WO2023132556A1 - 영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체 - Google Patents

영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체 Download PDF

Info

Publication number
WO2023132556A1
WO2023132556A1 PCT/KR2022/021532 KR2022021532W WO2023132556A1 WO 2023132556 A1 WO2023132556 A1 WO 2023132556A1 KR 2022021532 W KR2022021532 W KR 2022021532W WO 2023132556 A1 WO2023132556 A1 WO 2023132556A1
Authority
WO
WIPO (PCT)
Prior art keywords
candidate
control point
affine
candidates
block
Prior art date
Application number
PCT/KR2022/021532
Other languages
English (en)
French (fr)
Inventor
박내리
임재현
유선미
남정학
장형문
최장원
Original Assignee
엘지전자 주식회사
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 엘지전자 주식회사 filed Critical 엘지전자 주식회사
Publication of WO2023132556A1 publication Critical patent/WO2023132556A1/ko

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/103Selection of coding mode or of prediction mode
    • H04N19/105Selection of the reference unit for prediction within a chosen coding or prediction mode, e.g. adaptive choice of position and number of pixels used for prediction
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/103Selection of coding mode or of prediction mode
    • H04N19/109Selection of coding mode or of prediction mode among a plurality of temporal predictive coding modes
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/184Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being bits, e.g. of the compressed video stream
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/503Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
    • H04N19/51Motion estimation or motion compensation
    • H04N19/577Motion compensation with bidirectional frame interpolation, i.e. using B-pictures
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/70Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards

Definitions

  • the present invention relates to a video encoding/decoding method and apparatus, and a recording medium storing a bitstream.
  • HD High Definition
  • UHD Ultra High Definition
  • An inter-prediction technique for predicting pixel values included in the current picture from pictures before or after the current picture as an image compression technique an intra-prediction technique for predicting pixel values included in the current picture using pixel information within the current picture, and an appearance frequency.
  • an entropy coding technique in which a short code is assigned to a value with a high frequency of occurrence and a long code is assigned to a value with a low frequency of occurrence.
  • the present disclosure intends to provide a method and apparatus for generating various affine candidates for affine modes.
  • An object of the present disclosure is to provide a method and apparatus for constructing/rearranging a candidate list for an affine mode.
  • An object of the present disclosure is to provide a method and apparatus for constructing a candidate list based on checking duplication between affine candidates.
  • the present disclosure intends to provide a method and apparatus for deriving motion information of an affine candidate.
  • An object of the present disclosure is to provide a method and apparatus for generating a merge candidate for a merge mode.
  • An object of the present disclosure is to provide a method and apparatus for constructing/rearranging a candidate list for merge mode.
  • An image decoding method and apparatus generate a candidate list of a current block, add an average candidate to the candidate list, and perform inter prediction of the current block based on the candidate list including the average candidate.
  • the candidate list includes a plurality of affine candidates, and the plurality of affine candidates may include at least one of an inherited candidate, a configured candidate, and a history-based candidate.
  • the average candidate may be generated based on two affine candidates selected from the candidate list in a predetermined order of priority.
  • the type of the affine model for the average candidate is the selected two affine candidates. It can be determined as the type of an affine model for any one of
  • the image decoding method and apparatus may rearrange a plurality of affine candidates belonging to the candidate list in consideration of the type of an affine model for at least one of the plurality of affine candidates belonging to the candidate list. there is.
  • the average candidate is determined by the availability of the two affine candidates, whether reference picture indices between the two affine candidates are the same, or a reference between the two affine candidates.
  • Pictures may be adaptively added to the candidate list based on at least one of whether the pictures are identical or not.
  • the inherited candidates include whether a spatial neighboring block of the current block is a block encoded in an affine mode, the number of inherited candidates previously included in the candidate list, Alternatively, the candidate list may be configured in consideration of at least one of the shapes of the current block.
  • the average candidate is added to the candidate list based on a duplication check with at least one of a plurality of affine candidates belonging to the candidate list, and the duplication check comprises: It can be performed based on the type of affine model for the mean candidate.
  • rearrangement may be performed on at least one of the inherited candidate, the configured candidate, and the average candidate included in the candidate list based on a template-based cost. there is.
  • the motion information of the average candidate is determined based on motion information of at least one of the two selected affine candidates, and the motion information determines whether local illumination compensation is applied. It may include at least one of a flag indicating whether or not or weight information for bi-directional weighted prediction.
  • the video decoding method and apparatus may obtain a candidate index of the current block from a bitstream, and the candidate index specifies one of some affine candidates in the candidate list including the average candidate. can do.
  • the candidate list further includes a combined candidate
  • the combined candidate includes motion information in an L0 prediction direction for any one of the plurality of affine candidates and the It may be generated by combining motion information in the L1 prediction direction for another one of a plurality of affine candidates.
  • An image encoding method and apparatus generate a candidate list of a current block, add an average candidate to the candidate list, and perform inter prediction of the current block based on the candidate list including the average candidate.
  • the candidate list includes a plurality of affine candidates, and the plurality of affine candidates may include at least one of an inherited candidate, a configured candidate, and a history-based candidate.
  • the average candidate may be generated based on two affine candidates selected from the candidate list in a predetermined order of priority.
  • a computer-readable digital storage medium in which encoded video/image information that causes an image decoding method to be performed by a decoding device according to the present disclosure is stored.
  • a computer-readable digital storage medium in which video/image information generated by the video encoding method according to the present disclosure is stored is provided.
  • a method and apparatus for transmitting video/image information generated by the video encoding method according to the present disclosure are provided.
  • accuracy and reliability of inter prediction may be improved by generating various affine candidates or merge candidates.
  • the candidate index encoding efficiency can be improved by constructing a candidate list through redundancy check and rearranging the candidate list.
  • the present disclosure by determining a target candidate for generation of an average candidate or combined candidate in consideration of the type of an affine model, reliability of the average candidate or combined candidate may be increased and accuracy of inter prediction may be improved.
  • FIG. 1 illustrates a video/image coding system according to the present disclosure.
  • FIG. 2 shows a schematic block diagram of an encoding device to which an embodiment of the present disclosure may be applied and encoding of a video/video signal is performed.
  • FIG. 3 is a schematic block diagram of a decoding device to which an embodiment of the present disclosure may be applied and decoding of a video/image signal is performed.
  • FIG. 4 illustrates an affine model-based inter prediction method performed by a decoding apparatus as an embodiment according to the present disclosure.
  • FIG. 5 relates to a method of deriving an affine candidate from a control point vector of a spatial/temporal neighboring block as an embodiment according to the present disclosure.
  • FIG. 6 illustrates a method of deriving a candidate constructed based on a combination of motion vectors of spatial/temporal neighboring blocks as an embodiment according to the present disclosure.
  • FIG. 7 illustrates a method of deriving a motion vector in units of sub-blocks of a current block as an embodiment of the present disclosure.
  • FIG 8 illustrates a schematic configuration of an inter prediction unit 332 performing inter prediction based on an affine model as an embodiment according to the present disclosure.
  • FIG 9 illustrates an affine model-based inter prediction method performed by an encoding apparatus as an embodiment according to the present disclosure.
