied may be added to the candidate list. Rounding is to adjust the accuracy of candidate motion information so as to correspond to the accuracy of the motion vector of the current block. Referring to Equation 3, mvX _R may mean a motion vector to which a rounding process is applied, and mvX may mean a motion vector stored in an H buffer. Further, at least one of the shift value rightShift or leftShift may be determined in consideration of the accuracy (or resolution) of the motion vector. For example, when the accuracy of the motion vector is 1/4 sample, the shift value may be determined as 2, and in the case of 1/2 sample, the shift value may be determined as 3. When the accuracy of the motion vector is 1 sample, the shift value may be determined as 4, and in the case of 4 samples, the shift value may be determined as 6. RightShift and leftShift can be set to the same value.

When adding motion information stored in the H buffer to the Merge/AMVP candidate list, the number of motion information that can be added may be limited. For example, although the maximum number of candidates in the Merge/AMVP candidate list may be filled using motion information in the H buffer, only (maximum number of candidates-1) may be filled.

The number of candidate motion information stored in the H buffer may be determined under the same conditions in the video encoding device and the video decoding device, or may be transmitted from the upper header to the video decoding device.

Specifically, in the case of the Merge candidate list, only (maximum number of candidates-n) can be filled using the motion information of the H buffer. Here, n may be an integer of 1, 2, or more. The maximum number of candidates is determined by a fixed number (for example, 5, 6, 7, 8) pre-defined in the encoding/decoding device, or in signaled information to indicate the maximum number of candidates It may be determined variably on the basis of. Meanwhile, in the case of the AMVP candidate list, the maximum number of candidates may be filled using motion information of the H buffer. The maximum number of candidates in the AMVP candidate list may be 2, 3, 4, or more. In the case of the AMVP candidate list, unlike the Merge candidate list, the maximum number of candidates may not be variable.

For a first method of updating the H buffer, see FIG. 14. The H buffer updates motion information in the coding order of blocks in the first CTU row. From the second CTU row, it is the same to update the motion information in the coding order of the block, but the H buffer may be updated by further considering motion information stored in reconstructed blocks adjacent to the current CTU row in the upper CTU. In FIG. 14, mi in CTU is an abbreviation for Motion Information, and is reconstructed motion information stored in the bottom blocks in the remaining CTU lines except the last CTU line. Up to P (P is an integer greater than or equal to 1) motion information may be updated in the H buffer, and the update method may vary according to the unit in which the H buffer is initialized. In this embodiment, an H buffer update method according to initialization in CTU units and initialization in CTU row units will be described. First, when the H buffer is initialized in CTU units, each CTU in the second CTU row can initialize the H buffer using mi before starting coding. Here, the meaning of initialization means that any motion information is updated again in the completely empty H buffer. For example, before coding CTU ₈ , the H buffer can be initialized using the lower 4 mi of CTU ₃ . The order of updating mi can also be determined in various ways. You can update from the mi on the left now, or vice versa. When the H buffer is initialized in units of CTU rows, only the first CTU in each CTU row can be updated using the mi of the upper CTU. In addition, the H buffer may be emptied for each initialization unit to initialize. Meanwhile, when the most recently encoded/decoded motion information is the same as the motion information pre-stored in the H buffer, the most recent motion information may not be added to the H buffer. Alternatively, the same motion information as the most recent motion information may be removed from the H buffer, and the most recent motion information may be stored in the H buffer. At this time, the most recent motion information may be stored in the last position of the H buffer.

