RU2787882C2 - Method and device for processing video signals using inter-prediction - Google Patents

Abstract

FIELD: vector prediction technologies.

SUBSTANCE: inventions group relates to technologies for performing inter-prediction using history-based motion vector prediction. A method for processing a video signal based on inter-prediction is proposed. The method comprises generating a first merge candidate list for the current block based on the spatial merge candidate and the temporal merge candidate of the current block. Next, according to the method, a second merge candidate list is generated by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list. In doing so, a pruning check is performed for a history-based merge candidate only on specific spatial merge candidates included in the first list of merge candidates.

EFFECT: improving the efficiency of video signal processing based on inter-prediction.

11 cl, 50 dwg, 8 tbl

Description

Technical field

[1] An embodiment of the disclosure relates to a method and apparatus for processing a video signal using inter prediction, and more specifically, to a method and apparatus for performing inter prediction using history-based motion vector prediction.

Prior Art

[2] Compression encoding means a series of signal processing methods for transmitting digitized information over a communication line, or methods for storing information in a form suitable for a storage medium. A medium including a picture, image, audio, etc. may be a target for compression encoding, and specifically, a method for performing compression encoding on a picture is referred to as video image compression.

[3] Next generation video content is expected to have the characteristics of high spatial resolution, high frame rate, and high dimensionality of scene representation. To process such contents, a significant increase in memory capacity, memory access speed and processing power would be required.

[4] Accordingly, it is necessary to develop an encoding tool for efficient processing of next generation video content.

Disclosure

Technical problem

[5] An embodiment of the disclosure proposes restrictions on the redundancy check for adding a potentially suitable HMVP variant (candidate) to the merge list (or AMVP list).

[6] The technical objectives to be achieved in the embodiment of the disclosure are not limited to the aforementioned technical objectives, and other technical objectives not described above may be readily apparent to one skilled in the art to which the disclosure relates from the following description.

Technical solution

[7] In one aspect of an embodiment of the disclosure, a video signal processing method based on inter-prediction includes generating a first merge (merge) candidate list of a current block using a spatial merge candidate and a temporal merge candidate of the current block, generating a second merge candidate list by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list, obtaining a merge index indicating the merge candidate applied to the current block in the second merge candidate list, and generating a prediction block of the current block using movement information of the merge candidate indicated by the merge index. A history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of the predetermined merge candidate among the merge candidates included in the first merge candidate list.

[8] Preferably, a history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of a predetermined specific number of merge candidates among the merge candidates included in the first merge candidate list.

[9] Preferably, a history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of a specific spatial merge candidate included in the first merge candidate list.

[10] Preferably, a history-based merge candidate is derived from a predetermined number of candidates in a history-based buffer storing history-based merge candidates.

[11] Preferably, generating the second merge candidate list may further include adding a history-based spatio-temporal merge candidate. A history-based spatial temporal candidate can be derived using two history-based merging candidates and a temporal merging candidate.

[12] Preferably, the motion vector of the history-based spatial temporal candidate may be derived as the average of the motion vectors of the two history-based merge candidates and the average of the motion vectors of the temporal merge candidates.

[13] In one aspect of the embodiment, an apparatus for processing a video signal based on inter prediction includes a memory storing the video signal and a processor associated with the memory. The processor may be configured to generate a first merge candidate list of the current block using the spatial merge candidate and the temporal merge candidate of the current block, generate a second merge candidate list by adding a history-based merge candidate indicating motion information of the block encoded before the current block to the first a merge candidate list, obtain a merge index indicating a merge candidate applied to the current block in the second merge candidate list, and generate a current block prediction block using motion information of the merge candidate indicated by the merge index. A history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of the predetermined merge candidate among the merge candidates included in the first merge candidate list.

Useful results

[14] According to an embodiment of the disclosure, the complexity score according to the redundancy check can be improved, and the efficiency can be improved by limiting the redundancy check to be added to the merge list (or AMVP list).

[15] The effects that can be obtained in the disclosure are not limited to the above effects, and other technical effects not described above can be clearly understood by a person skilled in the art to which the disclosure pertains from the following description.

Description of drawings

[16] The accompanying drawings, which are included as part of the detailed description to assist in understanding the disclosure, provide embodiments of the disclosure and describe the specifications of the disclosure along with the detailed description.

[17] FIG. 1 illustrates an example of the functional elements of an encoder as an example of a device for processing a video signal according to an embodiment of the disclosure.

[18] FIG. 2 is an embodiment in which the disclosure is applied, and is a block diagram of a decoding apparatus in which a video/image signal is decoded.

[19] FIG. 3 is an embodiment in which expansion can be applied, and is a diagram illustrating an example of a multiple type tree structure.

[20] FIG. 4 is an embodiment in which expansion can be applied, and is a diagram illustrating a mechanism for signaling partitioning information of splitting a multiple type nested tree quadtree structure.

[21] FIG. 5 is an embodiment in which the disclosure may be applied, and is a diagram illustrating a method for dividing a CTU into a plurality of CUs based on a quadtree and nested multiple type tree structure.

[22] FIG. 6 is an embodiment in which expansion can be applied, and is a diagram illustrating a method for limiting the splitting of a binary tree.

[23] FIG. 7 is an embodiment in which expansion may be applied, and is a diagram illustrating patterns of redundant splitting that may occur in binary tree splitting and ternary tree splitting.

[24] FIG. 8 and 9 are diagrams illustrating an inter prediction-based video/image encoding method according to an embodiment of the disclosure and an inter predictor in an encoding apparatus according to an embodiment of the disclosure.

[25] FIG. 10 and 11 are diagrams illustrating an inter-prediction-based video/image decoding method according to an embodiment of the disclosure and an inter-predictor in a decoding apparatus according to an embodiment of the disclosure.

[26] FIG. 12 is an embodiment in which expansion is applied, and is a diagram for describing adjacent blocks used in the combine mode or the skip mode.

[27] FIG. 13 is a flowchart illustrating a merge candidate list generation method according to an embodiment in which disclosure is applied.

[28] FIG. 14 is a flowchart illustrating a merge candidate list generation method according to an embodiment in which disclosure is applied.

[29] FIG. 15 illustrates an example of motion patterns in accordance with an embodiment of the disclosure.

[30] FIG. 16 illustrates an example of a reference point motion vector for affine motion prediction in accordance with an embodiment of the disclosure.

[31] FIG. 17 illustrates an example of a motion vector for each sub-block of a block to which affine motion prediction has been applied in accordance with an embodiment of the disclosure.

[32] FIG. 18 illustrates an example of adjacent blocks used for affine motion prediction in affine pooling mode, in accordance with an embodiment of the disclosure.

[33] FIG. 19 illustrates an example of a block on which affine motion prediction is performed using neighboring blocks to which affine motion prediction has been applied in accordance with an embodiment of the disclosure.

[34] FIG. 20 is a diagram for describing a method for generating a merge candidate list using a neighboring affine coding block according to an embodiment of the disclosure.

[35] FIG. 21 and 22 are diagrams for describing a method for generating a merge affine candidate list using an affine prediction encoded neighbor block according to an embodiment of the disclosure.

[36] FIG. 23 illustrates an example of neighboring blocks used for affine motion prediction in the affine inter-mode, in accordance with an embodiment of the disclosure.

[37] FIG. 24 illustrates an example of neighboring blocks used for affine motion prediction in the affine inter-mode, in accordance with an embodiment of the disclosure.

[38] FIG. 25 and 26 are diagrams illustrating a motion vector candidate derivation method using neighbor block motion information in affine inter mode according to an embodiment of the disclosure.

[39] FIG. 27 illustrates an example of a subblock unit affine motion vector field in accordance with an embodiment of the disclosure.

[40] FIG. 28 illustrates a method for generating a prediction block and an inter-prediction motion vector to which an affine motion model has been applied in accordance with an embodiment of the disclosure.

[41] FIG. 29 is a diagram illustrating a method for performing motion compensation based on a motion vector of a reference point according to an embodiment of the disclosure.

[42] FIG. 30 is a diagram illustrating a method for performing motion compensation based on the motion vector of a reference point in a non-square block according to an embodiment of the disclosure.

[43] FIG. 31 is a diagram illustrating a method for performing motion compensation based on the motion vector of a reference point in a non-square block according to an embodiment of the disclosure.

[44] FIG. 32-38 are diagrams illustrating a method for performing motion compensation based on the motion vector of a reference point in a non-square block according to an embodiment of the disclosure.

[45] FIG. 39 is a flowchart for describing an HMVP storage method according to an embodiment of the disclosure.

[46] FIG. 40 is a diagram for describing an HMVP table operating in an unrestricted FIFO method according to an embodiment of the disclosure.

[47] FIG. 41 is a diagram for describing an HMVP table operating in a limited FIFO method according to an embodiment of the disclosure.

[48] FIG. 42 is a diagram illustrating an HMVP LUT and a long-term HMVP LUT according to an embodiment of the disclosure.

[49] FIG. 43 is a diagram illustrating an example of a method for updating an HMVP LUT according to an embodiment of the disclosure.

[50] FIG. 44 is a diagram illustrating a method for limiting the number of HMVP candidates, that is, a pruning test target, according to an embodiment of the disclosure.

[51] FIG. 45 is a diagram illustrating an example of a method for performing clipping checking according to an embodiment of the disclosure.

[52] FIG. 46 is a diagram for describing a method for deriving an H-STMVP candidate using motion vectors that refer to different reference pictures according to an embodiment of the disclosure.

[53] FIG. 47 is a flowchart illustrating an inter-prediction based video signal processing method according to an embodiment to which the disclosure applies.

[54] FIG. 48 illustrates an example block diagram of an apparatus for processing an image signal according to an embodiment of the disclosure.

[55] FIG. 49 illustrates a video coding system in which expansion is applied.

[56] FIG. 50 is an embodiment in which the disclosure is applied and illustrates the structure of a content streaming (streaming) system.

Mode of carrying out the invention

[57] Next, preferred embodiments of the disclosure will be described by reference to the accompanying drawings. The description that will be set forth below with the accompanying drawings is intended to describe exemplary embodiments of the disclosure and is not intended to describe a single embodiment in which the disclosure may be implemented. The description below includes specific details to provide a perfect understanding of the disclosure. However, those skilled in the art will appreciate that the disclosure may be practiced without specific details.

[58] In some cases, to prevent obscurity of the technical concept of the disclosure, structures or devices that are commonly known may be omitted or may be depicted as a block diagram centered on the basic functions of the structures or devices.

[59] Additionally, although general terms currently in common use are chosen as terms in the disclosure, to the extent possible, a term that is arbitrarily chosen by the Applicant is used on a case-by-case basis. Since the meaning of the term will be clearly described in the relevant part of the description in such a case, it is understood that the disclosure should not simply be interpreted in terms only used in the description of the disclosure, but the meaning of the terms should be clarified.

[60] Specific terms used in the description below may be provided to aid understanding of the disclosure. Moreover, specific terms may be modified into other forms within the scope of the technical concept of the disclosure. For example, signal, data, sample, picture, frame, block, etc. may be appropriately replaced and interpreted in each encoding process.

[61] Hereinafter, in this specification, a “processing unit” means a unit in which an encoding/decoding processing such as prediction, transformation, and/or quantization is performed. Hereinafter, for convenience of description, a processing unit may also be referred to as a “processing unit” or a “block”.

[62] A processing unit can be understood as having a value including a unit for a luma component and a unit for a chrominance component. For example, a processing unit may correspond to a coding tree unit (CTU), a coding unit (CU), a prediction unit (PU), or a transformation unit (TU).

[63] Moreover, a processing unit can be understood as being a unit for a luma component or a unit for a chrominance component. For example, a processing unit may correspond to a coding tree block (CTB), a coding block (CB), a prediction block (PB), or a transform block (TB) for a luminance component. Alternatively, a processing unit may correspond to a coding tree block (CTB), a coding block (CB), a prediction block (PB), or a transform block (TB) for a chrominance component. Moreover, the disclosure is not limited to this, and a processing unit can be understood as a value including a unit for a luminance component and a unit for a chrominance component.

[64] Moreover, the processing unit is not necessarily limited to a square block and may be created in the form of a polygon having three or more vertices.

[65] Moreover, further in this specification, a pixel, a picture element, etc. usually called a sample. Moreover, using sampling can mean using a pixel value, a picture element value, or the like.

[66] FIG. 1 illustrates an example of the functional elements of an encoder as an example of a device for processing a video signal according to an embodiment of the disclosure.

[67] With reference to FIG. 1, encoder 100 may be configured to include an image splitter 110, a subtractor 115, a transform 120, a quantizer 130, a dequantizer 140, an inverse transform 150, an adder 155, a filter 160, a memory 170, an inter predictor 180, an intra predictor 185 and entropy encoder 190. Inter predictor 180 and intra predictor 185 may be collectively referred to as a predictor. In other words, the predictor may include an inter predictor 180 and an intra predictor 185. Transformer 120, quantizer 130, dequantizer 140, and inverse transform 150 may be included in the residual processor. The residual processor may further include a subtractor 115. In one embodiment, image splitter 110, subtractor 115, transform 120, quantizer 130, dequantizer 140, inverse transform 150, adder 155, filter 160, inter-predictor 180, intra-predictor 185 and entropy encoder 190 may be configured as a single hardware component (eg, encoder or processor). Moreover, memory 170 may include a decoded picture buffer (DPB) and may be configured with a digital storage medium.

[68] The image separator 110 may divide the input image (or picture or frame) input to the encoding device 100 into one or more processing units. For example, a processing unit may be referred to as a coding unit (CU). In this case, a coding unit may be recursively divided from a coding tree unit (CTU) or a largest coding unit (LCU) based on a quadtree/binary tree (QTBT) structure. For example, one coding unit may be partitioned into multiple deeper depth coding units based on a quadtree structure and/or a binary tree structure. In this case, for example, the quadtree structure may be applied first, and then the binary tree structure may be applied. Alternatively, the binary tree structure may be applied first. The coding procedure according to the disclosure may be performed based on a final coding unit that is no longer divisible. In this case, the largest coding unit may be directly used as the final coding unit based on coding efficiency according to image characteristics, or the coding unit may be recursively divided into deeper coding units, if necessary. Accordingly, a coding unit having an optimal size can be used as a final coding unit. In this case, the encoding procedure may include a procedure such as prediction, transformation, or restoration described later. As another example, the processing unit may further include a prediction unit (PU) or a transformation unit (TU). In this case, each of the prediction unit and the transformation unit may be split or split from each final coding unit. The prediction unit may be a unit for predicting a sample, and the transform unit may be a unit from which a transform coefficient is derived and/or a unit in which a residual signal is derived from a transform coefficient.

[69] A unit may be used interchangeably with a block or area, as appropriate. In general, an M×N block may indicate a set of samples configured with M columns and N rows or a set of transform coefficients. In general, a sample may indicate a pixel or pixel value, and may indicate only a luminance component pixel/pixel value, or only a chrominance component pixel/pixel value. In a sample, one picture (or image) can be used as a term corresponding to a pixel or pel (image element).

[70] Encoder 100 can generate a residual signal (residual block or residual sample array) by subtracting the prediction signal (predicted block or prediction sample array) output by inter-predictor 180 or intra-predictor 185 from an input image signal (original block or array of initial samples). The generated residual signal is passed to transducer 120. In this case, as illustrated, the unit in which the prediction signal (prediction block or array of prediction samples) is subtracted from the input image signal (source block or array of source samples) at encoder 100 may be referred to as subtractor 115 The predictor may perform prediction on a target processing block (hereinafter referred to as the current block) and may generate a predicted block including prediction samples for the current block. The predictor may determine whether intra-prediction is applied or whether inter-prediction is applied in the current block or CU. The predictor may generate various pieces of prediction information, such as prediction mode information, as will be described later in the description of each prediction mode, and may transmit the information to the entropy encoder 190. The prediction information may be encoded in the entropy encoder 190 and may be output in the form of a bitstream. .

[71] Intra-predictor 185 may predict the current block based on the samples in the current picture. The reference samples may be adjacent to the current block or may be spaced apart from the current block depending on the prediction mode. In intra-prediction, the prediction modes may include a plurality of non-angular modes and a plurality of angular modes. The non-angular mode may include, for example, a DC mode and a planar mode. The angle mode may include, for example, 33 angle prediction modes or 65 angle prediction modes, depending on the accuracy degree of the prediction direction. In this case, angle prediction modes greater than or less than 33 prediction angle modes or 65 prediction angle modes may be used, for example, depending on the configuration. The intra predictor 185 may determine the prediction mode applied to the current block using the prediction mode applied to the adjacent block.

[72] The inter predictor 180 may output a predicted block for the current block based on a reference block (reference sample array) specified by a motion vector in the reference picture. In this case, in order to reduce the amount of motion information transmitted in the inter-prediction mode, the motion information may be predicted as a block, sub-block, or sample unit based on the correlation of the motion information between the neighboring block and the current block. The motion information may include a motion vector and a reference picture index. The motion information may further include inter prediction direction information (L0 prediction, L1 prediction, Bi prediction). In the case of inter-prediction, the neighboring block may include a spatial neighboring block in the current picture and a temporal neighboring block in a reference picture. The reference picture including the reference block and the reference picture including the temporal neighbor block may be the same or different. The temporary neighbor block may be referred to as a so-called collocated reference block or collocated CU (colCU). A reference picture including a temporal neighboring block may be referred to as a colPic. For example, inter predictor 180 may generate a list of candidate (i.e., potentially suitable) motion information based on adjacent blocks, and may generate information indicating which candidate is used to derive the motion vector and/or the reference picture index of the current block. Inter-prediction may be performed based on various prediction modes. For example, in the case of the skip mode and the merge mode, the inter predictor 180 may use the motion information of the neighboring block as the motion information of the current block. In the case of the skip mode, unlike the combine mode, the residual signal may not be transmitted. In the case of a motion information prediction (MVP) mode, a neighboring block motion vector may be used as a motion vector predictor. The motion vector of the current block may be indicated by motion vector difference signaling.

[73] The prediction signal generated by the inter-predictor 180 or intra-predictor 185 may be used to generate the recovered signal or the residual signal.