  • FIG. 10 illustrates a schematic configuration of an inter prediction unit 221 performing inter prediction based on an affine model as an embodiment according to the present disclosure.
  • FIG. 11 shows an example of a content streaming system to which embodiments of the present disclosure may be applied.
  • first and second may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present disclosure.
  • the terms and/or include any combination of a plurality of related recited items or any of a plurality of related recited items.
  • This disclosure relates to video/image coding.
  • the method/embodiment disclosed herein may be applied to a method disclosed in a versatile video coding (VVC) standard.
  • VVC versatile video coding
  • the method/embodiment disclosed in this specification is an essential video coding (EVC) standard, an AOMedia Video 1 (AV1) standard, a 2nd generation of audio video coding standard (AVS2), or a next-generation video/video coding standard (ex. H.267 or H.268, etc.).
  • EVC essential video coding
  • AV1 AOMedia Video 1
  • AVS2 2nd generation of audio video coding standard
  • next-generation video/video coding standard ex. H.267 or H.268, etc.
  • a video may mean a set of a series of images over time.
  • a picture generally means a unit representing one image in a specific time period
  • a slice/tile is a unit constituting a part of a picture in coding.
  • a slice/tile may include one or more coding tree units (CTUs).
  • CTUs coding tree units
  • One picture may consist of one or more slices/tiles.
  • One tile is a rectangular area composed of a plurality of CTUs in a specific tile column and a specific tile row of one picture.
  • a tile column is a rectangular area of CTUs with a height equal to that of the picture and a width specified by the syntax request of the picture parameter set.
  • a tile row is a rectangular area of CTUs with a height specified by the picture parameter set and a width equal to the width of the picture.
  • CTUs within one tile are consecutively arranged according to the CTU raster scan, whereas tiles within one picture may be consecutively arranged according to the raster scan of the tile.
  • One slice may contain an integer number of complete tiles or an integer number of contiguous complete CTU rows within a tile of a picture that may be exclusively included in a single NAL unit. Meanwhile, one picture may be divided into two or more subpictures.
  • a subpicture can be a rectangular area of one or more slices within a picture.
  • a pixel, pixel, or pel may mean a minimum unit constituting one picture (or image). Also, 'sample' may be used as a term corresponding to a pixel.
  • a sample may generally represent a pixel or a pixel value, may represent only a pixel/pixel value of a luma component, or only a pixel/pixel value of a chroma component.
  • a unit may represent a basic unit of image processing.
  • a unit may include at least one of a specific region of a picture and information related to the region.
  • One unit may include one luma block and two chroma (eg cb, cr) blocks. Unit may be used interchangeably with terms such as block or area depending on the case.
  • an MxN block may include samples (or a sample array) or a set (or array) of transform coefficients consisting of M columns and N rows.
  • a or B may mean “only A”, “only B”, or “both A and B”.
  • a or B (A or B)" in the present specification may be interpreted as “A and/or B (A and/or B)”.
  • A, B or C as used herein means “only A”, “only B”, “only C”, or “any and all combinations of A, B and C ( any combination of A, B and C)”.
  • a slash (/) or a comma (comma) used in this specification may mean “and/or”.
  • A/B can mean “A and/or B”. Accordingly, “A/B” may mean “only A”, “only B”, or “both A and B”.
  • A, B, C may mean “A, B or C”.
  • At least one of A and B may mean “only A”, “only B”, or “both A and B”. Also, in this specification, the expression “at least one of A or B” or “at least one of A and/or B” means “at least one It can be interpreted the same as "A and B (at least one of A and B) of
  • At least one of A, B and C means “only A”, “only B”, “only C", or “A, B and C” It may mean “any combination of A, B and C”. Also, “at least one of A, B or C” or “at least one of A, B and/or C” means It can mean “at least one of A, B and C”.
  • parentheses used in this specification may mean “for example”. Specifically, when it is indicated as “prediction (intra prediction)”, “intra prediction” may be suggested as an example of “prediction”. In other words, “prediction” in this specification is not limited to “intra prediction”, and “intra prediction” may be suggested as an example of “prediction”. Also, even when indicated as “prediction (ie, intra prediction)”, “intra prediction” may be suggested as an example of “prediction”.
  • FIG. 1 illustrates a video/image coding system according to the present disclosure.
  • a video/image coding system may include a first device (source device) and a second device (receive device).
  • the source device may transmit encoded video/image information or data to a receiving device in a file or streaming form through a digital storage medium or network.
  • the source device may include a video source, an encoding device, and a transmission unit.
  • the receiving device may include a receiving unit, a decoding device, and a renderer.
  • the encoding device may be referred to as a video/image encoding device, and the decoding device may be referred to as a video/image decoding device.
  • a transmitter may be included in an encoding device.
  • a receiver may be included in a decoding device.
  • the renderer may include a display unit, and the display unit may be configured as a separate device or an external component.
  • a video source may acquire video/images through a process of capturing, synthesizing, or generating video/images.
  • a video source may include a video/image capture device and/or a video/image generation device.
  • a video/image capture device may include one or more cameras, a video/image archive containing previously captured video/images, and the like.
  • Video/image generating devices may include computers, tablets and smart phones, etc., and may (electronically) generate video/images.
  • a virtual video/image may be generated through a computer or the like, and in this case, a video/image capture process may be replaced by a process of generating related data.
  • An encoding device may encode an input video/picture.
  • the encoding device may perform a series of procedures such as prediction, transformation, and quantization for compression and coding efficiency.
  • Encoded data (encoded video/video information) may be output in the form of a bitstream.
  • the transmission unit may transmit the encoded video/image information or data output in the form of a bit stream to the receiving unit of the receiving device in the form of a file or streaming through a digital storage medium or a network.
  • Digital storage media may include various storage media such as USB, SD, CD, DVD, Blu-ray, HDD, and SSD.
  • the transmission unit may include an element for generating a media file through a predetermined file format, and may include an element for transmission through a broadcasting/communication network.
  • the receiving unit may receive/extract the bitstream and transmit it to a decoding device.
  • the decoding device may decode video/images by performing a series of procedures such as inverse quantization, inverse transformation, and prediction corresponding to operations of the encoding device.
  • the renderer may render the decoded video/image.
  • the rendered video/image may be displayed through the display unit.
  • FIG. 2 shows a schematic block diagram of an encoding device to which an embodiment of the present disclosure may be applied and encoding of a video/video signal is performed.
  • the encoding device 200 includes an image partitioner 210, a predictor 220, a residual processor 230, an entropy encoder 240, It may include an adder 250, a filter 260, and a memory 270.
  • the prediction unit 220 may include an inter prediction unit 221 and an intra prediction unit 222 .
  • the residual processing unit 230 may include a transformer 232 , a quantizer 233 , a dequantizer 234 , and an inverse transformer 235 .
  • the residual processing unit 230 may further include a subtractor 231 .
  • the adder 250 may be called a reconstructor or a reconstructed block generator.
  • the above-described image segmentation unit 210, prediction unit 220, residual processing unit 230, entropy encoding unit 240, adder 250, and filtering unit 260 may be one or more hardware components ( For example, it may be configured by an encoding device chipset or processor).