15 for a second method of updating the H buffer. The second method places an additional motion candidate storage buffer in addition to the H buffer. This buffer is a buffer (hereinafter referred to as a'V buffer') that stores the reconstructed motion information of the upper CTU (hereinafter referred to as'Vmi'). This V buffer may be available when the aforementioned H buffer is initialized in units of CTU rows or slices, and the V buffer may be initialized in units of CTU rows. In the current picture except for the last CTU line, Vmi in the V buffer must be updated for the lower CTU line. Here, up to Q (Q is an integer of 1 or more) motion information can be updated in the V buffer, and motion information corresponding to Vmi can be determined in various ways. For example, Vmi may be reconstructed motion information of a block containing the center coordinate of the CTU, or may be the most recent motion information included in the H buffer when coding a block containing the center coordinate in the upper CTU. have. There may be more than one Vmi updated in one CTU, and the updated Vmi is used to update the H buffer of the lower CTU. In the remaining CTU rows except the first CTU row, Vmi stored in the V buffer of the upper CTU row is updated to the current H buffer for each CTU row. When there are multiple Vmis stored in the V buffer of the upper CTU row, the first updated motion information can be retrieved first and updated in the H buffer or vice versa. The update time may be before coding each CTU, or before coding blocks that border the upper CTU in each CTU. At various times, the Vmi of the upper CTU stored in the V buffer can be updated to the H buffer. Can. In addition, the Vmi of the upper left CTU as well as the upper left CTU and the upper right CTU can be updated in the H buffer.

The above-described candidate motion information of the V buffer may be independently added to the MERGE/AMVP candidate list derivation process of FIG. 9. When adding V buffer candidate motion information, the corresponding priority in the MERGE/AMVP candidate list can be variously determined. For example, if there is valid motion information in the V buffer after step S904, the candidate motion information may be added between steps S904 and S905.

(Example 3)

In this embodiment, step 905 of FIG. 9 will be described in detail. If there is less than one motion information filled in the Merge/AMVP candidate list through steps S901 to S904, this step is omitted. If there is more than two motion information filled in the Merge/AMVP candidate list, the average candidate motion information may be generated between each candidate to fill the candidate list. The motion vector of the average candidate motion information is derived for each prediction direction (List 0 or List 1) of the motion information, and means a motion vector generated by averaging motion vectors in the same direction stored in the candidate list. When deriving the average candidate motion information, if the reference picture index information of the motion information used when averaging is different, the reference picture information of the motion information with high priority may be determined as the reference picture index information of the average candidate motion information. Alternatively, the reference picture information of the motion information having a low priority may be determined as reference picture index information of the average candidate motion information.

When adding the average candidate motion information to the Merge/AMVP candidate list, the generated average candidate motion information may be used when generating another average candidate motion information. See FIG. 16 to illustrate this example. In FIG. 16, the left table is a list of Merge/AMVP candidates before step S905 of FIG. 9, and the right table is a list of Merge/AMVP candidates after step S905 of FIG. Looking at the table on the left, the candidate motion information corresponding to 0 and 1 is filled in the two candidate lists. Using these two pieces of motion information, it is possible to generate average candidate motion information corresponding to number two. In the example of FIG. 16, the motion information of List 0 direction motion vector (1, 1) of candidate 0, the motion information of reference picture index 1 and the motion vector of List 0 direction of candidate 1 (3, -1), motion of reference picture index 0 The information is averaged to fill the motion information of candidate 2 in the List 0 direction. In the case of List 1, there is no motion information of candidate 0. In this case, the motion information of candidate No. 1 in which the candidate in the direction of List 1 is present is taken as it is and the motion information of candidate No. 1 in the direction of No. 2 is filled. If there is no motion information in one direction when deriving the average candidate motion information, the corresponding direction is not separately derived. Additional average candidate motion information may be generated using the derived average motion information of the two times. The average candidate motion information of number 3 is the average candidate motion information of candidates 0 and 2, and the average candidate motion information of number 4 is the average candidate motion information of candidates 1 and 2. The method for generating the average candidate motion information is as described above.

Duplicate candidate motion information may exist in the Merge/AMVP candidate list before step S905 of FIG. 9. The average candidate motion information may be used to remove such duplicate candidate motion information. 17 shows this example. The left and right tables are as described in FIG. 16. Looking at the table on the left, the candidate motion information for 0 and 2 is completely the same. In this case, the second candidate motion information having a low priority may be replaced with the average candidate motion information. In the example of FIG. 17, the existing candidate motion information is replaced with the average candidate motion information of 0 and 1, 3 is the average candidate motion information of 0 and 2, and 4 is the candidate using the average candidate motion information of 1 and 2 The list is filled.