[74] Converter 120 can generate transform coefficients by applying a transform circuit to the residual signal. For example, the transformation scheme may include at least one of a discrete cosine transform (DCT), a discrete sine transform (DST), a Karhunen-Loeve transform (KLT), a graph-based transform (GBT), or a conditionally non-linear transform (CNT). In this case, GBT means a transformation obtained from a graph if the relationship information between pixels is represented as a graph. CNT means a transformation obtained based on a prediction signal generated using all previously reconstructed pixels. Moreover, the transformation process may be applied to blocks of pixels having the same size of a square shape, or may be applied to blocks having variable sizes of a non-square shape.

[75] The quantizer 130 may quantize the transform coefficients and transmit them to the entropy encoder 190. The entropy encoder 190 may encode the quantized signal (quantized transform coefficient information) and output it in the form of a bitstream. Information about the quantized transform coefficients may be referred to as residual information. The quantizer 130 may reorder the quantized transform coefficients of the block shape in the form of a one-dimensional vector based on the scan sequence of the coefficients, and may generate information about the quantized transform coefficients based on the quantized transform coefficients of the one-dimensional vector shape. Entropy encoder 190 may perform various coding techniques such as exponential Golomb coding, context adaptive variable length coding (CAVLC), and context adaptive binary arithmetic coding (CABAC). Entropy encoder 190 may encode information (eg, syntax element values) needed to reconstruct the video/image, in addition to the quantized transform coefficients, either alone or separately. Encoded information (eg, encoded video/image information) may be transmitted or stored in a network abstraction layer (NAL) module in the form of a bitstream. The bitstream may be transmitted over a network or may be stored in a digital storage medium. In this case, the network may include a broadcast network and/or a communication network. The digital storage media may include various storage media such as USB, SD, CD, DVD, Blueray, HDD, and SSD. A transmitter (not illustrated) that transmits the signal outputted by the entropy encoder 190 and/or a store (not illustrated) for storing the signal may be configured as an internal/external element of the encoder 100, or the transmitter may be an element of the entropy encoder 190.

[76] The quantized transform coefficients output by quantizer 130 may be used to generate a prediction signal. For example, the residual signal can be recovered by applying dequantization and inverse transform to the quantized transform coefficients using dequantizer 140 and inverse transform 150 in the loop. Summer 155 may add the reconstructed residual signal to the prediction signal output by inter-predictor 180 or intra-predictor 185 so that a reconstructed signal (reconstructed picture, reconstructed block, or reconstructed sample array) can be generated. The predicted block may be used as the reconstructed block if there is no remainder for the target processing block, as in the case where the skip mode was applied. The adder 155 may be referred to as a recovery module or a recovery block generator. The generated reconstructed signal may be used to intra-predict the next processing target in the current picture, and may be used to inter-predict the next picture through filtering, as will be described later.

[77] Filter 160 can improve subjective/objective picture quality by applying filtering to the reconstructed signal. For example, filter 160 may generate a modified reconstructed picture by applying various filtering methods to the reconstructed picture. The modified reconstructed picture may be stored in memory 170, in particular in DPB memory 170. Various filtering techniques may include, for example, deblocking filtering, adaptive sample bias, adaptive loop filter, and bilateral filter. The filter 160 may generate various pieces of information for filtering, as will be described below in describing each filtering method, and may transmit them to the entropy encoder 190. The filter information may be encoded by the entropy encoder 190 and output in the form of a bitstream.

[78] The modified reconstructed picture transmitted to the memory 170 may be used as a reference picture in the inter-predictor 180. The encoder may avoid prediction inconsistency in the encoder 100 and the decoder, and improve encoding efficiency if inter-prediction is applied.

[79] The DPB of the memory 170 may store a modified reconstructed picture to use the modified reconstructed picture as a reference picture in the inter predictor 180. The memory 170 may store the motion information of the block from which the motion information in the current picture has been derived (or encoded), and /or information about the movement of blocks in the already restored picture. The stored motion information may be transmitted to the inter predictor 180 to be used as spatial neighbor block motion information or temporal neighbor block motion information. Memory 170 may store reconstructed samples of reconstructed blocks in the current picture and may transmit them to intra-predictor 185.

[80] FIG. 2 is an embodiment in which the disclosure is applied, and is a block diagram of a decoding apparatus in which a video/image signal is decoded.

[81] With reference to FIG. 2, decoder 200 may be configured to include entropy decoder 210, dequantizer 220, inverse transform 230, adder 235, filter 240, memory 250, inter predictor 260, and intra predictor 265. Inter predictor 260 and intra predictor 265 may be collectively referred to as a predictor. That is, the predictor may include an inter predictor 180 and an intra predictor 185. Dequantizer 220 and inverse transform 230 may be collectively referred to as a residual processor. That is, the residual processor may include a dequantizer 220 and an inverse transform 230. Entropy decoder 210, dequantizer 220, inverse transform 230, adder 235, filter 240, inter predictor 260, and intra predictor 265 may be configured as a single hardware component ( eg, decoder or processor) according to an embodiment. Moreover, memory 170 may include a decoded picture buffer (DPB) and may be configured with a digital storage medium.

[82] When a bitstream including video/image information is input, the decoding apparatus 200 may reconstruct the image according to the video/image information processing process in the encoding apparatus of FIG. 1. For example, the decoding device 200 may perform decoding using a processing unit applied in the encoding device. Accordingly, the processing unit for decoding may be, for example, a coding unit. The coding unit may be divided from the coding tree unit or the largest coding unit depending on the quadtree structure and/or binary tree structure. Moreover, the reconstructed image signal decoded and output by the decoding apparatus 200 can be reproduced by the playback apparatus.

[83] The decoder 200 may receive the signal output by the encoder of FIG. 1 in the form of a bitstream. The received signal may be decoded by entropy decoder 210. For example, entropy decoder 210 may output information (eg, video/image information) for image reconstruction (or picture reconstruction) by bitstream parsing. For example, entropy decoder 210 may decode information in the bitstream based on an encoding method such as Exponential Golomb coding, CAVLC, or CABAC, and may output a syntax element value for image reconstruction or quantized values of transform coefficients relative to the remainder. More specifically, in the CABAC entropy decoding method, a bin corresponding to each syntax element may be received from a bitstream; the context model may be determined using decoding the target syntax element information and decoding the neighborhood information and decoding the target block or character/bin information decoded in the previous step; the probability that the bin will appear may be predicted based on a certain context model, and a symbol corresponding to the value of each syntax element may be generated by performing arithmetic decoding on the bin. In this case, in the CABAC entropy decoding method, after determining the context model, the context model may be updated using the symbol/bin information decoded for the context model of the next symbol/bin. The prediction information among the information decoded in the entropy decoder 2110 may be provided to a predictor (inter-predictor 260 and intra-predictor 265). Parameter information related to the residual value on which entropy decoding was performed in the entropy decoder 210, i.e., quantized transform coefficients, may be input to the dequantizer 220. Moreover, filtering information among the information decoded in the entropy decoder 210 may be provided to the filter 240. Meanwhile, a receiver (not illustrated) that receives a signal outputted by the encoder may be further configured as an internal/external element of the decoder 200, or the receiver may be an element of the entropy decoder 210.

[84] The dequantizer 220 may dequantize the quantized transform coefficients and output the transform coefficients. The dequantizer 220 may reorder the quantized transform coefficients in the form of a two-dimensional block. In this case, the reordering may be performed based on the coefficient scan sequence performed in the encoder. The dequantizer 220 may perform dequantization on the quantized transform coefficients using a quantization parameter (eg, quantization step size information) and may obtain the transform coefficients.

[85] The detransformer 230 obtains the residual signal (residual block or array of residual samples) by inverse transforming the transform coefficients.

[86] The predictor may perform prediction on the current block and may generate a predicted block including the prediction samples for the current block. The predictor may determine whether intra-prediction is applied or whether inter-prediction is applied to the current block based on the prediction information output by the entropy decoder 210, and may determine the detailed intra/inter prediction mode.

[87] Intra-predictor 265 may predict the current block based on the samples in the current picture. The reference samples may be adjacent to the current block or may be spaced apart from the current block depending on the prediction mode. In intra-prediction, the prediction modes may include a plurality of non-angular modes and a plurality of angular modes. The intra-predictor 265 may determine the prediction mode applied to the current block using the prediction mode applied to the neighboring block.

[88] The inter predictor 260 may derive a predicted block for the current block based on a reference block (reference sample array) specified by a motion vector in the reference picture. In this case, in order to reduce the amount of motion information transmitted in the inter-prediction mode, the motion information may be predicted as a block, sub-block, or sample unit based on the correlation of the motion information between the neighboring block and the current block. The motion information may include a motion vector and a reference picture index. The motion information may further include inter prediction direction information (L0 prediction, L1 prediction, Bi prediction). In the case of inter-prediction, the neighboring block may include a spatial neighboring block in the current picture and a temporal neighboring block in a reference picture. For example, inter predictor 260 may configure a list of motion information candidates based on adjacent blocks, and may derive a motion vector and/or a reference picture index of the current block based on the received candidate selection information. Inter-prediction may be performed based on various prediction modes. The prediction information may include information indicating the inter-prediction mode for the current block.

[89] The adder 235 may generate a reconstructed signal (reconstructed picture, reconstructed block, or reconstructed array of samples) by summing the obtained residual signal with the prediction signal (predicted block or array of prediction samples) output by inter-predictor 260 or intra-predictor 265. Predicted Block can be used as a reconstructed block if there is no remainder for the target processing block, as in the case where the skip mode was applied.

[90] The adder 235 may be referred to as a recovery module or a recovery block generator. The generated reconstructed signal may be used to intra-predict the next processing target in the current picture, and may be used to inter-predict the next picture through filtering, as will be described later.

[91]

[92] Filter 240 can improve subjective/objective picture quality by applying filtering to the reconstructed signal. For example, filter 240 may generate a modified reconstructed picture by applying various filtering methods to the reconstructed picture, and may transmit the modified reconstructed picture to memory 250, specifically, DPB memory 250. Various filtering methods may include, for example, deblocking filtering, adaptive sampling bias , adaptive loop filter and bidirectional filter.

[93] The reconstructed picture stored (modified) in the DPB memory 250 may be used as a reference picture in the inter predictor 260. The memory 250 may store the motion information of the block from which the motion information in the current picture has been derived (or decoded), and /or information about the movement of blocks in the already restored picture. The stored motion information may be transmitted to the inter predictor 260 to be used as spatial neighbor block motion information or temporal neighbor block motion information. Memory 170 may store reconstructed samples of reconstructed blocks in the current picture and may transmit them to intra-predictor 265.

[94] In the disclosure, the embodiments described for filter 160, inter-predictor 180, and intra-predictor 185 of encoder 100 may be applied to filter 240, inter-predictor 260, and intra-predictor 265 of decoder 200, in the same or corresponding manner. .

[95]

[96] Block splitting

[97] The video/image coding method according to the present document may be performed based on various detailed technologies. Each of the detailed technologies is schematically described below. Those skilled in the art will appreciate that the following technologies may be associated with related procedures such as prediction, residual processing ((inverse) transform, (de-)quantization), syntax element encoding, filtering, and partitioning/split in the encoding procedure /video/image decoding, which has been described and/or will be described later.

[98] The block splitting procedure according to the present document may be performed in the image separator 110 of the encoder. The partition-related information may be (encoded) processed in the entropy encoder 190 and transmitted to the decoder in the form of a bitstream. The entropy decoder 210 of the decoding apparatus may derive a block split structure of the current picture based on split related information obtained from the bitstream, and may perform a series of procedures (e.g., prediction, residual processing, block recovery, and loop filtering) based on the block split structure.

[99] Splitting the picture into CTUs

[100] Pictures may be partitioned into a sequence of coding tree units (CTUs). The CTU may correspond to a coding tree block (CTB). Alternatively, the CTU may include a luminance sample coding tree block and two corresponding chrominance sample coding tree blocks. In other words, with respect to a picture including three sample arrays, the CTU may include an N×N luma sample block and two corresponding chrominance sample blocks.

[101] The maximum allowed CTU size for encoding and prediction may be different from the maximum allowed CTU size for transformation. For example, the maximum allowed luminance block size in a CTU may be 128x128.

[102] Partitioning a CTU Using a Tree Structure

[103] The CTU may be partitioned based on a quadtree (QT) structure. The quadtree structure can be referred to as a quaternary (quaternary) tree structure. This is meant to introduce various local characteristics. Meanwhile, herein, the CTU may be partitioned based on the division of a multiple type tree structure including a binary tree (BT) and a ternary tree (TT) in addition to a quadtree. Further, the QTBT structure may include a partition structure based on a quadtree and a binary tree. QTBTTT may include a partition structure based on a quadtree, a binary tree, and a ternary tree. Alternatively, the QTBT structure may include a partition structure based on a quadtree, a binary tree, and a ternary tree. In the coding tree structure, the CU may have a square shape or a rectangular shape. The CTU may be split as a quadtree structure. Thereafter, the leaf nodes of the quadtree structure may be further split by the multiple type tree structure.

[104] FIG. 3 is an embodiment in which expansion can be applied, and is a diagram illustrating an example of a multiple type tree structure.

[105] In an embodiment of the disclosure, the multiple type tree structure may include 4 partition types such as those illustrated in FIG. 3. The 4 split types may include SPLIT_BT_VER vertical binary split, SPLIT_BT_HOR horizontal binary split, SPLIT_TT_VER vertical ternary split, and SPLIT_TT_HOR horizontal ternary split. The leaf nodes of a multiple type tree structure may be referred to as CUs. Such CUs may be used for the prediction and transformation procedure. In this document, in general, a CU, PU, or TU may have the same block size. In this case, if the maximum supported transform length is less than the CU's color component width or height, then the CU and TU may have different block sizes.

[106] FIG. 4 is an embodiment in which expansion can be applied, and is a diagram illustrating a mechanism for signaling partitioning information of splitting a multiple type nested tree quadtree structure.

[107] In this case, the CTU is considered the root of the quadtree and is first split as a quadtree structure. Each leaf node of the quadtree can be further decomposed into a multi-type tree structure. In a multiple type tree structure, a first flag (eg, mtt_split_cu_flag) is signaled to indicate whether the corresponding node is further split. If the corresponding node is further split, a second flag (eg, mtt_split_cu_vertical_flag) may be signaled to indicate the split direction. Thereafter, a third flag (eg, mtt_split_cu_binary_flag) may be signaled to indicate whether the split type is binary split or ternary split. For example, the split tree mode of the multiple type (MttSplitMode) of the CU may be derived as Table 1 based on mtt_split_cu_vertical_flag and mtt_split_cu_binary_flag.

[108] Table 1

MttSplitMode mtt_split_cu_vertical_flag mtt_split_cu_binary_flag SPLIT_TT_HOR 0 0 SPLIT_BT_HOR 0 1 SPLIT_TT_VER 1 0 SPLIT_BT_VER 1 1

[109] FIG. 5 is an embodiment in which the disclosure may be applied, and is a diagram illustrating a method for dividing a CTU into a plurality of CUs based on a quadtree and nested multiple type tree structure.

[110] In this case, the edges of the blocks shown in thick lines indicate a quadtree split, and the remaining edges indicate a multiple-type tree split. Splitting a quadtree with a multiple type tree can provide a content-adapted coding tree structure. A CU may correspond to a coding block (CB). Alternatively, the CU may include a luminance sample coding block and two corresponding chrominance sample coding blocks. The CU size may be larger, such as CTU, or may be small, such as 4×4, per luminance sampling unit. For example, in the case of a 4:2:0 color format (or chrominance format), the maximum chroma CB size may be 64x64, and the minimum chroma CB size may be 2x2.

[111] Herein, for example, the maximum allowed luminance TB size may be 64x64, and the maximum allowed chrominance TB size may be 32x32. If the width or height of a CB split based on the tree structure is greater than the maximum width or height of the transformation, the corresponding CB may be split until the constraint on the size of the TB in the horizontal and vertical direction is automatically (or explicitly) satisfied.

[112] Meanwhile, for the multiple type tree quadtree coding tree scheme, the following parameters may be defined and identified as SPS syntax elements.

[113] - CTU size: quad tree root node size

[114] - MinQTSize: minimum allowed quad tree leaf node size

[115] - MaxBtSize: maximum allowed size of the root node of a binary tree

[116] - MaxTtSize: maximum allowed size of a ternary tree root node

[117] - MaxMttDepth: maximum allowed depth of split tree splitting hierarchy of multiple type from quadtree leaf

[118] - MinBtSize: minimum allowed size of a binary tree leaf node

[119] - MinTtSize: minimum allowed ternary tree leaf node size

[120] For example, in the case of a multi-type tree quadtree coding tree structure, the CTU size may be set to 128x128 luma samples and 64x64 blocks of two corresponding chroma samples (in a 4:2:0 chroma format). In this case, MinQTSize can be set to 16x16, MaxBtSize can be set to 128x128, MaxTtSzie can be set to 64x64, and MinBtSize and MinTtSize (for width and length) can be set to 4x4, and MaxMttDepth can be set to 4. Quadtree partitioning can be applied to the CTU to generate leaf quadtree nodes. A leaf node of a quadtree may be referred to as a QT leaf node. The leaf nodes of the quadtree can be 16x16 (ie, MinQTSize) up to 128x128 (ie, CTU size). If a QT leaf node is 128x128, it cannot further split into a binary tree/ternary tree. The reason for this is that although the QT leaf node is divisible, it exceeds MaxBtSize and MaxTtSize (ie, 64x64). In other cases, a QT leaf node may further be split into a multiple-type tree. Accordingly, a QT leaf node is the root node for a set type tree, and a QT leaf node may have a value of 0 depth of the set type tree (mttDepth). If the depth of the multiple type tree reaches MaxMttDepth (eg 4), the additional partition can no longer be considered. If the width of a multitype tree node is equal to MinBtSize and less than or equal to 2xMinTtSize, the additional horizontal split can no longer be considered. If the height of a multitype tree node is equal to MinBtSize and less than or equal to 2xMinTtSize, the additional vertical split can no longer be considered.

[121] FIG. 6 is an embodiment to which the expansion may be applied, and is a diagram illustrating a method for limiting the division of a binary tree.

[122] With reference to FIG. 6, in order to enable the 64×64 luma and 32×32 chrominance block pipeline scheme in a hardware decoder, the TT division may be limited in a special case. For example, if the width or height of the luminance coding block is larger than a preset specific value (eg, 32, 64), the TT division may be limited, as illustrated in FIG. 6.