  • the memory 270 may include a decoded picture buffer (DPB) and may be configured by a digital storage medium.
  • the hardware component may further include a memory 270 as an internal/external component.
  • the image divider 210 may divide an input image (or picture or frame) input to the encoding device 200 into one or more processing units.
  • the processing unit may be called a coding unit (CU).
  • the coding unit may be partitioned recursively from a coding tree unit (CTU) or a largest coding unit (LCU) according to a quad-tree binary-tree ternary-tree (QTBTTT) structure.
  • CTU coding tree unit
  • LCU largest coding unit
  • QTBTTT quad-tree binary-tree ternary-tree
  • one coding unit may be divided into a plurality of coding units having a deeper depth based on a quad tree structure, a binary tree structure, and/or a ternary structure.
  • a quad tree structure may be applied first and a binary tree structure and/or ternary structure may be applied later.
  • the binary tree structure may be applied before the quad tree structure.
  • a coding procedure according to the present specification may be performed based on a final coding unit that is not further divided.
  • the largest coding unit can be directly used as the final coding unit, or the coding unit is recursively divided into coding units of lower depths as needed, A coding unit having a size of may be used as a final coding unit.
  • the coding procedure may include procedures such as prediction, transformation, and restoration to be described later.
  • the processing unit may further include a prediction unit (PU) or a transform unit (TU).
  • the prediction unit and the transform unit may be divided or partitioned from the above-described final coding unit, respectively.
  • the prediction unit may be a unit of sample prediction
  • the transform unit may be a unit for deriving transform coefficients and/or a unit for deriving a residual signal from transform coefficients.
  • an MxN block may represent a set of samples or transform coefficients consisting of M columns and N rows.
  • a sample may generally represent a pixel or a pixel value, may represent only a pixel/pixel value of a luma component, or only a pixel/pixel value of a chroma component.
  • a sample may be used as a term corresponding to one picture (or image) to a pixel or pel.
  • the encoding device 200 subtracts the prediction signal (prediction block, prediction sample array) output from the inter prediction unit 221 or the intra prediction unit 222 from the input video signal (original block, original sample array) to obtain a residual signal (residual signal, residual block, residual sample array) may be generated, and the generated residual signal is transmitted to the conversion unit 232.
  • a unit for subtracting a prediction signal (prediction block, prediction sample array) from an input video signal (original block, original sample array) in the encoding device 200 may be called a subtraction unit 231 .
  • the prediction unit 220 may perform prediction on a block to be processed (hereinafter referred to as a current block) and generate a predicted block including predicted samples of the current block.
  • the predictor 220 may determine whether intra prediction or inter prediction is applied in units of current blocks or CUs.
  • the prediction unit 220 may generate and transmit various types of information related to prediction, such as prediction mode information, to the entropy encoding unit 240, as will be described later in the description of each prediction mode. Prediction-related information may be encoded in the entropy encoding unit 240 and output in the form of a bitstream.
  • the intra predictor 222 may predict a current block by referring to samples in the current picture.
  • the referenced samples may be located in the neighborhood of the current block or may be located apart from the current block by a predetermined distance according to a prediction mode.
  • prediction modes may include one or more non-directional modes and a plurality of directional modes.
  • the non-directional mode may include at least one of a DC mode and a planar mode.
  • the directional mode may include 33 directional modes or 65 directional modes according to the degree of detail of the prediction direction. However, this is an example, and more or less directional modes may be used according to settings.
  • the intra predictor 222 may determine a prediction mode applied to the current block by using a prediction mode applied to neighboring blocks.
  • the inter-prediction unit 221 may derive a prediction block for a current block based on a reference block (reference sample array) specified by a motion vector on a reference picture.
  • motion information may be predicted in units of blocks, subblocks, or samples based on correlation of motion information between neighboring blocks 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, etc.).
  • a neighboring block may include a spatial neighboring block present in the current picture and a temporal neighboring block present in the reference picture.
  • a reference picture including the reference block and a reference picture including the temporal neighboring block may be the same or different.
  • the temporal neighboring block may be called a collocated reference block, a collocated CU (colCU), and the like, and a reference picture including the temporal neighboring block may be called a collocated picture (colPic).
  • the inter-prediction unit 221 constructs a motion information candidate list based on neighboring blocks, and provides information indicating which candidate is used to derive the motion vector and/or reference picture index of the current block. can create Inter prediction may be performed based on various prediction modes. For example, in the case of skip mode and merge mode, the inter prediction unit 221 may use motion information of neighboring blocks as motion information of the current block. In the case of the skip mode, the residual signal may not be transmitted unlike the merge mode. In the case of motion vector prediction (MVP) mode, the motion vector of the current block is used as a motion vector predictor and the motion vector difference is signaled. can be instructed.
  • MVP motion
  • the prediction unit 220 may generate a prediction signal based on various prediction methods described later.
  • the predictor may apply intra-prediction or inter-prediction to predict one block, as well as apply intra-prediction and inter-prediction at the same time. This may be called a combined inter and intra prediction (CIIP) mode.
  • the prediction unit may be based on an intra block copy (IBC) prediction mode or a palette mode for block prediction.
  • IBC intra block copy
  • the IBC prediction mode or the palette mode can be used for video/video coding of content such as games, such as screen content coding (SCC).
  • IBC basically performs prediction within the current picture, but may be performed similarly to inter prediction in that a reference block is derived within the current picture.
  • IBC may use at least one of the inter prediction techniques described in this specification.
  • Palette mode can be viewed as an example of intra coding or intra prediction.
  • a sample value within a picture may be signaled based on information about a palette table and a palette index.
  • the prediction signal generated by the prediction unit 220 may be used to generate a restored signal or a residual signal.
  • the transform unit 232 may generate transform coefficients by applying a transform technique to the residual signal.
  • the transform technique uses 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).
  • DCT Discrete Cosine Transform
  • DST Discrete Sine Transform
  • KLT Karhunen-Loeve Transform
  • GBT Graph-Based Transform
  • CNT Conditionally Non-linear Transform
  • GBT means a transformation obtained from the graph when relation information between pixels is expressed as a graph.
  • CNT means a transformation obtained by generating a prediction signal using all previously reconstructed pixels, and based thereon.
  • the conversion process may be applied to square pixel blocks having the same size, or may be applied to non-square blocks of variable size.
  • the quantization unit 233 quantizes the transform coefficients and transmits them to the entropy encoding unit 240, and the entropy encoding unit 240 may encode the quantized signal (information on the quantized transform coefficients) and output it as a bitstream. there is. Information about the quantized transform coefficients may be referred to as residual information.
  • the quantization unit 233 may rearrange block-type quantized transform coefficients into a one-dimensional vector form based on a coefficient scan order, and the quantization may be performed based on the quantized transform coefficients of the one-dimensional vector form. Information on the transformed transform coefficients may be generated.
  • the entropy encoding unit 240 may perform various encoding methods such as exponential Golomb, context-adaptive variable length coding (CAVLC), and context-adaptive binary arithmetic coding (CABAC).