The number of Merge/AMVP candidate lists that can be filled using the average candidate motion information may also be limited. For example, it is possible to fill up to the maximum number of candidates in the Merge/AMVP candidate list using the average candidate motion information, but only up to (the maximum number of candidate lists-1). In addition, three or more candidate motion information may be used when calculating the average candidate motion information, and median information, not information obtained by averaging three or more candidate motion information, may be determined as the average candidate motion information.

Although the above-described embodiment has been mainly focused on the operation in the decoding apparatus, the same or similar operation can be performed in the encoding apparatus, and detailed description will be omitted.

Various embodiments of the present disclosure are not intended to list all possible combinations, but are intended to describe representative aspects of the present disclosure, and the items described in various embodiments may be applied independently or in combination of two or more.

In addition, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. For implementation by hardware, one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), Universal It can be implemented by a processor (general processor), a controller, a microcontroller, a microprocessor.

The scope of the present disclosure includes software or machine-executable instructions (eg, operating systems, applications, firmware, programs, etc.) that cause an operation according to the method of various embodiments to be executed on a device or computer, and such software or Instructions include a non-transitory computer-readable medium that is stored and executable on a device or computer.

Claims (15)

Generating a candidate list of the current block; And
And performing inter prediction of the current block by using any one of a plurality of candidates belonging to the candidate list,
The plurality of candidates include at least one of a spatial candidate, a temporal candidate, or a reconstruction information-based candidate,
The reconstructed information-based candidate is added from a buffer that stores motion information decoded before the current block. According to claim 1,
The motion information stored in the buffer is added to the candidate list in the order of motion information stored in the buffer later, or added to the candidate list in the order of motion information stored in the buffer first. According to claim 2,
The order or number of motion information stored in the buffer added to the candidate list is determined differently according to a prediction mode between screens of the current block. According to claim 3,
The candidate list is filled with motion information stored in the buffer until the maximum number of candidates in the candidate list is reached, or the motion information stored in the buffer is reached until the maximum number of candidates minus one is reached. A video decoding method, which is filled by using. According to claim 1,
The buffer is initialized in any one unit of a coding tree unit (CTU), CTU row, slice or picture. Generating a candidate list of the current block; And
And performing inter prediction of the current block by using any one of a plurality of candidates belonging to the candidate list,
The plurality of candidates include at least one of a spatial candidate, a temporal candidate, or a reconstruction information-based candidate,
The reconstruction information-based candidate is added from a buffer that stores motion information decoded before the current block. The method of claim 6,
The motion information stored in the buffer is added to the candidate list in the order of motion information stored in the buffer later, or added to the candidate list in the order of motion information stored in the buffer first. The method of claim 7,
The order or number of motion information stored in the buffer added to the candidate list is determined differently according to a prediction mode between screens of the current block. The method of claim 8,
The candidate list is filled with motion information stored in the buffer until the maximum number of candidates in the candidate list is reached, or the motion information stored in the buffer is reached until the maximum number of candidates minus one is reached. Filled using, video encoding method. The method of claim 6,
The buffer is initialized in any one unit of a coding tree unit (CTU), CTU row, slice, or picture. A computer-readable recording medium storing a bitstream decoded by an image decoding method, the method comprising:
The video decoding method,
Generating a candidate list of the current block; And
And performing inter prediction of the current block by using any one of a plurality of candidates belonging to the candidate list,
The plurality of candidates include at least one of a spatial candidate, a temporal candidate, or a reconstruction information-based candidate,
The reconstructed information-based candidate is a computer-readable recording medium added from a buffer storing motion information decoded before the current block. The method of claim 11,
The computer-readable recording medium, the motion information stored in the buffer is added to the candidate list in the order of motion information stored later in the buffer, or added to the candidate list in the order of motion information stored in the buffer first. The method of claim 12,
The order or number of motion information stored in the buffer added to the candidate list is determined differently according to a prediction mode between screens of the current block. The method of claim 13,
The candidate list is filled with motion information stored in the buffer until the maximum number of candidates in the candidate list is reached, or the motion information stored in the buffer is reached until the maximum number of candidates minus one is reached. A computer-readable recording medium that is filled using. The method of claim 11,
The buffer is initialized by a unit of a coding tree unit (CTU), CTU row, slice, or picture, and a computer-readable recording medium.

Images