[123] As used herein, the coding tree scheme may support that the luminance and chrominance block has a separate block tree structure. With respect to P- and B-slices, luma and chrominance CTBs in one CTU may be limited to have the same coding tree structure. However, with respect to I-cuts, the luma and chrominance blocks may have separate block tree structures. If a separate block tree mode is applied, the luminance CTB may be divided into CUs based on a specific coding tree structure. A chroma CTB may be divided into a chroma CU based on a different coding tree structure. This may mean that a CU in an I-slice may be configured with a luminance component coding block or two chroma component coding blocks, and a CU in a P- or B-slice may be configured with three chroma component blocks.

[124] In “CTU partitioning using a tree structure”, the coding tree structure of a multiple-type tree quadtree has been described, but the structure in which CUs are divided is not limited to this. For example, the BT structure and the TT structure may be interpreted as a concept included in the multiple partition tree (MPT) structure, and the CU may be interpreted as being shared by the QT structure and the MPT structure. In one example, in which a CU is split by a QT structure and an MPT structure, the split structure may be determined by signaling a syntax element (eg, MPT_split_type) including information indicating how many blocks the leaf node of the QT structure is divided into, and a syntax element (eg, , MPT_split_mode) including information indicating in which of the vertical and horizontal directions the leaf node of the QT structure is split.

[125] In another example, the CU may be partitioned using a method other than that of the QT structure, the BT structure, or the TT structure. That is, in contrast to the case where the deeper CU is split as 1/4 the size of the shallower CU based on the QT structure, or the deeper CU is split as 1/2 the size of the shallower CU based on the BT structure, or CU deeper depth is split as 1/4 or 1/2 size of shallower depth CU based on TT structure, deeper depth CU can be split as 1/5, 1/3, 3/8, 3/5, 2/3 or 5 /8 CU size shallower depth. The CU division method is not limited to this.

[126] If a portion of the tree node block exceeds the bottom or right boundary of the picture, the corresponding tree node block may be constrained such that samples of all encoded CUs are located within the picture boundaries. In this case, for example, the following separation rule may apply.

[127] - If part of the tree node box exceeds both the bottom and right borders of the picture,

[128] - If the block is a QT node, and the block size is greater than the minimum QT size, the block is forced to split in the QT division mode.

[129] - Otherwise, block is forced to split in SPLIT_BT_HOR mode

[130] - Otherwise, if the block part of the tree node exceeds the lower bounds of the picture,

[131] - If the block is a QT node, and the block size is greater than the minimum QT size, and the block size is greater than the maximum BT size, the block is forced to split in the QT division mode.

[132] - Otherwise, if the block is a QT node, and the block size is greater than the minimum QT size, and the block size is less than or equal to the maximum BT size, the block is forced to split in the QT split mode or SPLIT_BT_HOR mode.

[133] - Otherwise (the block is a BTT node, or the block size is less than or equal to the minimum QT size), the block is forced to split in SPLIT_BT_HOR mode.

[134] - Otherwise, if part of the tree node block exceeds the right borders of the picture,

[135] - If the block is a QT node, and the block size is greater than the minimum QT size, and the block size is greater than the maximum BT size, the block is forced to split in the QT split mode.

[136] - Otherwise, if the block is a QT node, and the block size is greater than the minimum QT size, and the block size is less than or equal to the maximum BT size, the block is forced to split in QT split mode or SPLIT_BT_VER mode.

[137] - Otherwise (the block is a BTT node, or the block size is less than or equal to the minimum QT size), the block is forced to split in SPLIT_BT_VER mode.

[138] Meanwhile, the coding block structure of a multiple tree quadtree coding block can provide a very flexible block splitting structure. Due to the split types supported by the set type tree, the same coding block structure result can potentially be obtained depending on different split patterns. The amount of split information data can be reduced by limiting the occurrence of such redundant split patterns. This is described with reference to the following drawing.

[139] FIG. 7 is an embodiment in which expansion can be applied and is a diagram illustrating redundant partitioning patterns that may appear in binary tree partitioning and ternary tree partitioning.

[140] As illustrated in FIG. 7, two levels of consecutive binary splits in the same direction have the same coding block structure as the binary split for center split after ternary split. In such a case, splitting the binary tree (in a given direction) for splitting at the split center of the ternary tree may be restricted. Such a restriction may apply to the CU of all pictures. If such a particular division is restricted, the signaling of the corresponding syntax elements may be modified to include the case of such restriction. Accordingly, the number of bits signaled for splitting can be reduced. For example, as in the example illustrated in FIG. 7, if the binary tree split for CU center split is limited, the mtt_split_cu_binary_flag syntax element indicating whether the split is binary split or ternary split is not signaled, and the corresponding value may be considered as 0 by the decoder.

[141]

[142] Divination

[143] To restore the current processing unit in which decoding is performed, the decoded portion of the current picture or other pictures including the current processing unit may be used.

[144] A picture (clipping) using only the current picture for reconstruction, that is, on which only intra-picture prediction is performed, may be called an intra-picture or an I-picture (clipping). A picture (clipping) using at most one motion vector and a reference index to predict each unit may be called a predictive picture or a P-picture (clipping). A picture (clipping) using a maximum of two motion vectors and reference indices can be called a bi-predictive picture or a B-picture (clipping).

[145] Intra-prediction means a prediction method by deriving the current processing block from a data element (eg, sample value) of the same decoded picture (or slice). That is, this method means a method for predicting a pixel value of the current processing block with reference to reconstructed areas in the current picture.

[146] Next, inter-prediction is described more specifically.

[147]

[148] Inter-prediction (or inter-prediction)

[149] Inter-prediction means a prediction method by deriving a current processing block based on a data element (eg, sample value or motion vector) of a picture other than the current picture. That is, this method means a method for predicting a pixel value of a current processing block with reference to reconstructed areas in a different reconstructed picture other than the current picture.

[150] Inter-prediction (or inter-picture prediction) is a technique for removing redundancy present between pictures, and is mainly performed by motion estimation and motion compensation.

[151] An embodiment of the disclosure describes the detailed technology of the inter-prediction method described in FIG. 1 and 2. In the case of a decoder, an embodiment can be described by the video/image decoding method based on inter-prediction according to FIG. 10 and an inter predictor in the decoder of FIG. 11. Further, in the case of an encoder, an embodiment can be described by the inter-prediction-based video/image coding method of FIG. 8 and an inter predictor in the encoder of FIG. 9. In addition, the data encoded according to FIG. 8 and 9 may be stored in the form of a bitstream.

[152] The encoder/decoder predictor may output a prediction sample by performing inter-prediction in a block unit. The inter-prediction may indicate a prediction derived using a method dependent on data elements (eg, sample values or motion information) of the picture(s) other than the current picture. If inter-prediction is applied to the current block, a predicted block (prediction sample array) for the current block may be derived based on a reference block (reference sample array) specified by the motion vector in the reference picture indicated by the reference picture index.

[153] In this case, in order to reduce the amount of motion information transmitted in the inter-prediction mode, the motion information of the current block may be predicted in a block, subblock, or sampling unit based on the correlation between the motion information of the neighboring block and the motion information of the current block. The motion information may include a motion vector and a reference picture index. The motion information may further include inter prediction type information (L0 prediction, L1 prediction, Bi prediction).

[154] If inter-prediction is applied, the neighboring block may include the spatial neighboring block represented in the current picture and the temporal neighboring block represented in the reference picture. The reference picture including the reference block and the reference picture including the temporal neighbor block may be the same or different. The temporary neighbor block may be referred to as a collocated reference block or collocated CU (colCU). A reference picture including a temporal adjacent block may be referred to as a colPic. For example, the motion information candidate list may be configured based on neighboring blocks of the current block. Flag or index information may be signaled indicating which candidate is selected (used) for deriving the motion vector and/or the reference picture index of the current block.

[155] Inter-prediction may be performed based on various prediction modes. For example, in the case of the skip mode and the combine mode, the motion information of the current block may be the same as the motion information of the selected neighboring block. In the case of the skip mode, unlike the combine mode, the residual signal may not be transmitted. In the case of a motion information prediction mode (motion vector prediction, MVP), a motion vector of a selected neighboring block may be used as a motion vector predictor, and a motion vector difference may be signaled. In this case, the motion vector of the current block may be output using the sum of the motion vector predictor and the motion vector difference.

[156] FIG. 8 and 9 are diagrams illustrating an inter prediction-based video/image encoding method according to an embodiment of the disclosure and an inter predictor in an encoding apparatus according to an embodiment of the disclosure.

[157] With reference to FIG. 8 and 9, S801 may be performed by the inter-predictor 180 of the encoder. S802 may be executed by the rest processor of the encoder. Specifically, S802 may be performed by the subtractor 115 of the encoder. In S803, prediction information may be output by inter predictor 180 and may be encoded by entropy encoder 190. In S803, residual information may be output by residual processor and may be encoded by entropy encoder 190. Residual information is information about residual samples. The residual information may include information about the quantized transform coefficients for the residual samples.

[158] As described above, the residual samples may be output as transform coefficients by the encoder converter 120. The transform coefficients may be output as quantized transform coefficients by a quantizer 130. Information about the quantized transform coefficients may be encoded in the entropy encoder 190 via a residual encoding procedure.

[159] The encoder performs inter prediction on the current block (S801). The encoder may output the inter-prediction mode and motion information of the current block, and may generate prediction samples of the current block. In this case, the inter-prediction mode determination procedure, the motion information output, and the prediction sample generation procedure may be performed at the same time, and any one procedure may be performed before the other procedure. For example, the encoder inter-predictor 180 may include a prediction mode determination unit 181, a motion information output unit 182, and a prediction sample output unit 183. The prediction mode determination unit 181 may determine the prediction mode for the current block. The motion information output unit 182 may output motion information of the current block. The prediction sample output unit 183 may output motion samples of the current block.

[160] For example, the encoder inter-predictor 180 may search a given area (search area) of reference pictures for a block similar to the current block through motion estimation, and may output a reference block having a minimum difference, or a given support difference or less relative to the current block. The inter predictor 180 may output a reference picture index indicating a reference picture in which the reference block is located based on the reference block, and may derive a motion vector based on a position difference between the reference block and the current block. The encoder may determine the mode applied to the current block among various prediction modes. The encoder may compare the RD costs for different prediction modes and may determine the optimal prediction mode for the current block.

[161] For example, if a skip mode or a merge mode is applied to the current block, the encoder may configure a merge candidate list, which will be described later, and may output a reference block having a minimum difference or difference of a given reference or less relative to the current block, among the reference blocs indicated by the merger candidates included in the list of merger candidates. In this case, a merge candidate associated with the derived reference block may be selected. Merge index information indicating a selected merge candidate may be generated and signaled to a decoder. The motion information of the current block may be output using the motion information of the selected merge candidate.

[162] As another example, if the (A)MVP mode is applied to the current block, the encoder may configure an (A)MVP candidate list, which will be described later, and may use a motion vector of a motion vector predictor (mvp) candidate selected among mvp candidates included in the list of (A)MVP candidates as the mvp of the current block. In this case, for example, a motion vector indicating a reference block derived by the motion estimation may be used as the motion vector of the current block. The mvp candidate including the motion vector having the smallest difference with respect to the motion vector of the current block among the mvp candidates may become the selected mvp candidate. The motion vector difference (MVD), that is, the difference obtained by subtracting mvp from the motion vector of the current block, can be output. In this case, the MVD information may be signaled to the decoder. Moreover, if the (A)MVP mode is applied, the reference picture index value may be configured as reference picture index information and may be separately signaled to the decoding apparatus.

[163] The encoder may output residual samples based on the prediction samples (S802). The encoder may derive the residual samples by comparing between the original samples of the current block and the prediction samples.

[164] The encoding device encodes image information including prediction information and residual information (S803). The encoding device may output the encoded image information in the form of a bitstream. The prediction information may include information about prediction mode information (eg, a skip flag, a union flag, or a mode index) and motion information as information related to a prediction procedure. The information associated with the motion information may include candidate selection information (eg, union index, mvp flag, or mvp index), that is, information for deriving a motion vector. Moreover, the information associated with the motion information may include MVD information and/or reference picture index information.

[165] Moreover, the information associated with the motion information may include information indicating whether L0 prediction, L1 prediction, or bidirectional (bi) prediction is applied. The residual information is information about the residual samples. The residual information may include information about the quantized transform coefficients for the residual samples.

[166] The outputted bitstream may be stored in a (digital) storage medium and transmitted to the decoder, or may be transmitted to the decoder over a network.

[167] Meanwhile, as described above, the encoding device can generate a reconstructed picture (including reconstructed samples and a reconstructed block) based on the reference samples and the residual samples. This is done to output, at the encoder, the same prediction results as those performed at the decoder. Accordingly, the encoding efficiency can be improved. Accordingly, the encoder may store the reconstructed picture (or the reconstructed samples and the reconstructed block) in memory and may use the reconstructed picture as a reference picture for inter-prediction. As described above, the loop filtering procedure may be further applied to the reconstructed picture.

[168] FIG. 10 and 11 are diagrams illustrating an inter-prediction-based video/image decoding method according to an embodiment of the disclosure, and an inter-predictor in a decoding apparatus according to an embodiment of the disclosure.

[169] With reference to FIG. 10 and 11, the decoding device can perform an operation corresponding to the operation performed in the encoding device. The decoding apparatus may perform prediction on the current block based on the received prediction information, and may output prediction samples.

[170] S1001-S1003 may be performed by the inter-predictor 260 of the decoder. The residual information in S1004 can be obtained from the bit stream by the entropy decoder 210 of the decoding apparatus. The decoder residual processor may output residual samples for the current block based on the residual information. Specifically, the residual processor dequantizer 220 may derive transform coefficients by performing dequantization based on the quantized transform coefficients derived from the residual information. Residual processor detransformer 230 may derive residual samples for the current block by performing an inverse transform on the transform coefficients. S1005 may be performed by the adder 235 or by the recovery unit of the decoder.

[171] Specifically, the decoding apparatus may determine the prediction mode for the current block based on the received prediction information (S1001). The decoding apparatus may determine which inter prediction mode is applied to the current block based on the prediction mode information in the prediction information.

[172] For example, the decoding device may determine whether the combining mode or the (A)MVP mode is applied to the current block based on the combining flag. Alternatively, the decoder may select one of various inter prediction mode candidates based on the mode index. The inter-prediction mode candidates may include a skip mode, a pooling mode, and/or an (A)MVP mode, or may include various inter-prediction modes, which will be described later.

[173] The decoding apparatus outputs motion information of the current block based on the determined inter prediction mode (S1002). For example, if a skip mode or a merge mode is applied to the current block, the decoding device may configure a merge candidate list to be described later and select one of the merge candidates included in the merge candidate list. The selection may be performed based on the selection information (combination index). The motion information of the current block may be output based on the motion information of the selected merge candidate. The movement information of the selected merging candidate may be used as the movement information of the current block.

[174] As another example, if the (A)MVP mode is applied to the current block, the decoding device may configure an (A)MVP candidate list, which will be described later, and may use a motion vector of a motion vector predictor (mvp) candidate selected among mvp candidates included in the list of (A)MVP candidates as the mvp of the current block. The selection may be made based on selection information (mvp flag or mvp index). In this case, the decoding apparatus may derive the MVD of the current block based on the MVD information. The decoder may derive a motion vector of the current block based on the mvp of the current block and the MVD. Further, the decoding apparatus may derive a reference picture index of the current block based on the reference picture index information. The picture indicated by the index of the reference picture in the reference picture list relative to the current block may be output as a reference picture as a reference for inter-prediction of the current block.

[175] Meanwhile, as will be described later, the motion information of the current block can be output without configuring the candidate list. In this case, motion information of the current block may be output according to the procedure disclosed in the prediction mode to be described later. In this case, the candidate list configuration as described above may be omitted.

[176] The decoding apparatus may generate prediction samples for the current block based on the motion information of the current block (S1003). In this case, the decoding apparatus may output a reference picture based on the reference picture index of the current block, and may output prediction samples of the current block indicated in the reference picture by a motion vector of the current block. In this case, as will be described later, the prediction sample filtering procedure may be further performed on some or all of the prediction samples of the current block, as appropriate.

[177] For example, the inter-predictor 260 of the decoding apparatus may include a prediction mode determination unit 261, a motion information output unit 262, and a prediction sample output unit 263. The decoding apparatus may determine the prediction mode of the current block based on the prediction mode information received from the prediction mode determination unit 261, may output motion information (motion vector and/or reference picture index) of the current block based on information related to the motion information received from the unit 262 motion information output. The prediction sample output module 263 may output the prediction samples of the current block.

[178] The decoder generates residual samples for the current block based on the received residual information (S1004). The decoding apparatus may generate reconstructed samples for the current block based on the prediction samples and residual samples, and may generate a reconstructed picture based on the reconstructed samples (S1005). Thereafter, as described above, the loop filtering procedure may be further applied to the reconstructed picture.

[179] As described above, the inter prediction procedure may include an inter prediction mode determination step, a motion information output step according to the determined prediction mode, and a prediction execution step (prediction sample generation) based on the derived motion information.

[180]

[181] Inter prediction mode determination

[182] Various inter-prediction modes may be used to predict the current block in the picture. For example, different modes may be used, such as pool mode, skip mode, MVP mode, and affine mode. Decoder side motion vector refinement mode (DMVR), adaptive motion vector resolution (AMVR) mode, etc. can additionally be used as additional modes. The affine mode may be referred to as the affine motion prediction mode. The MVP mode may be referred to as the Advanced Motion Vector Prediction (AMVP) mode.

[183] Prediction mode information indicating the inter-prediction mode of the current block may be signaled from the encoding device to the decoding device. The prediction mode information may be included in the bitstream and received by the decoder. The prediction mode information may include index information indicating one of the plurality of candidate modes. Alternatively, the inter-prediction mode may be indicated via hierarchical flag information signaling. In this case, the prediction mode information may include one or more flags.

[184] For example, a flag may be further signaled to indicate whether a skip mode is applied by signaling a skip flag, to indicate whether a pooling mode is applied by signaling a pooling flag if the skip mode is not applied, and to indicate that an MVP mode is applied. , if the merge mode is not applied, or for additional identification. The affine mode may be signaled as an independent mode, or may be signaled as a mode dependent on the union mode or the MVP mode. For example, the affine mode may be configured as one of a union candidate list or an MVP candidate list, as will be described later.