  • the entropy encoding unit 240 may encode together or separately information necessary for video/image reconstruction (eg, values of syntax elements, etc.) in addition to quantized transform coefficients.
  • Encoded information may be transmitted or stored in a network abstraction layer (NAL) unit unit in the form of a bitstream.
  • the video/video information may further include information on various parameter sets such as an adaptation parameter set (APS), a picture parameter set (PPS), a sequence parameter set (SPS), or a video parameter set (VPS).
  • the video/image information may further include general constraint information.
  • information and/or syntax elements transmitted/signaled from an encoding device to a decoding device may be included in video/image information.
  • the video/image information may be encoded through the above-described encoding procedure and included in the bitstream.
  • the bitstream may be transmitted through a network or stored in a digital storage medium.
  • the network may include a broadcasting network and/or a communication network
  • the digital storage medium may include various storage media such as USB, SD, CD, DVD, Blu-ray, HDD, and SSD.
  • a transmission unit (not shown) for transmitting the signal output from the entropy encoding unit 240 and/or a storage unit (not shown) for storing may be configured as internal/external elements of the encoding device 200, or the transmission unit It may also be included in the entropy encoding unit 240.
  • the quantized transform coefficients output from the quantization unit 233 may be used to generate a prediction signal.
  • a residual signal residual block or residual samples
  • the adder 250 adds the reconstructed residual signal to the prediction signal output from the inter predictor 221 or the intra predictor 222 to obtain a reconstructed signal (reconstructed picture, reconstructed block, reconstructed sample array) can be created
  • a predicted block may be used as a reconstruction block.
  • the adder 250 may be called a restoration unit or a restoration block generation unit.
  • the generated reconstructed signal may be used for intra prediction of a block to be processed next in the current picture, or may be used for inter prediction of the next picture after filtering as described below. Meanwhile, luma mapping with chroma scaling (LMCS) may be applied in a picture encoding and/or reconstruction process.
  • LMCS luma mapping with chroma scaling
  • the filtering unit 260 may improve subjective/objective picture quality by applying filtering to the reconstructed signal. For example, the filtering unit 260 may generate a modified reconstructed picture by applying various filtering methods to the reconstructed picture, and store the modified reconstructed picture in the memory 270, specifically the memory 270. It can be stored in DPB.
  • the various filtering methods may include deblocking filtering, sample adaptive offset, adaptive loop filter, bilateral filter, and the like.
  • the filtering unit 260 may generate various types of filtering-related information and transmit them to the entropy encoding unit 240 . Filtering-related information may be encoded in the entropy encoding unit 240 and output in the form of a bit stream.
  • the modified reconstructed picture transmitted to the memory 270 may be used as a reference picture in the inter prediction unit 221 .
  • the encoding device can avoid prediction mismatch between the encoding device 200 and the decoding device, and can also improve encoding efficiency.
  • the DPB of the memory 270 may store the modified reconstructed picture to be used as a reference picture in the inter prediction unit 221 .
  • the memory 270 may store motion information of a block in a current picture from which motion information is derived (or encoded) and/or motion information of blocks in a previously reconstructed picture.
  • the stored motion information may be transmitted to the inter prediction unit 221 to be used as motion information of a spatial neighboring block or motion information of a temporal neighboring block.
  • the memory 270 may store reconstructed samples of reconstructed blocks in the current picture and transfer them to the intra predictor 222 .
  • FIG. 3 is a schematic block diagram of a decoding device to which an embodiment of the present disclosure may be applied and decoding of a video/image signal is performed.
  • the decoding device 300 includes an entropy decoder 310, a residual processor 320, a predictor 330, an adder 340, and a filtering unit. (filter, 350) and memory (memoery, 360).
  • the prediction unit 330 may include an inter prediction unit 331 and an intra prediction unit 332 .
  • the residual processing unit 320 may include a dequantizer 321 and an inverse transformer 321 .
  • the aforementioned entropy decoding unit 310, residual processing unit 320, prediction unit 330, adder 340, and filtering unit 350 may be configured as one hardware component (for example, a decoding device chipset or processor).
  • the memory 360 may include a decoded picture buffer (DPB) and may be configured by a digital storage medium.
  • the hardware component may further include a memory 360 as an internal/external component.
  • the decoding device 300 may restore an image corresponding to a process in which the video/image information is processed by the encoding device of FIG. 2 .
  • the decoding device 300 may derive units/blocks based on block division related information obtained from the bitstream.
  • the decoding device 300 may perform decoding using a processing unit applied in the encoding device.
  • a processing unit of decoding may be a coding unit, and a coding unit may be one divided from a coding tree unit or a largest coding unit according to a quad tree structure, a binary tree structure, and/or a ternary tree structure.
  • One or more transform units may be derived from a coding unit.
  • the restored video signal decoded and output through the decoding device 300 may be reproduced through a playback device.
  • the decoding device 300 may receive a signal output from the encoding device of FIG. 2 in the form of a bitstream, and the received signal may be decoded through the entropy decoding unit 310 .
  • the entropy decoding unit 310 may parse the bitstream to derive information (eg, video/image information) required for image restoration (or picture restoration).
  • the video/video information may further include information on various parameter sets such as an adaptation parameter set (APS), a picture parameter set (PPS), a sequence parameter set (SPS), or a video parameter set (VPS).
  • the video/image information may further include general constraint information.
  • the decoding device may decode a picture further based on the information about the parameter set and/or the general restriction information.
  • Signaling/received information and/or syntax elements described later in this specification may be obtained from the bitstream by being decoded through the decoding procedure.
  • the entropy decoding unit 310 decodes information in a bitstream based on a coding method such as exponential Golomb coding, CAVLC, or CABAC, and values of syntax elements required for image reconstruction and quantized values of residual transform coefficients. can output them.
  • the CABAC entropy decoding method receives bins corresponding to each syntax element in a bitstream, and converts syntax element information to be decoded and decoding information of neighboring and decoding object blocks or symbol/bin information decoded in a previous step.
  • a symbol corresponding to the value of each syntax element can be generated by determining a context model, predicting the probability of occurrence of a bin according to the determined context model, and performing arithmetic decoding of the bin.
  • the CABAC entropy decoding method may update the context model by using information of the decoded symbol/bin for the context model of the next symbol/bin after determining the context model.
  • prediction-related information is provided to the prediction unit (inter prediction unit 332 and intra prediction unit 331), and entropy decoding is performed by the entropy decoding unit 310.
  • Dual values that is, quantized transform coefficients and related parameter information may be input to the residual processing unit 320 .
  • the residual processor 320 may derive a residual signal (residual block, residual samples, residual sample array). Also, among information decoded by the entropy decoding unit 310 , information about filtering may be provided to the filtering unit 350 . Meanwhile, a receiving unit (not shown) receiving a signal output from the encoding device may be further configured as an internal/external element of the decoding device 300, or the receiving unit may be a component of the entropy decoding unit 310.
  • the decoding device may be referred to as a video/video/picture decoding device, and the decoding device includes an information decoding device (video/video/picture information decoding device) and a sample decoding device (video/video/picture sample decoding). devices) can be distinguished.