[185]

[186] Derivation of motion information according to the inter-prediction mode

[187] Inter-prediction may be performed using motion information of the current block. The encoding device may output the optimal motion information for the current block in accordance with the motion estimation procedure. For example, the encoder may search for a reference block having a similar correlation using an original block in the original picture for the current block in a fractional pixel unit within a certain search range in the reference picture. Accordingly, the encoding device can output motion information. Block similarity may be inferred based on the difference between phase-based sample values. For example, block similarity may be calculated based on the SAD between the current block (or current block template) and a reference block (or reference block template). In this case, motion information may be derived based on the reference block with the lowest SAD in the search area. The derived motion information may be signaled to a decoder using several methods based on the inter-prediction mode.

[188]

[189] Combine Mode and Skip Mode

[190] FIG. 12 is an embodiment to which the disclosure is applied, and is a diagram for describing adjacent blocks used in the combine mode or the skip mode.

[191] If the combining mode is applied, the motion information of the current prediction block is not directly transmitted, and the motion information of the current prediction block is output using the motion information of the adjacent prediction block. Accordingly, the motion information of the current prediction block may be indicated by transmitting flag information to notify that the combining mode has been used and a combining index to notify which adjacent prediction block has been used.

[192] In an embodiment of the disclosure,

[193] In an embodiment of the disclosure, the encoder may search for a combining candidate block used to output motion information of the current prediction block to perform the combining mode. For example, up to a maximum of 5 merge candidate blocks may be used, but expansion is not limited to this. Moreover, the maximum number of merge candidate blocks can be transmitted in the slice header (or tile group header), and the disclosure is not limited thereto. After searching for merge candidate blocks, the encoder may generate a list of merge candidates and may select a merge candidate block having the least cost among merge candidate blocks as a final merge candidate block.

[194] An embodiment of the disclosure provides various embodiments of merge candidate blocks forming a merge candidate list.

[195] The merge candidate list may use, for example, 5 merge candidate blocks. For example, 4 spatial pooling candidates and 1 temporal pooling candidate may be used. For a detailed example, in the case of a spatial merging candidate, the blocks illustrated in FIG. 12 can be used as spatial pooling candidates.

[196] FIG. 13 is a flowchart illustrating a merge candidate list generation method according to an embodiment in which disclosure is applied.

[197] With reference to FIG. 13, an encoder (encoder/decoder) searches for spatial neighboring blocks of the current block, and inserts the derived spatial merging candidates into a merging candidate list (S1301). For example, spatial neighbors may include a bottom left corner neighbor, a left neighbor, a top right corner neighbor, a top neighbor, and a top left corner neighbor of the current block. In this case, this is an example, and additional neighbor blocks such as a right neighbor block, a lower neighbor block, and a lower right neighbor block in addition to the spatial neighbor blocks can be further used as spatial neighbor blocks. The encoder may detect available blocks by searching for spatial neighboring blocks based on priority, and may output motion information of the detected blocks as spatial merging candidates. For example, the encoder and decoder may search for the 5 blocks illustrated in FIG. 12 in sequence of A1, B1, B0, A0, and B2, and can configure a merge candidate list by sequentially indexing the available candidates.

[198] The encoding device searches for a temporary neighbor block of the current block, and inserts the deduced temporary merge candidate into the merge candidate list (S1302). The temporary neighbor block may be located in a reference picture, that is, a picture different from the current picture in which the current block is located. The reference picture in which the temporal adjacent block is located may be called a col-picture or a col-picture. The search for a temporary neighboring block can be carried out in the sequence: the lower right corner neighboring block and the lower right central block of the combined block for the current block in the col-picture.

[199] Meanwhile, if motion data compression is applied, specific motion information may be stored in the col picture as representative motion information for each given storage unit. In this case, it is not necessary to store motion information for all blocks in a given storage unit, and thus the motion data compression effect can be obtained. In this case, a given storage unit may be predefined as a 16x16 sampling unit or an 8x8 sampling unit, for example, or size information for a given storage unit may be signaled from an encoder to a decoder. If motion data compression is applied, the motion information of the temporal neighbor block may be replaced with representative motion information of the storage unit in which the temporal neighbor block resides.

[200] That is, in this case, in one aspect of the implementation, after an arithmetic shift to the right by a given value is performed based on the coordinates (upper left sample position) of the temporal neighbor block, and not the prediction block in which the coordinates of the temporal neighbor block are located , a temporal join candidate may be derived based on motion information of a prediction block that spans an arithmetically left-shifted location. For example, if this storage unit is a 2n×2n sampling unit, assuming that the coordinates of the temporal neighbor block are (xTnb, yTnb), motion information of the prediction block located at ((xTnb>>n)<<n), (yTnb> >n)<<n)), i.e., the modified location, can be used for the temporary merge candidate.

[201] Specifically, for example, if the storage unit is a 16×16 sampling unit, assuming that the coordinates of the temporal neighbor block are (xTnb, yTnb), motion information of the prediction block located at ((xTnb>>4)<<4 ), (yTnb>>4)<<4)), i.e., the modified location, can be used for the temporary merge candidate. Alternatively, for example, if the given storage unit is an 8×8 sampling unit, assuming that the coordinates of the temporal neighbor block are (xTnb, yTnb), motion information of the prediction block located at ((xTnb>>3)<<3), ( yTnb>>3)<<3)), that is, the modified location, can be used for the temporary merge candidate.

[202] The encoder may check whether the current number of merge candidates is less than the maximum number of merge candidates (S1303). The maximum number of merge candidates may be predefined or may be signaled from encoder to decoder. For example, the encoder may generate information about the maximum number of merge candidates, may encode the information, and may transmit the information to the decoder in the form of a bitstream. If the maximum number of merge candidates is full, the process of adding candidates cannot be performed.

[203] If, as a result of checking, the current number of merge candidates is less than the maximum number of merge candidates, the encoder inserts the added merge candidate into the merge candidate list (S1304). The added merge candidate may be, for example, ATMVP, a combined bi-predictive merge candidate (if the slice type for the current slice is B-type), and/or a null vector merge candidate.

[204] If, as a result of checking, the current number of merge candidates is not less than the maximum number of merge candidates, the encoding device may complete the configuration of the merge candidate list. In this case, the encoder may select the optimal merging candidate among the merging candidates forming the merging candidate list based on the cost of the distortion-rate (RD) criterion, and may signal to the decoder selection information (e.g., merging index) indicating the selected merging candidate. . The decoder may select the optimal merge candidate based on the merge candidate list and the selection information.

[205] The movement information of the selected merging candidate may be used as the movement information of the current block. As described above, the prediction samples of the current block may be derived based on motion information of the current block. The encoder may derive the residual samples of the current block based on the prediction samples and may signal, to the decoder, residual information associated with the residual samples. The decoder may generate recovered samples based on the residual samples derived from the residual information and the prediction samples. As described above, the decoder may generate a reconstructed picture based on the reconstructed samples.

[206] If the skip mode is applied, motion information of the current block may be output using the same method as in the case where the join mode is applied. In this case, if the skip mode is applied, the residual signal for the corresponding block is omitted. Accordingly, the prediction samples can be directly used as recovered samples.

[207]

[208] MVP Mode

[209] FIG. 14 is a flowchart illustrating a merge candidate list generation method according to an embodiment in which disclosure is applied.

[210] If a motion vector prediction (MVP) mode is applied, a motion vector predictor (mvp) candidate list may be generated based on the motion vector of the reconstructed spatial neighbor block (e.g., the neighbor block described in FIG. 12) and/or the motion vector corresponding to temporary adjacent block (or Col block). That is, the motion vector of the reconstructed spatial neighbor block and/or the motion vector of the temporal neighbor block may be used as a motion vector predictor candidate.

[211] The prediction information may include selection information (eg, an MVP flag or an MVP index) indicating an optimal motion vector predictor candidate selected among the motion vector predictor candidates included in the list. In this case, the predictor may select the motion vector predictor of the current block from among the motion vector predictor candidates included in the motion vector candidate list using the selection information. The encoder predictor may calculate a motion vector difference (MVD) between the motion vector of the current block and the motion vector predictor, may encode the MVD, and may output the encoded MVD in the form of a bitstream. That is, MVD may be calculated as a value obtained by subtracting the motion vector predictor from the motion vector of the current block. In this case, the predictor of the decoding apparatus may obtain the motion vector difference included in the prediction information, and may output the motion vector of the current block by summing the motion vector difference and the motion vector predictor. The predictor of the decoding device may obtain or derive a reference picture index indicating the reference picture from the prediction information. For example, the motion vector predictor candidate list may be configured as illustrated in FIG. fourteen.

[212]

[213] Affine Motion Prediction

[214] FIG. 15 illustrates an example of motion patterns in accordance with an embodiment of the disclosure.

[215] In conventional image compression technology (eg, High Efficiency Video Coding (HEVC)), one motion vector is used to represent the motion of a coding block. Although the optimal motion of a block unit can be represented using a method using one motion vector per block, this may not be the actual optimal motion of each picture element. Accordingly, if an optimal motion vector is determined in a picture element unit, encoding efficiency can be improved. Accordingly, an embodiment of the disclosure describes a motion prediction method for encoding or decoding a video signal using a multiple motion model. Specifically, a motion vector may be represented in each block picture element unit or sub-block unit using the motion vector at 2 to 4 breakpoints. A prediction scheme using a motion vector of a plurality of control points may be referred to as affine motion prediction or affine prediction.

[216] An affine motion model according to an embodiment of the disclosure may represent 4 motion models such as those illustrated in FIG. 15. An affine motion model for representing three motions (translation, scaling and rotation) among the motions that can be represented by an affine motion model is called a (simplified) affine motion similarity model. In describing embodiments of the disclosure, the (simplified) affine motion similarity model is generally described for convenience of description, but the disclosure is not limited thereto.

[217]

[218] FIG. 16 illustrates an example of a reference point motion vector for affine motion prediction in accordance with an embodiment of the disclosure.

[219] As in FIG. 16, in affine motion prediction, a motion vector of a location (or sub-block) of a picture element included in a block may be determined using a pair v_0 and v_1 of two control point motion vectors (CPMV). In this case, the set of motion vectors may be referred to as an affine motion vector field (MVF). In this case, the affine field of motion vectors can be determined using Equation 1.

[220] Equation 1

[221] In Equation 1, v_0(v_0={v_0x, v_0y}) indicates the motion vector CPMV0 at the first waypoint of the upper left location of the current block 1300. v_1(v_1={v_1x, v_1y}) indicates the motion vector CPMV1 at the second waypoint the top right location of the current block 1300. In addition, w indicates the width of the current block 1300. v(v={v_x, v_y}) indicates the motion vector at location {x, y}. The motion vector of a unit of a subblock (or picture element) may be output using Equation 1. In one embodiment, the motion vector precision may be rounded as a precision of 1/16.

[222]

[223] FIG. 17 illustrates an example of a motion vector for each sub-block of a block to which affine motion prediction has been applied in accordance with an embodiment of the disclosure.

[224] With reference to FIG. 17, during encoding or decoding, an affine motion vector field (MVF) may be defined in a picture element unit or a block unit. That is, in affine motion prediction, the motion vector of the current block may be output in a picture element unit or a sub-block unit.

[225] If the affine field of motion vectors is defined in the picture element unit, the motion vector can be obtained based on the value of each picture element. In the case of a unit block, the motion vector of the corresponding block can be obtained based on the value of the center picture element of the block. Herein, it is assumed that the motion vector affine field (MVF) is defined in a 4*4 block unit as in FIG. 17. In this case, it is for the convenience of description, and the embodiment of the disclosure is not limited to this. Fig. 17 illustrates an example of a case where a coding block is configured with 16*16 samples and an affine motion vector field (MVF) is defined in a 4*4 block unit.

[226] Affine motion prediction may include affine merging (or AF_MERGE) mode and affine inter mode (or AF_INTER). AF_INTER mode may include AF_4_INTER mode using a 4-parameter motion model and motion models based on 6 parameters.

[227]

[228] Affine Union Mode

[229] AF_MERGE determines a control point motion vector (CPMV) based on a neighboring block affine motion model encoded as affine motion prediction. The adjacent block, affine-coded in the search sequence, can be used for AF_MERGE. When one or more adjacent blocks are encoded as affine motion prediction, the current block may be encoded as AF_MERGE.

[230] That is, if the affine association mode is applied, the CPMVs of the current block may be output using the CPMVs of the neighboring block. In this case, the CPMVs of the neighboring block may be used as the CPMVs of the current block without any change, and the CPMVs of the neighboring block may be modified based on the size of the neighboring block and the size of the current block, and may be used as the CPMVs of the current block.

[231] FIG. 18 illustrates an example of adjacent blocks used for affine motion prediction in affine pooling mode, in accordance with an embodiment of the disclosure.

[232] In the affine merging (AF_MERGE) mode, the encoder may perform encoding as in the following process.

[233] Step-1: Scan adjacent blocks A-E 1810, 1820, 1830, 1840, and 1850 of the current alphabetical coding block 1800, and determine the block encoded first according to the affine prediction mode based on the scan sequence as an affine union candidate block (AF_MERGE)

[234] Step-2: Determine an affine motion model using the control point motion vector (CPMV) of the determined candidate block

[235] Step-3: Determine the checkpoint motion vector (CPMV) of the current block 1800 based on the affine motion model of the candidate block, and determine the MVF of the current block 1800

[236]

[237] FIG. 19 illustrates an example of a block on which affine motion prediction is performed using neighboring blocks to which affine motion prediction has been applied in accordance with an embodiment of the disclosure.

[238] For example, as in FIG. 19, if block A 1920 has been encoded according to the affine mode, after block A 1920 is determined to be a candidate block, an affine motion model may be derived using control point motion vectors (CPMV) (e.g., v2 and v3) of block A 1920 , and control point motion vectors (CPMV) v0 and v1 of the current block 1900 may be determined. An affine motion vector field (MVF) of the current block 1900 may be determined based on the control point motion vectors (CPMV) of the current block 1900, and encoding may be performed.

[239]

[240] FIG. 20 is a diagram for describing a method for generating a merge candidate list using a neighboring affine coding block according to an embodiment of the disclosure.

[241] With reference to FIG. 20, if a CPMV pair is defined using an affine pooling candidate, then candidates such as those illustrated in FIG. 20. In FIG. 20, it is assumed that the candidate list scanning sequence has been set to A, B, C, D, and E. In this case, the disclosure is not limited to this, and various sequences can be preset.

[242] In an embodiment, if the number of candidates (hereinafter referred to as affine candidates) encoded according to the affine mode (or affine prediction) available in neighboring blocks (i.e., A, B, C, D, E) , is 0, then the affine union mode of the current block can be skipped. If the number of available affine candidates is one (eg, A), then the motion model of the corresponding candidate may be used to derive the motion vectors of the control points (CPMV_0 and CPMV_1) of the current block. In this case, an index indicating the corresponding candidate may not be required (or encoded). If the number of available affine candidates is two or more, then two candidates in the scan sequence may be configured as a candidate list for AF_MERGE. In this case, candidate selection information such as an index indicating the candidate selected in the candidate list may be signaled. The selection information may be flag or index information and may be called AF_MERGE_flag or AF_merge_idx.

[243] In an embodiment of the disclosure, motion compensation for the current block may be performed based on the size of the subblock. In this case, the subblock size of the affine block (ie, the current block) is output. If each of the sub-block width and height is greater than 4 luma samples, a motion vector for each sub-block may be output, and motion compensation based on DCT-IF (1/16 pel for luma and 1/32 for chroma) may be performed on the sub-block. If not, motion compensation based on the extended bilinear interpolation filter can be performed on all affine blocks.

[244] In an embodiment of the disclosure, if the merge/skip flag is true and both the width and height of the CU are greater than or equal to 8, at the CU level, an affine flag is signaled via the bitstream indicating whether the affine merge mode is used. . When the CU is encoded as AF_MERGE, a merge candidate index having a maximum value of '5' is signaled to indicate that a motion information candidate is used for the CU in the affine merge candidate list.

[245]

[246] FIG. 21 and 22 are diagrams for describing a method for generating a merge affine candidate list using an affine prediction encoded neighbor block according to an embodiment of the disclosure.

[247] With reference to FIG. 21, the pool affine candidate list is configured through the following steps.

[248] 1) Inserting a model-based affine candidate

[249] A model-based affine candidate means that the candidate is derived from a real neighboring reconstructed block encoded according to the affine mode. As illustrated in FIG. 21, the scan sequence for a candidate block from left A, top B, top right C, and bottom left D to top left E.

[250] If the neighboring lower left block A is encoded in affine mode with 6 parameters, then motion vectors (v_4, v_5, v_6) of the upper left corner, upper right corner and lower left corner of the CU including block A are obtained. The motion vectors ( v_0, v_1, v_2) of the upper left corner of the current block are calculated based on the motion vectors (v_4, v_5, and v_6) according to a 6-parameter affine model.

[251] If the neighboring lower left block A is encoded in 4-parameter affine mode, then the motion vectors (v_4, v_5) of the upper left corner and the upper right corner of the CU including block A are obtained. The motion vectors (v_0, v_1) of the upper left the angles on the current block are calculated based on the motion vectors (v_4, v_5) in accordance with the 4-parameter affine model.

[252] 2) Insertion of checkpoint-based affine candidates

[253] With reference to FIG. 21, a breakpoint-based candidate means that a candidate is configured by combining adjacent breakpoint movement information.

[254] The checkpoint movement information is first derived from the assigned spatial neighbor blocks and the temporal neighbor block illustrated in FIG. 21. CP_k (k=1, 2, 3, 4) indicates the k-th control point. Moreover, A, B, C, D, E, F, and G are the spatial location for CP_k (k=1, 2, 3) prediction, and H is the temporal location for CP4 prediction.

[255] The coordinates of CP_1, CP_2, CP_3 and CP_4 are (0, 0), (W, 0), (H, 0) and (W, H), respectively. In this case, W and H are the width and height of the current block.

[256] The motion information of each checkpoint is obtained based on the next priority.

[257] With respect to CP_1, the priority check is A→B→C, and A is used if A is available. If not, and if B is available, then B is used. If both A and B are not available, then C is used. If all three candidates are not available, motion information CP_1 cannot be obtained.

[258] With respect to CP_2, the priority check is E→D.

[259] With respect to CP_3, the priority check is G→F.

[260] For CP_4, H is used.

[261] Second, combinations of control points are used to configure the movement model.