  • the information decoding device may include the entropy decoding unit 310, and the sample decoding device may include the inverse quantization unit 321, an inverse transform unit 322, an adder 340, a filtering unit 350, a memory 360, an inter predictor 332, and an intra predictor 331 may be included.
  • the inverse quantization unit 321 may inversely quantize the quantized transform coefficients and output transform coefficients.
  • the inverse quantization unit 321 may rearrange the quantized transform coefficients in a 2D block form. In this case, the rearrangement may be performed based on a coefficient scanning order performed by the encoding device.
  • the inverse quantization unit 321 may perform inverse quantization on quantized transform coefficients using a quantization parameter (eg, quantization step size information) and obtain transform coefficients.
  • a quantization parameter eg, quantization step size information
  • a residual signal (residual block, residual sample array) is obtained by inverse transforming the transform coefficients.
  • the predictor 320 may perform prediction on a current block and generate a predicted block including predicted samples of the current block.
  • the prediction unit 320 may determine whether intra prediction or inter prediction is applied to the current block based on the information about the prediction output from the entropy decoding unit 310, and determine a specific intra/inter prediction mode.
  • the prediction unit 320 may generate a prediction signal based on various prediction methods described later.
  • the predictor 320 may apply intra-prediction or inter-prediction to predict one block, and may simultaneously apply intra-prediction and inter-prediction. This may be called a combined inter and intra prediction (CIIP) mode.
  • the prediction unit may be based on an intra block copy (IBC) prediction mode or a palette mode for block prediction.
  • IBC intra block copy
  • the IBC prediction mode or the palette mode can be used for video/video coding of content such as games, such as screen content coding (SCC).
  • SCC screen content coding
  • IBC basically performs prediction within the current picture, but may be performed similarly to inter prediction in that a reference block is derived within the current picture. That is, IBC may use at least one of the inter prediction techniques described in this specification.
  • Palette mode can be viewed as an example of intra coding or intra prediction. When the palette mode is applied, information on a palette table and a palette index may be included in the video/
  • the intra predictor 331 may predict a current block by referring to samples in the current picture.
  • the referenced samples may be located in the neighborhood of the current block or may be located apart from the current block by a predetermined distance according to a prediction mode.
  • prediction modes may include one or more non-directional modes and a plurality of directional modes.
  • the intra prediction unit 331 may determine a prediction mode applied to the current block by using a prediction mode applied to neighboring blocks.
  • the inter-prediction unit 332 may derive a prediction block for a current block based on a reference block (reference sample array) specified by a motion vector on a reference picture.
  • motion information may be predicted in units of blocks, subblocks, or samples based on correlation of motion information between neighboring blocks 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, etc.).
  • a neighboring block may include a spatial neighboring block present in the current picture and a temporal neighboring block present in the reference picture.
  • the inter-prediction unit 332 may construct a motion information candidate list based on neighboring blocks and derive a motion vector and/or reference picture index of the current block based on the received candidate selection information. Inter prediction may be performed based on various prediction modes, and the prediction-related information may include information indicating an inter prediction mode for the current block.
  • the adder 340 adds the obtained residual signal to the prediction signal (prediction block, prediction sample array) output from the prediction unit (including the inter prediction unit 332 and/or the intra prediction unit 331) to obtain a reconstructed signal (reconstructed picture, reconstructed block, reconstructed sample array) can be created.
  • a prediction block may be used as a reconstruction block.
  • the adder 340 may be called a restoration unit or a restoration block generation unit.
  • the generated reconstructed signal may be used for intra prediction of the next block to be processed in the current picture, output after filtering as described below, or may be used for inter prediction of the next picture. Meanwhile, luma mapping with chroma scaling (LMCS) may be applied in a picture decoding process.
  • LMCS luma mapping with chroma scaling
  • the filtering unit 350 may improve subjective/objective picture quality by applying filtering to the reconstructed signal.
  • the filtering unit 350 may generate a modified reconstructed picture by applying various filtering methods to the reconstructed picture, and store the modified reconstructed picture in the memory 360, specifically the DPB of the memory 360. can be sent to
  • the various filtering methods may include deblocking filtering, sample adaptive offset, adaptive loop filter, bilateral filter, and the like.
  • a (modified) reconstructed picture stored in the DPB of the memory 360 may be used as a reference picture in the inter prediction unit 332 .
  • the memory 360 may store motion information of a block in the current picture from which motion information is derived (or decoded) and/or motion information of blocks in a previously reconstructed picture.
  • the stored motion information may be transmitted to the inter prediction unit 260 to be used as motion information of a spatial neighboring block or motion information of a temporal neighboring block.
  • the memory 360 may store reconstructed samples of reconstructed blocks in the current picture and transfer them to the intra prediction unit 331 .
  • the embodiments described in the filtering unit 260, the inter prediction unit 221, and the intra prediction unit 222 of the encoding device 200 are the filtering unit 350 and the inter prediction of the decoding device 300, respectively.
  • the same or corresponding to the unit 332 and the intra predictor 331 may be applied.
  • FIG. 4 illustrates an affine model-based inter prediction method performed by a decoding apparatus as an embodiment according to the present disclosure.
  • a candidate list for predicting motion information of a current block may be generated (S400).
  • the candidate list may be for an affine merge mode in inter prediction using an affine model. However, it is not limited thereto, and the candidate list may be for an affine inter mode.
  • an affine mode may be interpreted as an affine merge mode or an affine inter mode.
  • the motion information may include at least one of a motion vector, a reference picture index, inter-prediction direction information, weight information for bidirectional weighted prediction, and a flag indicating whether local illumination compensation is applied.
  • the candidate list may include one or more affine model-based candidates (hereinafter referred to as affine candidates).
  • An affine candidate may mean a candidate having a control point vector.
  • the control point vector means a motion vector of a control point for an affine model, and may be defined for a corner position of a block (eg, a position of at least one of upper left corner, upper right corner, lower left corner, or lower right corner). .
  • An affine candidate may include at least one of a spatial candidate, a temporal candidate, and a constructed candidate.
  • the spatial candidate may be derived from a vector of neighboring blocks spatially adjacent to the current block (hereinafter referred to as spatial neighboring blocks), and the temporal candidate may be derived from a vector of neighboring blocks temporally adjacent to the current block (hereinafter, referred to as temporal neighboring blocks). It can be derived from the vector of).
  • the neighboring block may mean a block encoded in an affine mode.
  • the vector may mean a control point vector or a motion vector.
  • a method of deriving a spatial/temporal candidate based on a vector of a spatial/temporal neighboring block will be described in detail with reference to FIG. 5 .
  • the configured candidate may be derived based on a combination of motion vectors of spatial/temporal neighboring blocks of the current block, where the neighboring blocks may not be blocks encoded in an affine mode.
  • the neighboring blocks may not be blocks encoded in an affine mode.
  • at least one neighboring block may be a block encoded in an affine mode. This will be examined in detail with reference to FIG. 6 .
  • the affine candidate may further include a history-based candidate.
  • Motion information of a block decoded before the current block (hereinafter, referred to as a previous block) may be sequentially stored in a buffer having a predetermined size according to a decoding order.
  • the motion information of the previous block may be used as a candidate for deriving the motion information of the current block, and a candidate having motion information of the previous block in this way is referred to as a history-based candidate.
  • the previous block may be a neighboring block adjacent to the current block or a neighboring block not adjacent to the current block.
  • the affine candidate may further include an average candidate.
  • the number of average candidates that can be generated may be determined based on at least one of the maximum number of affine candidates (MaxNumSubblockMergeCand) that can be included in the candidate list or the number of affine candidates previously included in the candidate list.
  • the maximum number may be a fixed number equally defined for the encoding device and the decoding device.
  • the maximum number may be variably determined based on the size of the candidate list or information indicating the maximum number of affine candidates that can be included in the candidate list.
  • the information indicating the maximum number may be signaled at the level of at least one of a sequence parameter set, a picture parameter set, a picture header, a slice header, and a coding block.
  • the average number of candidates that can be included in the candidate list is Min(N*(N-1) /2, MaxNumSubblockMergeCand-N).
  • the average number of candidates that can be included in the candidate list may be derived as Min(N*(N-1)/2, MaxNumSubblockMergeCand-N-P).
  • P may be an integer of 0, 1, 2, 3 or more.
  • P may be a fixed value identically defined in the encoding device and the decoding device, or may be variably determined based on the number of candidates not generated using the affine model among the pre-included candidates in the candidate list.
  • An average candidate may be generated based on two candidates in the candidate list, that is, Cand0 and Cand1.
  • Cand0 may be one of the candidates pre-included in the candidate list
  • Cand1 may be any other one of the pre-included candidates in the candidate list.
  • Cand0 and Cand1 may be determined in consideration of a predetermined priority order. As an example, the order of priority is shown in Table 1 below, which shows a combination of Cando and Candl to generate 10 average candidates as an example.
  • Idx represents a priority order, and the smaller the Idx, the higher the priority order.
  • Cand0Idx and Cand1Idx may each mean a position of a candidate used to generate an average candidate within the candidate list. For example, in the order of 0 to 9, an average candidate may be generated based on Cand0 and Cand1 corresponding to the corresponding Idx.
  • an average candidate may be generated based on a candidate (Cand0) having an index of 0 in the candidate list and a candidate (Cand1) having an index of 1 in the candidate list.
  • the two target candidates can be constrained to be selected from among inherited candidates, constructed candidates, and history-based candidates.
  • the inherited candidate may include at least one of the aforementioned spatial candidate and temporal candidate.
  • an average candidate may be generated based on two target candidates of different types.
  • one type of the two target candidates may be an inherited candidate and the other type may be a configured candidate.
  • one type of the two target candidates may be a configured candidate, and the other type may be a history-based candidate.
  • it may be constrained to generate an average candidate based on two target candidates of the same type. That is, the average candidate may be generated based on two target candidates, which are inherited candidates, or based on two target candidates, which are configured candidates.
  • the average candidate can be constrained to be generated based on the two target candidates being history-based candidates.
  • the two target candidates may be restricted so that they are selected within the inherited and configured candidates, and not selected within the history-based candidates.
  • the average candidate may be generated based on two target candidates of different types, or the average candidate may be generated based on two target candidates of the same type.
  • the two target candidates may be restricted so that they are selected from the inherited and history-based candidates, and not selected from the configured candidates.
  • the average candidate may be generated based on two target candidates of different types, or the average candidate may be generated based on two target candidates of the same type.
  • the two target candidates may be restricted to be selected only within any one type of an inherited candidate, a configured candidate, or a history-based candidate.
  • the two target candidates may be restricted to being selected only within inherited candidates.
  • the two target candidates may be restricted to being selected only from within the configured candidates.
  • the two target candidates may be limited to being selected only from among history-based candidates.
  • the type of one affine model of the two target blocks is 4-parameter and the type of the other affine model is 6-parameter.
  • a method of generating an average candidate will be described.
  • both of the two target candidates are 4-parameter
  • the first control point vectors (cpMV[0]) and the second control point vectors (cpMV[1]) of the two target candidates are averaged, respectively, and the 4-parameter
  • the type of the affine model can be stored/propagated.
  • the first control point vectors (cpMV[0]), the second control point vectors (cpMV[1]), and the third control point vectors (cpMV[2]) are Each average is averaged, and the type of affine model can be saved/propagated with 6-parameters.
  • the first control point vectors (cpMV[0]) and the second control point vectors (cpMV[1]) of the two target candidates are The first and second control point vectors of the average candidate may be respectively calculated by averaging each.
  • the type of the affine model for the average candidate may be set as a 4-parameter. If the average candidate is used to derive the control point vector of the current block, the type of affine model for the average candidate may be stored in the current block and may be propagated to blocks after the current block.
  • the first control point vectors (cpMV[0]), the second control point vectors (cpMV[1]), and the third control point vectors of the two target candidates The first to third control point vectors of the average candidate may be calculated by averaging each of (cpMV[2]).
  • the type of the affine model of the average candidate may be set as a 6-parameter.
  • the type of an affine model for the average candidate may be stored in the current block and may be propagated to blocks subsequent to the current block.
  • the first control point vector of the average candidate may be calculated as an average value of the first control point vector of Cand0 and the first control point vector of Cand1.
  • the second control point vector of the average candidate may be set to the second control point vector of Cando or may be calculated as an average value of the second control point vector of Cando and the third control point vector of Cando.
  • the second control point vector is derived based on the first control point vector and the third control point vector of Cand1, and the second control point of the average candidate is an average value of the second control point vector of Cand0 and the pre-derived second control point vector of Cand1.
  • a vector may be produced.
  • one of the two target candidates (Cand0) has first and second control point vectors, and the other one ( Cand1) may be restricted so that an average candidate is not generated based on these two target candidates when it has the first and third control point vectors.
  • the affine model type can be stored/propagated as either a 4-parameter or a 6-parameter.
  • a third control point vector can be derived using the first control point vector and the second control point vector.
  • the second control point vector can be derived using the first control point vector and the third control point vector.
  • the first to third control point vectors of the average candidate may be calculated by averaging the first to third control point vectors of other 6-parameter target candidates, respectively.
  • the type of the affine model of the average candidate may be set as a 6-parameter.
  • a type having a small number of parameters may be stored/propagated. That is, when the type of the affine model for Cand0 is 4-parameter and the type of the affine model for Cand1 is 6-parameter, the type of the affine model of the average candidate is the same as the type of the affine model for Cand0. It can be set as 4-parameter.
  • the type of an affine model for the average candidate may be stored in the current block and may be propagated to blocks subsequent to the current block.
  • the difference between the control point vectors may mean a first difference between the first control point vector of Cand0 and the first control point vector of Cand1.
  • the difference between the control point vectors may mean a second difference between the second control point vector of Cand0 and the second control point vector of Cand1.
  • the difference between the control point vectors may mean a third difference between the third control point vector of Cand0 and the third control point vector of Cand1.
  • the difference between the control point vectors may mean a fourth difference between the differential control point vector for Cand0 and the differential control point vector for Cand1.