[262] The motion vectors of the two control points are needed to calculate the transformation parameters in the 4-parameter affine model. The two control points can be selected from one of the following 6 combinations {CP_1, CP_4}, {CP_2, CP_3}, {CP_1, CP_2}, {CP_2, CP_4}, {CP_1, CP_3} and {CP_3, CP_4}. For example, the use of control points CP_1 and CP_2 in the formation of a 4-parameter affine motion model is referred to as “affine (CP_1, CP_2)”.

[263] The motion vectors of the three control points are needed to calculate the transformation parameters in the 6-parameter affine model. The three control points can be selected from one of the following 4 combinations {CP_1, CP_2, CP_4}, {CP_1, CP_2, CP_3}, {CP_2, CP_3, CP_4} and {CP_1, CP_3, CP_4}. For example, the use of control points CP_1, CP_2, and CPv3 in generating a 6-parameter affine motion model is referred to as “affine (CP_1, CP_2, CP_3)”.

[264] In addition, in an embodiment of the disclosure, if an affine join candidate is presented in affine join mode, this can always be considered as a 6-parameter affine mode.

[265]

[266] Affine Inter Mode

[267] FIG. 23 illustrates an example of neighboring blocks used for affine motion prediction in the affine inter-mode, in accordance with an embodiment of the disclosure.

[268] With reference to FIG. 23, affine motion prediction may include an affine merge mode (or AF_MERGE) and an affine inter mode (or AF_INTER). In the affine inter-mode (AF_INTER), after cue-point motion vector prediction (CPMVP) and CPMV 2 are determined, a cue-point motion vector difference (CPMVD) corresponding to the difference can be transmitted from an encoder to a decoder. The detailed affine inter-mode (AF_INTER) encoding process may be as described below.

[269] Step-1: Determine two CPMVP pair candidate

[270] Step-1.1: Determine a maximum of 12 CPMVP candidate combinations (see Equation 2)

[271] Equation 2

[272] In Equation 2, v_0 indicates the motion vector CPMV0 at the top left breakpoint 2310 of the current block 2300. v_1 indicates the motion vector CPMV1 at the top right breakpoint 2311 of the current block 2300. v_2 indicates the motion vector CPMV2 at the bottom left checkpoint 2312 of the current block 2300. v_A indicates the motion vector of the neighboring block A 2320 adjacent to the top left of the upper left breakpoint 2310 of the current block 2300. v_B indicates the motion vector of the neighboring block B 2322 adjacent to the top left breakpoint 2310 of the current block 2300. v_C indicates the motion vector of the neighboring block C 2324 adjacent to the left of the top left breakpoint 2310 of the current block 2300. v_D is the motion vector of the neighboring block D 2326 adjacent to the top right breakpoint 2311 of the current block 2300. v_E indicates the motion vector of the neighboring block E 2328 adjacent to the top right of upper right breakpoint 2311 of the current block 2300. v_F indicates the age the motion torus of the adjacent block F 2330 adjacent to the left of the bottom left breakpoint 2312 of the current block 2300. v_G indicates the motion vector of the neighboring block G 2332 adjacent to the left of the bottom left breakpoint 2312 of the current block 2300.

[273] Step-1.2: Sort combinations of CPMVP candidates based on a value having a small difference value (DV) and use the top two candidates (see Equation 3 below)

[274] Equation 3

[275] v_0x indicates the x-axis element of the motion vector (V0 or CPMV0) at the upper left breakpoint 2310 of the current block 2300. v_1x indicates the x-axis element of the motion vector (V1 or CPMV1) at the top right breakpoint 2311 of the current block 2300. v_2x indicates the element x axis of the motion vector (V_2 or CPMV_2) at the lower left control point 2312 of the current block 2300. v_0y indicates the y axis element of the motion vector (V_0 or CPMV_0) at the upper left control point 2310 of the current block 2300. v_1y indicates the y axis element of the motion vector (V_1 or CPMV_1) at the top right breakpoint 2311 of the current block 2300. v_2y indicates the motion vector y-axis element (V_2 or CPMV_2) at the bottom left breakpoint 2312 of the current block 2300. w indicates the width of the current block 2300. h indicates the height of the current block 2300.

[276] Step-2: Use AMVP candidate list when checkpoint motion vector predictor pair (CPMVP) candidate is less than 2

[277] Step-3: Determine the control point motion vector predictor (CPMVP) of each of the two candidates, and optimally select the candidate having the smaller value by comparing the cost of RD and CPMV

[278] Step-4: Transmit index corresponding to optimal candidate and checkpoint motion vector difference (CPMVD)

[279] In an embodiment of the disclosure, a CPMVP candidate generation process in AF_INTER is provided. Identical to AMVP, the number of candidates is 2, and an index indicating the location of the candidate list is signaled.

[280] The CPMVP candidate list generation process is as follows.

[281] 1) It is checked whether neighboring blocks are encoded as affine motion prediction by scanning neighboring blocks. If scanned blocks are encoded as affine prediction, a pair of motion vectors of the current block is derived from the affine motion model of the scanned adjacent block until the number of candidates becomes 2.

[282] 2) If the number of candidates is less than 2, a candidate configuration process is performed. Moreover, in an embodiment of the disclosure, a 4-parameter affine inter-mode (2 control points) is used to predict the approach/departure and rotation movement pattern and content. As illustrated in FIG. 1, the block affine motion field is described by two control point motion vectors.

[283] The motion vector field (MVF) of the block is described by Equation 1 described above.

[284] In conventional technology, an advanced motion vector prediction (AMVP) mode is needed to scan the motion vector prediction index (MVP) and motion vector difference (MVD). When the AMVP mode is applied in a discovery, an affine flag (affine_flag) is signaled to indicate whether affine prediction is used. If affine prediction is applied, then the syntaxes inter_dir, ref_idx, mvp_index and two MVDs (mvd_x and mvd_y) are signaled. A candidate list of a pair of affine MVPs is generated, which includes two pairs of affine MVPs. The signaled mvp_index is used to select one of two pairs of affine MVPs. A pair of affine MVPs is generated by two types of affine MVP candidates. One is a spatially inherited affine candidate and the other is an angle-derived affine candidate. If neighboring CUs are encoded in affine mode, then spatial legacy affine candidates may be generated. The affine motion model of a neighboring affine-coded block is used to generate the motion vectors of the MVP pair 2 control points. The MVs of the MVP 2 control point pair of the spatial legacy affine candidate are derived using the following equations.

[285] Equation 4

V _0x =V _B0x +(VB _{2_x} -V _B0x )*(posCurCU_Y-posRefCU_Y)/RefCU_height+ (V _B1x -V _B0x )*(posCurCU_X-posRefCU_X)/RefCU_width

[286] Equation 5

V _0y =V _B0y +(V _{B2_y} -V _B0y )*(posCurCU_Y-posRefCU_Y)/RefCU_height+ (V _B1y -V _B0y )*(posCurCU_X-posRefCU_X)/RefCU_width

[287] If V_B0, V_B1 and V_B2 can be replaced by the top left MV, top right MV and bottom left MV of a given reference/neighbor CU, (posCurCU_X, posCurCU_Y) is the top left sample location of the current CU for the top left frame sample. (posRefCU_X, posRefCU_Y) is the top left sample location of the reference/neighbor CU for the top left sample of the frame.

[288] Equation 6

V _{1 x} =V _{B 0 x} +(V _{B 1 x} -V _{B 0 x} )*CU_width/RefCU_width

[289] Equation 7

V _{1 y} =VB _{0 y} +(V _{B 1 y} -V _{B 0 y} )*CU_width/RefCU_width

[290] FIG. 24 illustrates an example of neighboring blocks used for affine motion prediction in the affine inter-mode, in accordance with an embodiment of the disclosure.

[291] With reference to FIG. 24, when the number of MVP pairs is less than 2, the corner-derived affine candidate is used. As illustrated in FIG. 24, adjacent motion vectors are used to derive a pair of affine MVPs. With respect to the first corner-derived affine candidate, the first available MV in set A (A0, A1 and A2) and the first available MV in set B (B0 and B1) are used to configure the first pair of MVPs. With respect to the second corner-derived affine candidate, the first available MV in set A and the first available MV in set C (C0 and C1) are used to calculate the MV of the upper right control point. The first available MV in set A and the computed MV of the top right checkpoint are the second pair of MVPs.

[292] In an embodiment of the disclosure, two sets of candidates including two (three) candidates {mv_0, mv_1} ({mv_0, mv_1, mv_2) are used to predict two (three) control points of the affine motion model. Difference data (mvd_0, mvd_1, mvd_2) of motion vectors and control points are calculated using the following equations.

[293] Equation 8

[294] FIG. 25 and 26 are diagrams illustrating a motion vector candidate derivation method using motion information of an adjacent block in the affine inter mode according to an embodiment of the disclosure.

[295] In an affine candidate list, affine motion continues from spatial neighbor blocks (extrapolated affine candidates), and the affine candidate list is augmented by a combination of motion vectors from spatial neighbor blocks (virtual affine candidates). Candidate sets are set up as follows:

[296] 1. A maximum of two different sets of affine MV predictor are derived from the affine motion of neighboring blocks. Neighbor blocks A0, A1, B0, B1 and B2 are checked as illustrated in FIG. 25. If an adjacent block is encoded with an affine motion model and the corresponding reference frame is the same as the current block's reference frame, two control points (for a 4-parameter affine model) of the current block or three control points (for a 6-parameter affine model ) of the current block are derived from the affine model of neighboring blocks.

[297] 2. FIG. 29 illustrates neighboring blocks used to generate a set of virtual affine candidates. Neighboring MVs are divided into three groups: S_0={mv_A, mv_B, mv_C}, S_1={mv_D, mv_E} and S_2={mv_F, mv_G}. mv_0 is the first MV that refers to the same reference picture as the current block in S0. mv_2 is the first MV that refers to the same reference picture as the current block in S1.

[298] If mv_0 and mv_1 are given, then mv_2 can be derived by Equation 9 below.

[299] Equation 9

[300] In Equation 9, the size of the current block is W×H.

[301] If only mv_0 and mv_2 are given, then mv_1 can be output by Equation 10 below.

[302] Equation 10

[303] In an embodiment of the disclosure, affine inter-prediction may be performed in accordance with the following sequence.

[304] Input: Affine Motion Parameters, Reference Picture Samples

[305] Output: CU prediction block

[306] Process

[307] - Print affine block subblock size

[308] - If both the sub-block width and height are greater than 4 luma samples,

[309] - For each sub-block,

[310] - Output motion vector of subblock

[311] - Perform motion compensation (1/16 pel for luma and 1/32 pel for chroma) based on DCT-IF on subblocks (enabled)

[312] - If not, perform (enable) compensation based on extended bilinear interpolation filter on all affine blocks

[313] Moreover, in an embodiment of the disclosure, if the merge/skip flag is false and the width and height of the CU is greater than or equal to 8, then an affine flag is signaled to indicate whether the affine inter-mode will be used at the CU level. If the CU is encoded in affine inter-mode, then a model flag is signaled to indicate whether a 4-parameter or 6-parameter affine CU model is applied. If the model flag is true, then the AF_6_INTER mode (6-parameter affine model) is applied and the MVD is parsed. If not, then the AF_4_INTER mode (4-parameter affine model) is applied and the two MVDs are parsed.

[314] In the AF_4_INTER mode, similarly to the affine combining mode, pairs of motion vectors extrapolated from neighboring blocks encoded by the affine mode are generated and preferably inserted into the candidate list.

[315] Thereafter, if the candidate list size is less than 4, candidates having a motion vector pair {(v_0,v_1)|v0={v_A,v_B, v_C}, v_1={v_D, v_E}} are generated using neighboring blocks . As illustrated in FIG. 22, v_0 is selected from motion vectors A, B, and C blocks. The motion vector from the neighbor block is scaled based on the relationship between the reference list, the POC for the neighbor block reference, the POC for the current CU reference, and the current CU. Moreover, the approach to choosing v_1 from adjacent blocks D and E is similar. When the list of candidates is greater than 4, the candidates are first sorted based on the consistency of neighboring motion vectors (similar to two motion vectors in a pair of candidates), and the first 4 candidates are kept.

[316] If the number of candidate lists is less than 4, the list is populated with a pair of motion vectors by duplicating AMVP candidates.

[317] In the AF_6_INTER mode, similarly to the affine pooling mode, triplets of motion vectors (affine motion vector triplets) extrapolated from adjacent blocks encoded in the affine pooling mode are generated and preferably inserted into the candidate list.

[318] Thereafter, when the candidate list size is less than 4, candidates including the triples {(v_0, v_1, v_2)|v0={v_A, v_B, v_C}, v1={v_D, v_E}, v2={v_G , v_H}} motion vectors are generated using neighboring blocks. As illustrated in FIG. 22, v_0 is selected from the motion vectors of block A, B, or C. The motion vector from the neighboring block is scaled based on the relationship between the reference list, the POC for the neighbor block reference, the POC for the current CU reference, and the current CU. Moreover, the approach for choosing v_1 from adjacent blocks D and E is similar to choosing v_2 from F and G. When the list of candidates is greater than 4, the candidates are sorted based on the consistency of neighboring motion vectors (similar to two motion vectors in three candidates), and the first 4 candidates are kept .

[319] When the number of candidate lists is less than 4, the list may be populated with motion vector triplets configured by duplicating the respective AMVP candidates.

[320] After the CPMV of the current CU is derived, the MVF of the current CU is generated in accordance with Equation 11 for the 4-parameter affine model, and generated in accordance with Equation 12 for the 6-parameter affine model, based on the number of affine parameters.

[321] Equation 11

[322] Equation 12

[323] In this case, the subblock size M×N is output in Equation 13, and MvPre is the motion vector fractional precision (1/16).

[324] Equation 13

[325] After derivation by Equation 12, M and N should be reduced, if necessary, to make them divisors of w and h. When M or N is less than 8, WIF is applied. If not, then sub-block based affine motion compensation is applied.

[326] FIG. 27 illustrates an example of a subblock unit motion vector affine field in accordance with an embodiment of the disclosure.

[327] With reference to FIG. 27 to output the motion vector of each sub-block M×N, the motion vector of the center sample of each sub-block, such as illustrated in FIG. 27 is calculated according to Equation 11 or Equation 12, and rounded to the nearest 1/16 part. Upsampling SHVC interpolation filters are used to generate a prediction of each subblock using the derived motion vector.

[328] Upsampling SHVC interpolation filters having the same filter length and normalization factor as the HEVC motion compensation interpolation filters can be used as motion compensation interpolation filters for additional fractional pel positions (with pel fraction accuracy). The motion vector accuracy of the chrominance component is 1/32 of the sample. Additional interpolation filters for fractional locations with 1/32 pel precision are derived using the average of the filters of two adjacent fractional locations with 1/16 pel accuracy.

[329] The AF_MERGE mode may be selected at the encoder side using the same general merging mode selection method. A list of candidates is preferably generated and a minimum RD cost is selected in the candidates for comparison with the RD cost of other inter-modes. The result of the comparison is to determine whether AF_MERGE is applied or not.

[330] For the AF_4_INTER mode, an RD cost check is used to determine whether a motion vector pair candidate is selected as a control point motion vector prediction (CPMVP) of the current CU. Once the CPMVP of the current affine CU is determined, the affine motion estimate is applied and a control point motion vector (CPMV) is obtained. Accordingly, the difference between CPMV and CPMVP is determined.

[331] On the encoder side, the AF_6_INTER mode is identified only when the AF_MERGE or AF_4_INTER mode is determined to be the optimal mode in the previous mode selection stage.

[332] In an embodiment of the disclosure, affine inter-mode (affine AMVP) can be performed as follows:

[333] 1) AFFINE_MERGE_IMPROVE: Instead of searching for the first adjacent block in affine mode, the improvement is to look for an adjacent block having the maximum coding unit size as an affine merge candidate.

[334] 2) AFFINE_AMVL_IMPROVE: Adjacent blocks in affine mode are added to the affine AMVP candidate list in a similar manner to the general AMVP procedure.

[335] The detailed process for generating an affine AMVP candidate list is as follows.

[336] First, it is identified whether the lower left adjacent block uses an affine motion model and has the same reference index as the current reference index. If there is no neighbor block, then the left neighbor block is identified using the same method. If there is no adjacent block, then it is identified whether the lower left adjacent block uses an affine motion model and has a different reference index. If a neighboring block is present, the scaled affine motion vector is added to the reference picture list. If there is no neighbor block, then the left neighbor block is identified using the same method.

[337] Next, the top right neighbor, the top neighbor, and the top left neighbor are identified using the same method.

[338] After these processes, if two candidates are found, the affine AMVP candidate list generation process ends. If two candidates are not found, then the initial operation in the JEM software is performed to generate affine AMVP candidate lists.

[339] 3) AFFINE_SIX_PARAM: In addition to the 4-parameter affine motion model, the 6-parameter affine motion model is added as an additional model.

[340] A 6-parameter affine motion model is derived using Equation 14.

[341] Equation 14

[342] Three motion vectors at the top left location MV_0, the top right location MV_1, and the bottom left location MV_2 are needed to determine the model because 6 parameters are present in the motion model. The three motion vectors may be determined using a method similar to the two motion vector method in the 4-parameter affine motion model. The affine pool model is always set as a 6-parameter affine motion model.

[343] 4) AFFINE_CLIP_REMOVE: Motion vector constraints for all affine motion vectors are removed. Motion compensation processes are performed to manage the motion vector constraints themselves.

[344]

[345] Affine Motion Model

[346] As described above, various models of affine motion may be used or considered in affine inter-prediction. For example, an affine motion model may represent four motions as shown in FIG. 15. An affine motion model capable of representing three motions (translation, scaling and rotation), among the motions that can be represented by an affine motion model, may be referred to as a (simplified) affine motion similarity model. The number of CPMVs and/or the way in which the sample unit/subblock MV of the current block is derived may be different depending on which of the affine motion models is used.

[347] In an embodiment of the disclosure, adaptive four- and six-parameter motion models are used. In AF_INTER, a 6-parameter motion model is offered in addition to the 4-parameter motion model provided in JEM. The 6-parameter affine motion model is described as Equation 15.

[348] Equation 15

[349] In this case, the coefficients a, b, c, d, e and f are parameters of the affine motion. (x, y) and (x',y') are the coordinates in pixel locations before and after the affine motion model transformation. In video coding, to use the affine motion model, if CPMV0, CPMV1 and CPMV2 are the MVs for CP0 (top left), CP1 (top right) and CP2 (bottom left), Equation 16 can be described as follows.

[350] Equation 16

[351] In this case, CPMV_0={v_0x, v_0y}, CPMV_1={v_1x, v_1y}, CPMV_2={v_2x, v_2y}, and w and h are the width and height of each coding block. Equation 16 is the motion vector field (MVF) of the block.

[352] A flag is parsed at the CU level to indicate whether a 4-parameter or 6-parameter affine motion model is used when neighboring blocks are encoded as affine prediction. If there is no neighbor block encoded as affine prediction, the flag is omitted and the 4-parameter model is used for affine prediction. In other words, a 6-parameter model is considered under the condition that one or more neighboring blocks are encoded in an affine motion model. With respect to the number of CPMVDs, each of two or three CPMVDs is signaled against 4-parameter and 6-parameter affine motion models.

[353] In addition, in an embodiment of the disclosure, motion vector refinement by pattern matching may be used. When deriving a motion vector by pattern matching (PMMVD, hereinafter abbreviated as PMVD in the JEM encoder description) JEM, the decoder needs to evaluate some motion vectors (MVs) to determine an initial MV candidate for searching at the CU level. When searching at the sub-CU level, some MV candidates are added in addition to the optimal MV at the CU level. The decoder needs to evaluate such MV candidates in order to perform a search for the optimal MV. This requires a large amount of memory. In the proposed motion vector refinement by pattern matching (PMVR), the concepts of pattern matching and bidirectional matching in PMVD are adopted in JEM. When skip mode or merge mode is selected to indicate whether PMVR is available, one PMVR_flag is signaled. To significantly reduce the memory bandwidth requirement compared to PMVD, a list of MV candidates is generated. If PMVR is applied, the initial MV candidate index is explicitly signaled.

[354] A list of candidates is generated using a merge candidate list generation process, but sub-CU merge candidates such as affine candidates and ATMVP candidates are excluded. For bidirectional matching, only the unidirectional (uni-) prediction MV candidate is included. The bi-prediction MV candidate is divided into two uni-prediction MV candidates. In addition, similar MV candidates (having MV differences less than a predetermined threshold) are also removed. For CU level searches, diamond search MV refinement is performed starting from the signaled MV candidate.

[355] Searching at the sub-CU level is only available in the two-way negotiation merge mode. The search window for sub-CU level search for all sub-CUs is the same as the search window for CU level search. Accordingly, additional bandwidth is not required in the sub-CU level search.

[356] To refine the MVP in some mode, pattern matching is also used. In AMVP mode, two MVPs are generated using the HEVC MVP generation process, and one MVP index is signaled to select one of the two MVPs. The selected MVP is further refined using pattern matching in PMVR. If adaptive motion vector resolution (AMVR) is applied, the MVP is rounded to the appropriate precision before template matching is refined. Such a refinement process is referred to as motion vector predictor pattern matching (PMVPR) refinement. Throughout the remainder of this document, PMVR includes PMVR pattern matching, PMVR bidirectional matching, and PMVPR unless specifically defined otherwise.

[357] To reduce the memory bandwidth requirement, PMVR is unavailable for 4×4, 4×8 and 8×4 CUs. To reduce the amount of additional memory bandwidth required, the search range {pattern matching, bidirectional matching} for a CU area of 64 may be reduced to {±2,±4}. The search range of {pattern matching, bidirectional matching} for a CU area larger than 64 may be reduced to {±6,±8}. Compared to the worst case in HEVC, the required memory bandwidth was reduced from 45.9x in PMVD JEM-7.0 to 3.1x in PMVR using all the methods described in the PMVR section of this document.

[358]

[359] Technique used when affine mode is used in a non-QT box

[360] FIG. 28 illustrates a method for generating a prediction block and an inter-prediction motion vector to which an affine motion model has been applied in accordance with an embodiment of the disclosure.

[361] FIG. 28, you can see the equation for deriving the motion vector if the affine motion model is applied. The motion vector can be derived based on Equation 17 below.

[362] Equation 17

[363] In this case, v_x indicates the x-component of the motion vector of the unit for the sample with coordinates (x, y) in the current block. v_y specifies the y-component of the sample unit motion vector for the sample at (x, y) coordinates in the current block. That is, (v_x, v_y) is the unit sample motion vector for the sample with coordinates (x, y). In this case, a, b, c, d, e, f indicate the parameters of an equation for outputting a motion vector of a sample unit (motion information) with coordinates (x, y) from control points (CP) of the current block. CP can be represented as a control element of the picture. The parameters may be derived from the CP motion information of each PU transmitted in the PU. An equation for deriving a motion vector of a sample unit derived from the CP motion information may be applied to each block sample, and may be output as a sample location in a reference picture depending on the relative location between the x-axis and the y-axis of each sample. The motion vector of the sample unit may be output differently depending on the block size according to the QTBT(TT) block division structure, asymmetric type or symmetric type, or block location. Detailed embodiments of this are illustrated in FIG. 29 - fig. 38 described below.

[364] FIG. 29 is a diagram illustrating a method for performing motion compensation based on a motion vector of a reference point according to an embodiment of the disclosure.

[365] With reference to FIG. 29, it is assumed that the current block is a 2N×2N block. For example, the motion vector of the top left sample in the current block could be v_0. In addition, the motion vectors of the CP may be v_1 and v_2 using neighbor block samples adjacent to the current block as the CP. That is, if the width and height of the current block are S, and the coordinates at the top left sample position of the current block are (xp, yp), the CP0 CP coordinates can be (xp, yp), the CP1 CP coordinates can be (xp+S, yp ), CP2 CP coordinates can be (xp, yp+S). The motion vector CP0 may be v_0, the motion vector CP1 may be v_1, and the motion vector CP2 may be v_2. The motion vector of the sample unit may be output using the motion vector of the CP. The sample unit motion vector can be derived based on Equation 18 below.

[366] Equation 18

[367] In this case, v_x, v_y indicate the x-component, y-component of the motion vector for the sample having coordinates (x, y) in each current block. v_x0, v_y0 indicate the x-component, y-component of the motion vector v_0 for CP0, respectively. v_x1, v_y1 indicate the x-component, y-component of the motion vector v_1 of motion for CP1, respectively. v_x2, v_y2 indicate the x-component, y-component of the motion vector v_2 for CP2, respectively. The motion vector of each sample in the current block may be derived based on the relative location in the current block using an equation for inferring the motion vector of a sample unit, such as Equation 18.

[368] FIG. 30 is a diagram illustrating a method for performing motion compensation based on the motion vector of a reference point in a non-square block according to an embodiment of the disclosure.

[369] FIG. 30 illustrates the CP of a block divided into N×2N. An equation for deriving the motion vector of a sample unit in the current block may be derived using the same method as in the case of the partition type 2N×2N. The equation derivation process can use a width value appropriate for the shape of the current box. To derive the motion vector of the sample unit, three CPs may be derived, and the CP locations may be adjusted as in FIG. 30. That is, if the width and height of the current block are S/2 and S, respectively, and the coordinates at the top left sample position of the current block are (xp, yp), the CP0 CP coordinates can be (xp, yp), the CP1 CP coordinates can be (xp+S/2, yp), and CP2 CP coordinates can be (xp, yp+S). The sample unit movement vector can be derived based on Equation 19 below.

[370] Equation 19

[371] In this case, vx, vy indicate the x-component, y-component of the motion vector to sample at the (x, y) coordinates in the current block, respectively. v_x0, v_y0 indicate the x-component, y-component of the motion vector v_0 for CP0, respectively. v_x1, v_y1 indicate the x-component, y-component of the motion vector v_1 for CP1, respectively. v_x2, v_y2 indicate the x-component, y-component of the motion vector v_2 for CP2, respectively. Equation 3 illustrates the equation for deriving the motion vector of a sample unit, given that the width of the current block is S/2. Each sample in the current block partitioned from the CU based on the N×2N partition type using an Equation to derive a motion vector of a sample unit, such as Equation 19, may be derived based on a relative location in the current block.

[372] FIG. 31 is a diagram illustrating a method for performing motion compensation based on the motion vector of a reference point in a non-square block according to an embodiment of the disclosure.

[373] FIG. 31 illustrates blocks divided based on the 2N×N partition type. To output the motion vector of a sample unit, three CPs can be output. The location of the CP can be adjusted as in FIG. 31, and the height can be adjusted up to S/2 based on the shape of the current block illustrated in FIG. 31. That is, if the width and height of the current block are S and S/2, respectively, and the coordinates at the upper left sample position of the current block are (xp, yp), the CP0 CP coordinates can be (xp, yp), the CP1 CP coordinates can be (xp+S, yp), and CP2 CP coordinates can be (xp, yp+S/2). The sample unit motion vector can be derived as in Equation 20 below.

[374] Equation 20

[375] In this case, v_x, v_y indicate the x-component, y-component of the motion vector to be sampled at the (x, y) coordinates in the current block, respectively. v_x0, v_y0 indicate the x-component, y-component of the motion vector v_0 for CP0, respectively. v_x1, v_y1 indicate the x-component, y-component of the motion vector v_1 for CP1, respectively. v_x2, v_y2 indicate the x-component, y-component of the motion vector v_2 for CP2, respectively. Equation 4 illustrates the equation for deriving the motion vector of a sample unit by considering that the height of the current block is S/2. Each sample in the current block partitioned from the CU based on the Nx2N partition type using an equation for deriving a motion vector of a sample unit such as Equation 4.18 can be derived based on the relative location in the current block.

[376] FIG. 32-38 are diagrams illustrating a method for performing motion compensation based on the motion vector of a reference point in a non-square block according to an embodiment of the disclosure.

[377] FIG. 32 illustrates the CP of the current blocks of the asymmetric type. As illustrated in FIG. 32, the width and height of current asymmetric type blocks may be referred to as W and H, respectively. In order to output the motion vector of the sample unit, three CPs for each current block can be output, and the coordinates of the CP can be adjusted based on the width and height according to the shape of the current block as in FIG. 32. That is, if the width and height of the current block are W and H, and the coordinates at the top-left sampling position of each current block are (xp, yp), the CP0 CP coordinates can be set as (xp, yp), the CP1 CP coordinates can be set as (xp+W, yp) and the CP2 CP coordinates can be set as (xp, yp+H). In this case, the motion vector of the sample unit in the current block may be derived based on Equation 21 below.

[378] Equation 21

[379] In this case, v_x, v_y indicate the x-component, y-component of the motion vector to be sampled at the (x, y) coordinates in the current block, respectively. v_x0, v_y0 indicate the x-component, y-component of the motion vector v_0 for CP0, respectively. v_x1, v_y1 indicate the x-component, y-component of the motion vector v_1 for CP1, respectively. v_x2, v_y2 indicate the x-component, y-component of the motion vector v_2 for CP2, respectively. Equation 21 illustrates the equation for deriving the motion vector of a sample unit by taking into account the width and height of the current asymmetric type blocks.

[380] Meanwhile, according to the disclosure, in order to reduce the amount of CP motion information data indicated in a block unit, a motion information prediction candidate for at least one CP may be selected based on the motion information of a neighboring block or a neighboring sample of the current block. The motion information prediction candidate may be referred to as a motion information affine candidate or a motion vector affine candidate. The affine motion information candidates may include the content disclosed in, for example, FIG. 33-38.

[381]

[382] ATMVP Simplification

[383] Extended temporal motion vector prediction (ATMVP) is inter-prediction using a temporal motion information candidate, but may indicate inter-prediction using a motion information candidate proposed to improve existing TMVP. In an embodiment of the disclosure, ATMVP is not limited by its name, and ATMVP may be referred to as a temporal merge candidate subblock, subblock based temporal merge candidate, subblock temporal motion vector predictor, or subblock based temporal motion vector predictor.

[384] Specifically, intra-frame motion may not be included in TMVP using the motion vector colPB at the bottom right block of the current block or at the center location of the current block as a temporal motion information candidate. On the contrary, if ATMVP is applied, the motion vector colPB at the location indicated by the motion vector of the neighboring block may be used as an ATMVP candidate.

[385] In the example in which ATMVP is applied, after the initial search for the motion vector of the available spatial neighbor block while sequentially checking the combination candidate configurations, the location indicated by the temporal vector in the reference picture may be output as col-PB (ATMVP candidate). Moreover, the motion vector of the corresponding block can be used in each sub-block unit using the time vector. In this case, if the MV is not present in a particular sub-block, the MV of the block located in the center of the corresponding block is used as the MV for the unavailable sub-block and stored as the representative MV.

[386] In addition, various ATMVP simplification schemes for using ATMVP can be proposed, but capable of reducing the amount of memory used.

[387] In one embodiment, memory usage can be reduced because the reference picture (indicated as the slice segment header) for finding the reference block is limited to the merged picture. For example, a maximum of 4 reference pictures can be used. To search for a reference block, a temporal vector may be output in the scan sequence from one of the spatial candidates. If the current candidate has the same reference picture as the merged picture, the search process may end.

[388] In one embodiment, a modification may be applied to the ATMVP and/or STMVP subunit combining modes.

[389] 1. Signaling is added at the picture/cut level, and thus different sub-block sizes can be used to output motion information based on ATMVP/STMVP.

[390] 2. ATMVP-based simplified traffic information output can be used because one merged picture is used.

[391] In one embodiment, support is provided for adapting, at the picture/clipping level, the granularity of the motion field output in the ATMVP and STMVP modes. Specifically, in a slice that indicates a sequence parameter set (SPS), a default value of the subblock size used to derive a motion parameter for ATMVP and STMVP may be signaled in the SPS. Additionally, one flag is signaled in the slice segment header. If the flag is 0, it indicates that a default sub-block size, such as specified in the SPS, is used to infer traffic based on ATMVP/STMVP in the current slice. If not (ie, if the flag is set to 1), another syntax element is signaled in the slice segment header and may specify the size of the ATMVP/STMVP subblock used for the slice.

. Соответственно, когда (i+1)-ая картинка кодируется в том же самом k-ом временном уровне, соответствующий размер

подблока ATMVP/STMVP может определяться в соответствии с Уравнением 22 ниже.[392] In one embodiment, a simple encoder method is used to determine the ATMVP/STMVP subblock size for the current picture. One of the two ATMVP/STMVP subblock sizes 4 and 8 is selected from the picture finally encoded at the same time level based on the average ATMVP/STMVP block size. It is assumed that the current encoded picture is the i-th picture in the k-th time layer and includes N CUs encoded in the ATMVP and STMVP modes. Further, it is assumed that the size of the CU is S_0, S_1, ..., S_(N-1). The average ATMVP/STMVP CU size is calculated as

. Accordingly, when the (i+1)th picture is encoded in the same kth time layer, the corresponding size

ATMVP/STMVP subblock may be determined in accordance with Equation 22 below.

[393] Equation 22

[394] In one embodiment, the ATMVP/STMVP subblock size may always be set to 4 with respect to the first picture in each temporal layer.

[395] In one embodiment, ATMVP may be derived using the following two steps.

[396] (1) If a neighbor CU is available and the MV of the CU is different from the MV in the existing candidate list, the encoder/decoder may add a motion vector to the sequence A1, B1, B0, and A0 of the spatial neighbor CUs, as illustrated in FIG. 12. The number of available spatial candidates can be represented as N0.

[397] (2) The first MV candidate of the N0 spatial candidates may be used to determine the location from which the motions for the combined picture and each sub-block will be extracted (by adding an offset according to the MV). If N0 is 0, a co-merged location having a co-located picture and zero motion signaled in the slice header can be used to extract the motion for each sub-block.

[398] The combined picture of different CUs for ATMVP may not always be the same if multiple reference pictures are used. If the merged pictures for ATMVP output are different for different CUs in the current picture, it means that motion fields of a plurality of reference pictures do not need to be extracted. This is not preferred due to the increased memory bandwidth. Accordingly, one embodiment proposes a simplified design using the same alias picture as the alias picture for ATMVP output as in HEVC signaled in the slice header. At the block level, if the adjacent block A's reference picture is different from the superimposed picture, the MV of block A is scaled using the HEVC temporal MV scaling method, and the scaled MV of block A can be used for ATMVP.

[399] The motion vector used to extract the motion field in the combined picture Rcol is labeled MVcol. To minimize the impact caused by MV scaling, the MV in the spatial candidate list used to derive MVcol is selected using the following method. If the candidate reference picture MV is a composite picture, the MV is selected and used as the MVcol without any scaling. If not, the MV having the reference picture closest to the combined picture is used to output the MVcol having the scaling.

[400]

[401] History Based General Motion Vector Prediction (HMVP)

[402] In general, image compression technology uses the use of spatial and temporal redundancy as two main schemes. For example, both High Efficiency Video Coding (HEVC) and VVC use two motion compression schemes based on inter-coding. One is motion join and the other is advanced motion vector prediction (AMVP). In order to improve the two prediction modes, various modifications are considered. Various modifications include increasing the number of candidates to search for more advanced spatial candidates and testing temporary candidates at non-traditional locations. The two schemes mainly include list generation using available candidates, distortion-rate (RD) cost minimization, and signaling of the selected candidate in the bitstream.

[403] In particular, in a recent image compression technology, an HMVP is considered, in which motion information of a previously encoded block is stored, and the stored motion information is used to predict the motion of a sequentially encoded block. Such an HMVP may be added to a merge list (or merge candidate list) or an AMVP list (or AMVP candidate list).

[404] The decoder supports a lookup table (LUT) operating in a first-in-first-out (FIFO) system (or method) for HMVP. In the disclosure, the LUT is not limited by its name and may be referred to as a table, HMVP table, HMVP candidate table, buffer, HMVP buffer, HMVP candidate buffer, HMVP list, or HMVP candidate list. Specifically, when a non-affine prediction unit (PU) (or coding unit (CU)) is decoded, the corresponding motion information is stored in the LUT. The decoder performs decoding at the next PU. In this case, the stored motion information may include motion vectors, reference index information, and mode information in the x- (horizontal) and y- (vertical) directions.

[405] The decoder maintains a LUT that stores motion information of a progressively decoded non-affine candidate. The LUT size may be limited by predefined S candidates. In one embodiment, the LUT may be reset at the start of a slice, the start of a CTU line, or the start of a CTU.