  • the difference control point vectors are (cpMV[0]-cpMV[1]), (cpMV[0]-cpMV[2]), (cpMV[1]-cpMV[2]), (cpMV[1]-cpMV [0]), (cpMV[2]-cpMV[0]), or (cpMV[2]-cpMV[1]).
  • the second control point vector may be derived based on the first control point vector and the third control point vector of the target candidate.
  • Cand0 and Cand1 may not have the third control point vector, and in this case, the third control point vector may be derived based on the first control point vector and the second control point vector of the target candidate.
  • the difference between the control point vectors may mean any one of the first difference, the second difference, the third difference, or the fourth difference, and may include at least two of the first to fourth differences.
  • the threshold value may be the same pre-defined value for the encoding device and the decoding device.
  • an average candidate may be generated.
  • affine candidates of the candidate list may be rearranged, and the average candidate may be generated based on the rearranged candidate list.
  • the candidate list may be rearranged by prioritizing the type of a 6-parameter affine model.
  • the candidate list may be rearranged by prioritizing the type of the 4-parameter affine model.
  • the candidate list may be rearranged by prioritizing the type of the affine model of the first affine candidate in the candidate list.
  • the first affine candidate may mean a candidate to which the smallest index is assigned within the candidate list.
  • the candidate list may be rearranged by prioritizing the type of the affine model having the highest frequency among affine candidates existing in the candidate list.
  • An average candidate may be generated using only two target candidates having the same affine model type in the rearranged candidate list.
  • an average candidate may be generated based on two target candidates selected according to a predetermined priority order from the rearranged candidate list regardless of whether the two target candidates have the same affine model type.
  • a flag (LIC_flag) indicating whether local illumination compensation is applied to the average candidate can be derived as follows.
  • the LIC_flag for the average candidate can be stored/propagated to TRUE. If local roughness compensation is limited to be applied only to blocks of unidirectional prediction, even if it is stored as TRUE according to the aforementioned inheritance rule, it may be changed to FALSE and stored/propagated.
  • a control point vector of any one of two target candidates may be stored. In this case, the LIC_flag of the average candidate may follow the LIC flag included in the motion information of any one of the two target candidates.
  • weight information (BCW_index) for bi-directional weighted prediction of the average candidate can be derived as follows.
  • the BCW_index of the average candidate may be stored/propagated as the BCW_index of the target candidate. Otherwise, BCW_index for the average candidate may be stored/propagated with a default value (eg, 0).
  • BCW_index is a default value
  • a final prediction block may be generated through an average of the L0 prediction block and the L1 prediction block.
  • LIC_flag and BCW_index of the average candidate may be set to FALSE and default values, respectively, without inheriting from the target candidate.
  • the average candidate may be adaptively added/saved to the candidate list in consideration of specific pre-defined conditions.
  • the specific condition is whether a target candidate is available (condition 1), whether two target candidates have the same reference picture index (condition 2), or whether two target candidates have the same reference picture (condition 2). 3) may include at least one of them.
  • variable list has a value of 0 or 1, and may indicate either an L0 prediction direction or an L1 prediction direction.
  • the variable idx has a value of 0 to 2, which may indicate the first to third control point vectors, respectively.
  • the variable idx has values of 0 and 1, which may indicate the first and second control point vectors, respectively.
  • Avail0[list] may indicate the availability of the first target candidate for each prediction direction
  • Avail1[list] may indicate the availability of the second target candidate for each prediction direction. For example, if AvailX[0] is 1 (where X is 0 or 1), this may indicate that the target candidate is available for the L0 prediction direction. In other words, this may mean that the target candidate has motion information in the L0 prediction direction. Conversely, if AvailX[0] is 0, this may indicate that the target candidate is not available for the L0 prediction direction. In other words, this may mean that the target candidate does not have motion information in the L0 prediction direction. Similarly, if AvailX[1] is 1, this may indicate that the target candidate is available for the L1 prediction direction.
  • this may mean that the target candidate has motion information in the L1 prediction direction.
  • AvailX[1] 0 when the target candidate is not available for the L1 prediction direction. In other words, this may mean that the target candidate does not have motion information in the L1 prediction direction.
  • ref_idx0[list] may indicate a reference picture index of a first target candidate for each prediction direction
  • ref_idx1[list] may indicate a reference picture index of a second target candidate for each prediction direction.
  • Whether or not the average is determined may be determined for at least one of the L0 prediction direction and the L1 prediction direction.
  • an average candidate may be generated based on the two target candidates.
  • the control point vector in the L0 prediction direction for the average candidate may be calculated as an average value between the control point vector in the L0 prediction direction for the first target candidate and the control point vector in the L0 prediction direction for the second target candidate.
  • an average candidate may not be generated based on two target candidates. That is, the control point vector in the L0 prediction direction for the average candidate may not be calculated as an average value between the control point vector in the L0 prediction direction for the first target candidate and the control point vector in the L0 prediction direction for the second target candidate. Instead, the control point vector of the L0 prediction direction for the average candidate may be set/stored as a control point vector of the L0 prediction direction of the available target candidate for the L0 prediction direction.
  • the above method can be equally applied to the L1 prediction direction. That is, based on whether two target candidates are available for the L1 prediction direction, it is possible to determine whether to average the two target candidates.
  • an average candidate can be generated based on the two target candidates.
  • the control point vector of the L1 prediction direction for the average candidate may be set/saved as the control point vector of the L1 prediction direction of the available target candidate for the L1 prediction direction.
  • each prediction direction it is possible to determine whether to average two target candidates based on whether the two target candidates have the same reference picture index. Whether or not the average is determined may be determined for at least one of the L0 prediction direction and the L1 prediction direction.
  • an average candidate may be generated based on the two target candidates.
  • the control point vector in the L0 prediction direction for the average candidate may be calculated as an average value between the control point vector in the L0 prediction direction for the first target candidate and the control point vector in the L0 prediction direction for the second target candidate.
  • an average candidate may not be generated based on the two target candidates. That is, the control point vector in the L0 prediction direction for the average candidate may not be calculated as an average value between the control point vector in the L0 prediction direction for the first target candidate and the control point vector in the L0 prediction direction for the second target candidate. Instead, the control point vector of the L0 prediction direction for the average candidate may be set/stored as a control point vector of the L0 prediction direction for either one of the two target candidates.
  • any one of the two target candidates may mean a candidate having a smaller candidate list index (ie, Cando) among the two target candidates.
  • one of the two target candidates may mean a candidate having a smaller reference picture index among the two target candidates.
  • the above method can be equally applied to the L1 prediction direction. That is, based on whether the reference picture index of the first target candidate in the L1 prediction direction is the same as the reference picture index of the second target candidate in the L1 prediction direction, whether to average the two target candidates may be determined.
  • an average candidate may be generated based on the two target candidates.
  • the reference picture indices of the L1 prediction direction for the two target candidates are different from each other (ie, ref_idx0[1] ⁇ ref_idx1[1])
  • an average candidate may not be generated based on the two target candidates.
  • the control point vector of the L1 prediction direction for the average candidate may be set/saved as the control point vector of the L1 prediction direction for either of the two target candidates.