[406] HMVP can be used in both merge mode and AMVP mode. The merge list may have B candidates, and the AMVP list may have two candidates. In the conventional image compression technology, the merge list is configured with the following candidates: i) spatial candidate, ii) temporal candidate, iii) bi-directional prediction (Bi-Pred) candidate, iv) zero motion candidate. Recently, a method has been considered to further consider advanced motion vector prediction (ATMVP) as a candidate. For example, the ATMVP candidate may be inserted into the merge list before the temporary candidate. Merge list candidates are added to the merge list until they reach the maximum size of the merge list. A duplicate candidate may not be added to the merge list. Two candidates can be inserted into the AMVP list. For example, the first candidate of two candidates may be selected from available spatial candidates, and the second candidate may be selected from temporal candidates. If the list is empty, a null motion vector candidate may be added.

[407] HMVP is applied based on the FIFO order in which candidates exit the LUT is identical to their entry order.

[408] In one embodiment, when the HMVP is applied to the merge list configuration, the HMVP candidate may be inserted (or added) to the third list location as follows:

[409] 1. Spatial Candidate

[410] 2. Provisional Candidate

[411] 3. Up to S LUT HMVP Candidates

[412] 4. Combined Bi-Pred Candidate

[413] 5. Zero Motion Vector Candidate

[414] In one embodiment, when an HMVP is applied to an AMVP list configuration, the HMVP may be inserted in a third location after the temporary candidate as follows:

[415] 1. Spatial Candidate

[416] 2. Provisional Candidate

[417] 3. Up to K HMVP candidates

[418] 4. Zero Motion Vector Candidate

[419] FIG. 39 is a flowchart for describing a method for storing an HMVP according to an embodiment of the disclosure.

[420] With reference to FIG. 39, the decoder decodes the current PU (or CU) (S3901).

[421] The decoder checks whether the current PU is a non-affine encoded block (S3902). To facilitate the use of HMVP candidates, if the current PU is an affine-mode encoded block, the decoder does not store the current PU's motion information in a table.

[422] If the current PU is a block encoded in a non-affine mode, the decoder stores (or updates) the motion information of the current PU in a table (S3903).

[423] In an embodiment of the disclosure, the HMVP table may be updated using two methods, i.e., methods i) unlimited FIFO ii) limited FIFO. In the first, redundant information can be presented, but no pruning process is applied. This helps to reduce the complexity of the entire process. This is described with reference to the following drawing.

[424] FIG. 40 is a diagram for describing an HMVP table applied in an unconstrained FIFO according to an embodiment of the disclosure.

[425] With reference to FIG. 40, the candidate added to the table is added at the end (right) of the table. In contrast, the candidate removed from the table according to the FIFO method is located at the front end (left side, oldest candidate) of the table.

[426] If the table is not completely filled with the maximum number of predefined candidates with index L-1 (ie, at the end), new candidates are added without removing the candidate. On the contrary, if the table has already been completely filled, that is, if the maximum number of the table is satisfied, the candidate located at the front end, that is, the oldest one in the table, is removed and a new candidate is added.

[427] FIG. 41 is a diagram for describing an HMVP table operating in a restricted FIFO according to an embodiment of the disclosure.

[428] With reference to FIG. 41, if a limited FIFO is used, then pruning is performed if the addition of a new candidate causes any redundancy. In an embodiment, if a candidate having redundant motion information is present in the table, the redundant candidate in the table is removed and the current candidate's motion information may be added.

[429]

[430] Embodiment 1

[431] With respect to HMVP candidates, in many cases, the MVs of the most recent history may overlap the motion information of the spatial candidate (or spatial neighbor candidate). Accordingly, the present embodiment proposes a method for setting the additional candidate order differently from the HMVP LUT index order when an HMVP candidate is added to an AMVP list or a merge list.

[432] According to an embodiment of the disclosure, the list of candidates can be efficiently configured by adaptively adjusting the HMVP candidate. Accordingly, the number of signaling bins used for binarization can be reduced and coding efficiency can be improved.

[433] That is, the HMVP candidate added to the merge list or the AMVP list may not be limited to an index in the HMVP list. In one embodiment, Table 2 below illustrates a method for changing the order in which an HMVP candidate is added to an AMVP list or merge list.

[434] Table 2

HMVP LUT index (0 denotes the MV of the most recent story) Order to add to AMVP list or merge list 0 2 1 0 2 1 3 3 .. ..

[435] With reference to Table 2 as described above, there is a high probability that the most recently entered HMVP candidate may have the same motion information as the spatial candidate. Accordingly, the addition order of the HMVP candidate can be predetermined regardless of the HMVP index in view of the probability.

[436] In addition, in one embodiment, the HMVP candidate may be added to the merge list or the AMVP list from the HMVP candidate starting with the nth candidate in the list. Table 3 below illustrates the modified order for adding a candidate to an AMVP list or pool list.

[437] Table 3

HMVP LUT Index
(0 denotes the MV of the most recent story) Order to add to AMVP list or merge list 0 1 0 2 1 3 2 .. ..

[438] With reference to Table 3, the HMVP candidate may be added to the merge list or the AMVP list from the second index.

[439] In one embodiment, information about the order of adding the HMVP candidate in the table (LUT) can be signaled from the encoder to the decoder. For example, such order information may be conveyed via High Level Syntax (HLS). The high level syntax may be, for example, a sequence parameter set, a picture parameter set, a slice header, a coding tree unit, an coding unit, and/or other suitable syntax data header.

[440] Table 4 illustrates the structure of the high-level syntax to which the method proposed in the disclosure can be applied.

[441] Table 4

[442] With reference to Table 4, set_HMVP_order_flag equal to 1 indicates that set_HMVP_order_flag is present in the slice header in non-IDR pictures in CVS. set_HMVP_order_flag equal to 0 indicates that set_HMVP_order_flag is not present in slice headers and adaptive HMVP is not used in VCS.

[443] Table 5 illustrates a slice segment header syntax structure to which the method proposed in the disclosure can be applied.

[444] Table 5

[445] With reference to Table 5, slice_HMVP_idx means an index for the sequence of candidates used. For example, slice_HMVP_idx equal to 0 may represent a basic HMVP sequence such as 0, 1, 2, 3. Similarly, an index value of 1 may be used to represent an HMVP sequence of 3, 2, 1, 0.

[446]

[447] Embodiment 2

[448] In an embodiment of the disclosure, in addition to the HMVP LUT, a method is provided for using a long-term list for motion prediction. Accordingly, the number of supported HMVP candidates may increase. In an embodiment, 2 HMVP tables may be considered. In this case, one can be used to store generic HMVP candidates and the other can be used as a long-term list that stores candidates that need to be further supported.

[449] The following illustrates methods for resetting and generating a long-term list (or long-term HMVP list).

[450] - After the first CTU of the CTU string is decoded, one or more successive CTU history MVs may be added to the long-term HMVP LUT. Such a long-term HMVP LUT may not be used or updated to the next CTU line.

[451] - At the start of the next CTU line, a long-term HMVP LUT may be used to reset the overall HMVP LUT. The reason for this is that the HMVP CTU candidates at the beginning of a CTU row may be more related than the history MV at the end of the previous CTU row.

[452] - The above process can be repeated.

[453] FIG. 42 is a diagram illustrating an HMVP LUT and a long term HMVP LUT according to an embodiment of the disclosure.

[454] With reference to FIG. 42, two LUTs are shown. One is an HMVP LUT (either a general HMVP LUT or a short term HMVP LUT) and the other is a long term HMVP LUT. When an HMVP candidate is added to the merge and AMVP lists, it may be added from an HMVP LUT or a long-term LUT, as illustrated in FIG. 42.

[455] The use of a long-term LUT can be indicated by high-level syntax by introducing a new syntax element. For example, a syntax element may be represented in a sequence parameter set, a picture parameter set, a slice header, a coding tree unit, an coding unit, and/or other suitable syntax data header.

[456]

[457] Embodiment 3

[458] In an embodiment of the disclosure, a method is provided in which flexibility for decoding is taken into account in adding an HMVP candidate to an HMVP LUT. The encoder/decoder may consider a decision for one or more PU (or CU) characteristics.

[459] In an embodiment, the encoder/decoder may consider the following content when adding the HMVP candidate to the table. The encoder/decoder may add a candidate to the table based on characteristics such as mode (eg, merge mode, affine mode, or AMVP mode) of the PU and/or block size, individually or in combination. In one embodiment, in addition to these characteristics, other characteristics may be considered. For example, the type of association (eg, spatial candidate or temporal candidate) in which the update of the HMVP LUT is considered, whether it is a sub-PU, etc., may be considered as a criterion for selecting a candidate. The selection criterion may be defined to reduce redundancy with the previous history (or previous HMVP). For example, if the PU is encoded in merging mode and the merging type is spatial merging, the decoder may not update the HMVP LUT with motion information of the corresponding PU.

[460] FIG. 43 is a diagram illustrating an example of a method for updating an HMVP LUT according to an embodiment of the disclosure.

[461] With reference to FIG. 43, the encoder/decoder obtains the motion information of the encoded candidate (S4301).

[462] The encoder/decoder judges whether to update the LUT with candidate motion information based on a predetermined decision criterion (S4302). The decision criterion may include characteristics associated with one or more of a candidate mode (eg, merge mode, affine mode, or AMVP mode), candidate block size, and/or candidate merge type.

[463] The encoder/decoder updates the LUT based on the decision criterion (S4303). That is, if the candidate satisfies the predetermined decision criterion, the encoder/decoder may add motion information of the candidate to the LUT.

[464]

[465] Embodiment 4

[466] In an embodiment of the disclosure, redundancy check restrictions for adding an HMVP candidate to a merge list (or AMVP list) are proposed. Redundancy check constraints can be implemented in various ways.

[467] In one embodiment, the encoder/decoder may limit the number of clipping checks for the first specific number of candidates in the merge list. The encoder/decoder may limit the number of pruning checks for candidates from the first candidate of the merge list to a particular candidate. For example, the encoder/decoder may perform a pruning process on candidates from the first candidate of the merge list to a particular candidate. In this case, the HMVP candidate, that is, the target of the pruning check, may be limited to a predefined number.

[468] Moreover, in one embodiment, the encoder/decoder may limit the pruning check by performing a pruning check on a particular merge candidate type in the merge list. For example, the encoder/decoder may only perform pruning checks on the merge list spatial candidate when adding the HMVP candidate. Alternatively, for example, the encoder/decoder may perform a pruning check on only some of the spatial merge list candidates when adding the HMVP candidate. Some of the spatial candidates may be predefined. For example, the predefined spatial candidates may be at least one of a left neighbor spatial candidate and/or a top neighbor spatial candidate. The embodiment of the disclosure is not limited to this, and other types may be combined and limited as a pruning test target.

[469] FIG. 44 is a diagram illustrating a method for limiting the number of HMVP candidates, that is, a pruning test target, according to an embodiment of the disclosure.

[470] With reference to FIG. 44, in the disclosure embodiment, the number of HMVP candidates, that is, pruning test targets, may be limited to four times.

[471] The encoder/decoder may check for redundancy of motion information between the first M HMVP LUT candidates and the merge list merge candidates when generating the merge list using the HMVP candidates.

[472] Alternatively, the encoder/decoder may check for redundancy of motion information between the first M candidates in the HMVP LUT and the current decoded PU when adding the motion information of the decoded PU in the HMVP LUT.

[473] FIG. 45 is a diagram illustrating an example of a method for performing clipping checking according to an embodiment of the disclosure.

[474] With reference to FIG. 45, the encoder/decoder obtains motion information of the decoded candidate and determines (or decodes) the number of clipping checks (S4501, S4502).

[475] The encoder/decoder performs the clipping check based on the determined number of clipping checks (S4503).

[476] In one embodiment, as in Table 4 and Table 5, information related to the pruning check may be signaled through high-level syntax. In this case, the syntax element passed from the encoder to the decoder may be included in any header to indicate the number of clipping checks to be performed. The high-level syntax may be, for example, a sequence parameter set, a picture parameter set, a slice header, a coding tree unit, an coding unit, and/or other suitable syntax data header.

[477]

[478] Embodiment 5

[479] In an embodiment of the disclosure, an efficient method for selecting an HMVP candidate is provided. When a history motion vector candidate (ie, an HMVP candidate) is inserted into the merge list (or AMVP list), pruning check can be performed so that the HMVP candidate does not overlap the existing merge list.

[480] If full redundancy checks are to be performed between a merge list of size M and a history LUT of size N, checks (M-1)×N times are necessary. Accordingly, in an embodiment of the disclosure, the number of HMVP candidates may depend on the existing merge candidate. For example, the number of HMVP candidates may depend on the number of spatial candidates present in the merge list. Alternatively, for example, the number of HMVP candidates may depend on the number of spatial candidates and temporal candidates present in the merge list.

[481] If there is another merge candidate present in the merge list, the number of HMVP candidates to be pruned based on a particular criterion (or rule) based on the number of merge candidates and/or the number of HVMP merge list may be reduced. . Accordingly, the number of redundancy checks in the worst case can be reduced.

[482] For example, in the case of a merge list having a size (or length) of 6, if the merge list is full, the merge list may include a maximum of 5 spatial or other merge candidates. To insert an HMVP candidate into 6 HMVP lists, 30 redundancy checks may be required in the worst case.

[483] In one embodiment, an example related to restrictions on the number of HMVPs to be verified is illustrated in Equation 23 and Table 6.

[484] Equation 23

if (existing_candidates >= 3)

number_hist_to_check=7 - existing_candidates;

[485] Table 6

# of an existing candidate # of an existing candidate to test # MV stories
for check # checks 1 1 6 6 2 2 6 12 3 3 four 12 four four 3 12 five five 2 10

[486] With reference to Table 6, the number of redundancy checks for adding an HMVP in the worst case can be reduced to 12 times rather than 30 times by limiting the number of HMVPs, that is, the goal of the pruning check, to 2.

[487]

[488] Embodiment 6

[489] In an embodiment of the disclosure, a method for creating a merge list using history-based spatial-temporal motion vector prediction (H-STMVP) is provided. H-STMVP indicates a candidate derived as the average of two spatial MVPs and a history-based TMVP. Two spatial HMVPs can be obtained from the HMVP buffer. The TMVP can be obtained from the current merge list. In this case, the spatial candidate may be a candidate obtained from the last 2 encoded MVs in the decoding sequence before the current block.

[490] For example, the last encoded MV (referred to as MV_L in the disclosure), the MV that is penultimate (referred to as MV_(L-1) in the disclosure), and MV_TMVP may be used to generate an H-STMVP candidate for insertion into the merge list. .

[491] If all three candidates can be used, the MVs added to the pooling list can be calculated by Equation 24 below.

[492] Equation 24

[493] In one embodiment, if only two of the three candidates are available, only two candidates may be averaged to generate an H-STMVP. Similarly, if only one candidate is available, that one candidate may be used. If there is no available candidate, then the H-STMVP is not used to configure the merge list.

[494] In an embodiment of the disclosure, a method is provided for obtaining a motion vector of an H-STMVP candidate using a method other than Equation 24.

[495] For example, instead of averaging three or more candidates at once, it may be computationally easier to first average the spatial candidates and then average the two candidates again using the result of the averaging. An example of this is illustrated in the following equations.

[496] Equation 25

[497] Alternatively, the average value can be obtained as follows.

[498] Equation 26

[499] Equation 27

[500] Equation 28

[501] The encoder/decoder may first average the two candidates, as in Equations 25-27, and then may average the result values using the third candidate. Alternatively, the encoder/decoder may assign a higher importance/weight to the candidate, i.e., MV_L, by applying the shift operation twice, as in Equation 28. The average value may be derived without the division operation by only the shift operation using Equations 25-28.

[502]

[503] Embodiment 7

[504] In an embodiment of the disclosure, a method is provided using a given number (n) of spatial candidates instead of two history-based spatial candidates in H-STMVP derivation. The n candidates need not be a substantially sequential decoding sequence. The n candidates may be chosen arbitrarily or according to some rule.

[505] Accordingly, Equation 24 can be represented using a more general method, as in Equation 29 below.

[506] Equation 29

[507] In another embodiment, assuming that 5 spatial candidates are used, the enhancement of the influence of spatial candidates to generate an H-STMVP candidate can be minimized, and the spatial candidate and temporal candidate can be appropriately introduced by improving the weight applied to the temporal candidate.

[508] Accordingly, to this end, after the spatial candidates are averaged together using Equation 30, the above goal can be achieved by averaging MV_TMVP using the result of the averaging.

[509] Equation 30

[510]

[511] Embodiment 8

[512] In an embodiment of the disclosure, a method is provided for adding a weight (or weighting factor) to a motion vector candidate used to derive an H-STMVP. In this case, the weight may be determined empirically, or may be determined based on the temporal distance to the fixed reference frame, or may be determined based on the location in the history table. For example, a new candidate may carry more weight than a previous candidate.

[513] That is, in the present embodiment, Equation 24 can be represented as Equation 31 below.

[514] Equation 31

[515] In this case, the weight may have the same value or an unevenly distributed value.

[516]

[517] Embodiment 9

[518] In an embodiment of the disclosure, a method is provided for scaling a motion vector used to derive an H-STMVP candidate as one reference picture.

[519] FIG. 46 is a diagram for describing a method for deriving an H-STMVP candidate using motion vectors that refer to different reference pictures according to an embodiment of the disclosure.

[520] With reference to FIG. 46, candidates MV_L, MV_L-1, and MV_TMVP are assumed to refer to (or point to) respective different reference pictures. That is, Fig. 46 illustrates that the candidates used to generate the H-STMVP candidates may have different reference indices and as a result have different reference frames.

[521] The averaging over Equations 24-31 may produce an unequal result because a frame having a close reference frame may substantially have a greater effect on the H-STMVP motion vector. Accordingly, a method is provided for scaling all motion vectors into one reference frame for equal comparison and inclusion.

[522] In this case, the encoder may determine which one frame, performed as part of the RD optimization, is the most suitable for use as a reference frame. In an embodiment, the selected reference frame may be signaled in the slice header in a manner similar to the TMVP array index present in the slice header. For example, the reference frame to be used may be generated using a fixed rule. Alternatively, for example, the list may be scaled as the first available reference frame from L0, or the list may be scaled based on the current count of the picture sequence.

[523] In one embodiment, to achieve the above goal, the encoder may send, to the decoder, information for one fixed picture using high-level syntax (HLS), which may be part of a sequence parameter set, a picture parameter set, a slice header, a unit coding tree and/or other data header. For example, a high level syntax structure may be defined, such as Table 7 and/or Table 8 below.

[524] Table 7

high_level_parameter_set() { Description … set_HSTMVP_ref_pic_flag u(1) …

[525] With reference to Table 7, set_HSTMVP_ref_pic_flag equal to 1 indicates that set_HSTMVP_idx is present in the header of a non-IDR picture slice in CVS. set_HSTMVP_ref_pic_flag equal to 0 indicates that set_HSTMVP_idx is not present in the slice header.

[526] Table 8

slice_segment_header() { Description … if(set_HSTMVP_ref_pic_flag) slice_HSTMVP_idx u(1) … } …

[527] With reference to Table 8, slice_HMVP_idx indicates a reference index. In one embodiment, the reference index may be selected with respect to the L0 list.

[528]

[529] The above embodiments of the disclosure may be separated and described for convenience of description, but the disclosure is not limited thereto. That is, the embodiments described in Embodiments 1 to 9 may be performed independently, or one or more different embodiments may be combined and performed.

[530] FIG. 47 is a flowchart illustrating an inter-prediction based video signal processing method according to an embodiment to which the disclosure applies.

[531] With reference to FIG. 47, the decoder is described basically for the convenience of description, but the disclosure is not limited thereto. The inter-prediction based video signal processing method according to the embodiment of the disclosure may be identically performed in the encoder and decoder.

[532] The decoder configures the first merge candidate list of the current block using the spatial merge candidate and the temporal merge candidate of the current block (S4701).

[533] The decoder configures the second merge candidate list by adding, to the first merge candidate list, a history-based merge candidate indicating motion information of a block encoded before the current block (S4702).

[534] The decoder obtains a merge index indicating a merge candidate applied to the current block in the second merge candidate list (S4703).

[535] The decoder generates a prediction block of the current block using the merge candidate motion information indicated by the merge index (S4704).

[536] As described above, a history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of the predetermined merge candidate among the merge candidates included in the first merge candidate list.

[537] As described above, a history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of a predetermined specific number of merge candidates among merge candidates included in the first merge candidate list. .

[538] As described above, a history-based merge candidate may be added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of a specific spatial merge candidate included in the first merge candidate list.

[539] As described above, a history-based merge candidate may be derived from a predetermined number of candidates in a history-based buffer that stores history-based merge candidates. For example, the predetermined number could be 2.

[540] As described above, the step of generating the second merge candidate list further includes the step of adding a history-based spatial temporal merge candidate. A history-based spatial temporal candidate can be derived using two history-based merging candidates and a temporal merging candidate.

[541] As described above, the motion vector of the history-based spatial temporal candidate may be derived as the average of the motion vectors of the two history-based merge candidates and the average of the motion vectors of the temporal merge candidates.

[542] FIG. 48 illustrates an exemplary block diagram of an apparatus for processing an image signal in accordance with an embodiment of the disclosure. The image signal processor in FIG. 48 may correspond to the encoder 100 in FIG. 1 or decoder 200 in FIG. 2.

[543] An image processor 4800 for processing an image signal includes a memory 4820 for storing an image signal and a processor 4810 coupled to the memory for processing the image signal.

[544] A processor 4810 according to an embodiment of the disclosure may be configured with at least one processing circuit for processing an image signal, and may process the image signal by executing instructions for encoding or decoding the image signal. That is, the processor 4810 may encode the original image data or decode the encoded image signal by executing the above encoding or decoding methods.

[545] FIG. 49 illustrates a video coding system in which expansion is applied.

[546] The video coding system may include a source device and a reception device. The source device may transmit, to the receiver device, the encoded information or video/image data via a digital storage medium or over a network in a file or stream form.

[547] The source device may include a video source, an encoder, and a transmitter. The receiving device may include a receiver, a decoding device, and a playback module (renderer). The encoding device may be referred to as a video/image encoding device. The decoding device may be referred to as a video/image decoding device. The transmitter may be included in the encoder. The receiver may be included in the decoding device. The rendering device may include a display. The display can be configured for each device or external component.

[548] A video source may acquire a video/image through a video/image capture, synthesis, or generation process. The video source may include a video/image capture device and/or a video/image generation device. The video/image capture device may include one or more cameras, a video/image archive including, for example, previously captured videos/images, and so on. The video/image generating device may include, for example, a computer, a tablet, and a smartphone, and may (electronically) generate a video/image. For example, the virtual video/image may be computer generated. In this case, the related data generation process may be replaced by a video/image capture process.

[549] The encoding device may encode the input video/image. The encoding device may perform a series of procedures such as prediction, transformation, and quantization for compression and encoding efficiency. Encoded data (encoded video/image information) may be output in the form of a bitstream.

[550] The transmitter may transmit, to the receiver of the receiving device, encoded video/image information or data output in the form of a bitstream via a digital storage medium or over a network in a file or stream form. The digital storage media may include various storage media such as USB, SD, CD, DVD, Blueray, HDD, and SSD. The transmitter may include an element for generating a media file with a predefined file format and may include an element for transmission over a broadcast/communications network. The receiver can extract the bit stream and pass it to the decoder.

[551] The decoding device can decode the video/image by performing a series of procedures such as dequantization, inverse transformation, and prediction corresponding to the operations of the encoding device.

[552] The playback device can play the decoded video/image. The played video/image can be displayed on the display.

[553] FIG. 50 is an embodiment in which the disclosure is applied and illustrates the structure of a content streaming system.

[554] With reference to FIG. 50, the content streaming system to which the disclosure applies may mainly include an encoding server, a streaming (streaming) server, a web server, a media storage, a user equipment, and a media input device.

[555] The encoding server basically functions to generate a bitstream by compressing content input from a media input device such as a smartphone, camera, or camcorder into digital data and transmitting the bitstream to the streaming server. As another example, if media input devices such as a smartphone, camera, or camcorder directly generate the bitstream, the encoding server may be omitted.

[556] The bitstream may be generated by an encoding method or a bitstream generation method in which expansion is applied. The streaming server may temporarily store the bitstream while transmitting or receiving the bitstream.

[557] The streaming server transmits media data to the user equipment based on a user request via a web server. The web server acts as a carrier to notify the user which service is being provided. When a user requests a desired service from a web server, the web server forwards the request to the streaming server. The streaming server transmits media data to the user. In this case, the content streaming system may include a separate control server. In this case, the control server functions to manage the instruction/response between devices in the content streaming system.

[558] The streaming server may receive content from storage media and/or an encoding server. For example, if content is received from an encoding server, the streaming server may receive the content in real time. In this case, in order to provide a smooth streaming service, the streaming server may store the bitstream for a predetermined time.

[559] Examples of user equipment may include a mobile phone, smartphone, laptop, digital broadcast terminal, personal digital assistants (PDA), portable multimedia player (PMP), navigator, compact tablet PC, tablet PC, ultrabook, wearable device ( such as watch type terminal (smart watch), glasses type terminal (smart glasses) and head-mounted display (HMD)), digital TV, desktop computer and digital signage.

[560] The servers in the content streaming system may operate as distributed servers. In this case, the data received from the servers can be distributed and processed.

[561] As described above, the embodiments described in the disclosure may be implemented and executed on a processor, microprocessor, controller, or chip. For example, the functional modules illustrated in the drawings may be implemented or executed on a computer, processor, microprocessor, controller, or chip.

[562] Moreover, the decoder and encoder in which the disclosure is applied can be included in a multimedia broadcast transmission and reception device, a mobile communication terminal, a home theater device, a digital cinema device, a monitoring camera, a video dialog device, a communication device in real-time device such as video communication, mobile streaming device, storage media, camcorder, video service on demand (VoD) device, over-the-carrier (OTT) video delivery device from content provider to user, Internet streaming service providing device, 3D ( 3D) video, video telephony device and medical video device and can be used to process a video signal or a data signal. For example, an OTT video device may include a game console, a Blueray player, a TV with Internet access, a home theater, a smartphone, a tablet PC, and a digital video recorder (DVR).

[563] Moreover, the processing method in which the disclosure is applied may be implemented in the form of a program executable by a computer and may be stored in a computer-readable recording medium. Multimedia data having a data structure in accordance with the disclosure may also be stored in a computer-readable recording medium. Computer-readable recording media includes all types of storage devices that store computer-readable data. The computer-readable storage medium may include a Blueray (BD) disk, Universal Serial Bus (USB), ROM, PROM, EPROM, EEPROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical storage, for example. Moreover, a computer-readable recording medium includes media implemented in the form of carriers (eg, transmissions over the Internet). Moreover, the bit stream generated using the encoding method may be stored in a computer-readable recording medium, or may be transmitted over wired and wireless communication networks.

[564] In addition, an embodiment of the disclosure may be implemented as a computer program product using program code. The program code may be executed by a computer in accordance with an embodiment of the disclosure. The program code may be stored on a computer-readable medium.

[565] In the above embodiments, the elements and characteristics of the disclosure have been combined in a specific form. Each of the elements or characteristics can be considered as optional, unless otherwise explicitly described. Each of the elements or characteristics may be implemented in such a way as not to be combined with other elements or characteristics. In addition, some of the elements and/or characteristics may be combined to form an embodiment of the disclosure. The sequence of operations described in the embodiments of the disclosure may vary. Some of the elements or characteristics of an embodiment may be included in another embodiment or may be substituted for corresponding elements or characteristics of another embodiment. Obviously, an embodiment may be created by combining claims that do not have an explicit relationship of reference in the claims, or may be incorporated into a new claim by post-filing changes.

[566] An embodiment in accordance with the disclosure may be implemented by various means, such as in hardware, firmware, software, or a combination thereof. In the case of a hardware implementation, an embodiment of the disclosure may be implemented using one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), field programmable logic devices (PLDs), field programmable gate arrays (FPGAs). ), processors, controllers, microcontrollers, microprocessors, etc.

[567] In the case of implementation through firmware or software, an embodiment of the disclosure may be implemented in the form of a module, procedure, or function to perform the aforementioned functions or operations. The software code may be stored in memory and run by a processor. The memory may be located within or outside the processor and may communicate with the processor through various known means.

[568] Those skilled in the art will appreciate that the disclosure may be embodied in other specific forms without deviating from the essential characteristics of the disclosure. Accordingly, the detailed description is not to be understood as limiting, but is to be understood as illustrative in all aspects. The scope of the disclosure is to be determined by rational analysis of the appended claims, and all changes within an equivalent range of the disclosure are included within the scope of the disclosure.

Industrial Applicability

[569] The foregoing preferred embodiments of the disclosure have been disclosed for illustrative purposes, and various other embodiments may be improved, modified, replaced or added to by those skilled in the art without deviating from the technical spirit and scope of the disclosure set forth in the accompanying claims.

Statement 1. A method for processing a video signal based on inter-prediction, the method comprising:

generating a first merge candidate list of the current block using a spatial merge candidate and a temporal merge candidate of the current block;

generating a second merge candidate list by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list;

obtaining a merge index indicating a merge candidate applied to the current block in the second merge candidate list; and

generating a prediction block of the current block using the motion information of the merge candidate indicated by the merge index,

wherein the history-based merge candidate is added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of the predetermined merge candidate among the merge candidates included in the first merge candidate list.

Statement 2: The method according to Statement 1, wherein a history-based merge candidate is added to the first merge candidate list when the history-based merge candidate has motion information not overlapping the motion information of a predetermined specific number of merge candidates among merge candidates included in the first merge candidate list .

Statement 3: The method according to Statement 1, wherein a history-based merge candidate is added to the first merge candidate list when the history-based merge candidate has motion information that does not overlap the motion information of a specific spatial merge candidate included in the first merge candidate list.

Statement 4: The method according to Statement 1, wherein a history-based merge candidate is derived from a predetermined number of candidates in a history-based buffer storing history-based merge candidates.

Statement 5. Method according to Statement 1,

wherein generating the second list of merge candidates further comprises adding a space-time merge candidate based on history, and

wherein a history-based spatial temporal candidate is derived using two history-based merging candidates and a temporal merging candidate.

Statement 6. The method of Statement 5, wherein the motion vector of the history-based spatial temporal candidate is derived as the average of the motion vectors of the two history-based merge candidates and the mean of the motion vectors of the temporal merge candidates.

Statement 7. An apparatus for processing a video signal based on inter-prediction, the apparatus comprising:

a memory storing the video signal; and

memory-bound processor

wherein the processor is configured to:

generating a first merge candidate list of the current block using a spatial merge candidate and a temporal merge candidate of the current block;

generating a second merge candidate list by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list;

obtaining a merge index indicating a merge candidate applied to the current block in the second merge candidate list; and

generate a prediction block of the current block using the motion information of the merge candidate indicated by the merge index,

wherein the history-based merge candidate is added to the first merge candidate list if the history-based merge candidate has motion information that does not overlap the motion information of the predetermined merge candidate among the merge candidates included in the first merge candidate list.

Statement 8: The device according to Statement 7, wherein a history-based merge candidate is added to the first merge candidate list when the history-based merge candidate has motion information not overlapping the motion information of a predetermined specific number of merge candidates among merge candidates included in the first merge candidate list .

Statement 9: The device according to Statement 7, wherein a history-based merge candidate is added to the first merge candidate list when the history-based merge candidate has motion information that does not overlap the motion information of a specific spatial merge candidate included in the first merge candidate list.

Statement 10: The device according to Statement 7, wherein a history-based merge candidate is derived from a predetermined number of candidates in a history-based buffer storing history-based merge candidates.

Statement 11: The apparatus of Statement 7, wherein the processor is configured to generate a second list of merge candidates further comprises adding a history-based spatial temporal merge candidate, and wherein the history-based spatial temporal candidate is derived using two history-based merge candidates and a temporal candidate associations.

Statement 12: The apparatus according to Statement 11, wherein the motion vector of the history-based spatial temporal candidate is derived as the average of the motion vectors of the two history-based merge candidates and the average of the motion vectors of the temporal merge candidates.

Claims (36)

1. A method for decoding a video signal, comprising the steps of: generating a first merge candidate list for the current block based on a plurality of merge candidates including spatial merge candidates and a temporal merge candidate of the current block, the spatial merge candidates preceding the temporal merge candidate in the first merge candidate list; generating a second merge candidate list by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list; obtaining a merge index indicating a merge candidate applied to the current block in the second merge candidate list; generating a prediction block of the current block based on motion information of the merge candidate indicated by the merge index; outputting a residual block of the current block from the video signal; and generating a reconstructed block of the current block based on the prediction block and the residual block, moreover, the formation of the second list of candidates for the association includes the stages at which: checking the motion information of the merge candidate based on the history with respect to the movement information of only a part of the plurality of merge candidates that constitute the first number of merge candidates from the first merge candidate list, and adding a history-based merge candidate only when said check shows that the history-based merge candidate has motion information that does not overlap with motion information of said part of the set of spatial merge candidates in the first merge candidate list. 2. The method of claim 1, wherein the history-based merge candidate is derived from a predetermined number of candidates in a history-based buffer storing the history-based merge candidates. 3. The method of claim 1, wherein generating the second list of merge candidates further comprises adding a spatial temporal history based merge candidate, wherein the spatial temporal history based merge candidate is derived using two history based merge candidates and a temporal merge candidate. 4. The method of claim 3, wherein the motion vector of the history-based spatial temporal candidate is derived as the average of the motion vectors of the two history-based merge candidates and the average of the motion vectors of the temporal merge candidates. 5. The method of claim 1, wherein generating the second list of merge candidates further comprises determining whether to add the history based merge candidate to the first merge candidate list, starting with the last history based merge candidate added last to the list. history-based merge candidates among the history-based merge candidates included in the list of history-based merge candidates. 6. A method for encoding a video signal, comprising the steps of: generating a first merge candidate list for the current block based on a plurality of merge candidates including spatial merge candidates and a temporal merge candidate of the current block, the spatial merge candidates preceding the temporal merge candidate in the first merge candidate list; generating a second merge candidate list by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list; reporting in the bitstream a merge index indicating a merge candidate applied to the current block in the second merge candidate list; generating a prediction block of the current block based on motion information of the merge candidate indicated by the merge index; and generating a residual block based on the current block and the prediction block, At the same time, the formation of the second list of candidates for the association includes the stages at which: checking the motion information of the merge candidate based on the history with respect to the movement information of only a part of the plurality of merge candidates that constitute the first number of merge candidates from the first merge candidate list, and adding a history-based merge candidate only when said check shows that the history-based merge candidate has motion information that does not overlap with motion information of said part of the set of spatial merge candidates in the first merge candidate list. 7. The method of claim 6, wherein the history-based merge candidate is derived from a predetermined number of candidates in a history-based buffer storing the history-based merge candidates. 8. The method of claim 6, wherein generating the second list of merge candidates further comprises adding a history-based spatial temporal merge candidate, wherein the history-based spatial temporal merge candidate is derived using two history-based merge candidates and a temporal merge candidate. 9. The method of claim 8, wherein the motion vector of the history-based spatial temporal candidate is derived as the average of the motion vector of the temporal candidate and the mean of the motion vectors of the two history-based merge candidates. 10. The method of claim 6, wherein generating the second merge candidate list further comprises determining whether to add the history based merge candidate to the first merge candidate list, starting with the last history based merge candidate added last to the list. history-based merge candidates among the history-based merge candidates included in the list of history-based merge candidates. 11. A computer-readable medium for storing a bit stream generated by the steps: generating a first merge candidate list for the current block based on a plurality of merge candidates including spatial merge candidates and a temporal merge candidate of the current block, the spatial merge candidates preceding the temporal merge candidate in the first merge candidate list; generating a second merge candidate list by adding a history-based merge candidate indicating motion information of a block encoded before the current block to the first merge candidate list; messages in the bitstream of a merge index indicating a merge candidate applied to the current block in the second merge candidate list; generating a prediction block of the current block based on the motion information of the merge candidate indicated by the merge index; and generating a residual block based on the current block and the prediction block, At the same time, the formation of the second list of candidates for the association includes: checking the motion information of the merger candidate based on the history with respect to the motion information of only a part of the plurality of merger candidates that constitute the first number of merger candidates from the first merger candidate list, and adding a history-based merge candidate only when said check shows that the history-based merge candidate has motion information that does not overlap with motion information of said part of the set of spatial merge candidates in the first merge candidate list.

Images