  • each prediction direction it is possible to determine whether to average two target candidates based on whether the two target candidates have the same reference picture. Whether two reference pictures are the same can be checked through picture order count (POC) assigned to each of the two reference pictures. That is, if the difference in output order information between two reference pictures is 0, the two reference pictures are the same reference picture. Otherwise, the two reference pictures may not be the same reference picture. Whether or not the average is determined may be determined for at least one of the L0 prediction direction and the L1 prediction direction.
  • POC picture order count
  • output order information (POC ref0 [Cand0]) of a reference picture in the L0 prediction direction for the first target candidate and output order information (POC ref0 [Cand1]) of a reference picture in the L0 prediction direction for the second target candidate If the difference between ⁇ is 0, an average candidate may be generated based on the two target candidates.
  • the control point vector in the L0 prediction direction for the average candidate may be calculated as an average value between the control point vector in the L0 prediction direction for the first target candidate and the control point vector in the L0 prediction direction for the second target candidate.
  • an average candidate may not be generated based on the two target candidates. That is, the control point vector in the L0 prediction direction for the average candidate may not be calculated as an average value between the control point vector in the L0 prediction direction for the first target candidate and the control point vector in the L0 prediction direction for the second target candidate. Instead, the control point vector of the L0 prediction direction for the average candidate may be set/stored as a control point vector of the L0 prediction direction for either one of the two target candidates.
  • any one of the two target candidates may mean a candidate having a smaller candidate list index (ie, Cando) among the two target candidates.
  • one of the two target candidates may mean a candidate having a smaller reference picture index among the two target candidates.
  • one of the two target candidates may mean a candidate having a smaller difference from output order information of the current picture to which the current block belongs among the two target candidates.
  • the above method can be equally applied to the L1 prediction direction. That is, output order information (POC ref1 [Cand0]) of a reference picture in the L1 prediction direction for the first target candidate and output order information (POC ref1 [Cand1]) of a reference picture in the L1 prediction direction for the second target candidate Based on whether the difference is zero, it can be determined whether or not to average the two target candidates.
  • an average candidate may be generated based on two target candidates.
  • the control point vector in the L1 prediction direction for the average candidate may be calculated as an average value between the control point vector in the L1 prediction direction for the first target candidate and the control point vector in the L1 prediction direction for the second target candidate.
  • the difference between POC ref1 [Cand0] and POC ref1 [Cand1] is not 0, an average candidate may not be generated based on the two target candidates. Instead, the control point vector of the L1 prediction direction for the average candidate may be set/saved as the control point vector of the L1 prediction direction for either of the two target candidates.
  • the pre-defined condition may be any one of the above-described (condition 1) to (condition 3).
  • the pre-defined condition may include at least two of the above (condition 1) to (condition 3).
  • condition 1 and condition 2 it is first determined whether two target candidates satisfy (condition 1), and only when (condition 1) is satisfied, two It may be determined whether the target candidate satisfies (Condition 2). Alternatively, it may be simultaneously determined whether (condition 1) and (condition 2) are satisfied for the two target candidates.
  • the corresponding average candidate may be added to the candidate list.
  • the average candidate may be added to the candidate list.
  • the affine candidate may include a combined candidate.
  • the combined candidate may be generated by combining two target candidates (Cand0 and Cand1) included in the candidate list.
  • the combined candidate may be generated by combining motion information of Cand0 in the first prediction direction and motion information of Cand1 in the second prediction direction.
  • the first prediction direction may be one of the L0 prediction direction and the L1 prediction direction
  • the second prediction direction may be the other one of the L0 prediction direction and the L1 prediction direction.
  • the combined candidate may be independently defined as an affine candidate for an affine mode, and may be used together with the above-described average sign.
  • a combined candidate or an average candidate may be selectively used, or the combined candidate may be substituted for the average candidate.
  • Idx represents a priority order of combinations, and a combined candidate may be generated based on two target candidates corresponding to Idx. For example, when Idx is 0, motion information in the L0 prediction direction of a merge candidate having an index of 0 in the candidate list is combined with motion information in the L1 prediction direction of a merge candidate having an index of 1 in the candidate list. candidates can be created.
  • the combined candidate may be generated when the types of affine models for the two specified target candidates are the same.
  • the affine model type may be stored/propagated as either a 4-parameter or a 6-parameter.
  • a third control point vector can be derived using the first control point vector and the second control point vector.
  • the second control point vector can be derived using the first control point vector and the third control point vector.
  • a combined candidate can be created through a combination with another 6-parameter target candidate.
  • the type of affine model for the combined candidate may be set as a 6-parameter.
  • the type of the affine model for the combined candidate may be stored in the current block and propagated to blocks subsequent to the current block.
  • a type having a small number of parameters may be stored/propagated. That is, when the type of the affine model for Cand0 is 4-parameter and the type of the affine model for Cand1 is 6-parameter, the type of the affine model for the combined candidate is the type of the affine model for Cand0 and the type of the affine model for Cand1. It can be set as 4-parameter in the same way.
  • the type of the affine model for the combined candidate may be stored in the current block and propagated to blocks subsequent to the current block.
  • the difference between the control point vectors may mean a first difference between the first control point vector of Cand0 and the first control point vector of Cand1.
  • the difference between the control point vectors may mean a second difference between the second control point vector of Cand0 and the second control point vector of Cand1.
  • the difference between the control point vectors may mean a third difference between the third control point vector of Cand0 and the third control point vector of Cand1.
  • the difference between the control point vectors may mean a fourth difference between the differential control point vector for Cand0 and the differential control point vector for Cand1.
  • the difference control point vectors are (cpMV[0]-cpMV[1]), (cpMV[0]-cpMV[2]), (cpMV[1]-cpMV[2]), (cpMV[1]-cpMV [0]), (cpMV[2]-cpMV[0]), or (cpMV[2]-cpMV[1]).
  • the second control point vector may be derived based on the first control point vector and the third control point vector of the target candidate.
  • Cand0 and Cand1 may not have the third control point vector, and in this case, the third control point vector may be derived based on the first control point vector and the second control point vector of the target candidate.
  • the difference between the control point vectors may mean any one of the first difference, the second difference, the third difference, or the fourth difference, and may include at least two of the first to fourth differences.
  • the threshold value may be the same pre-defined value for the encoding device and the decoding device.
  • a combined candidate may be generated.
  • affine candidates of the candidate list may be rearranged, and the combined candidate may be generated based on the rearranged candidate list.
  • the candidate list may be rearranged by prioritizing the type of a 6-parameter affine model.
  • the candidate list may be rearranged by prioritizing the type of the 4-parameter affine model.
  • the candidate list may be rearranged by prioritizing the type of the affine model for the first affine candidate in the candidate list.
  • the first affine candidate may mean a candidate to which the smallest index is assigned among a plurality of affine candidates belonging to the candidate list.
  • the candidate list may be rearranged by prioritizing the type of the affine model having the highest frequency among affine candidates existing in the candidate list.
  • embodiments of a method for generating an average candidate may be equally/similarly applied to generate a combined candidate, and a detailed description thereof will be omitted.
  • the candidate list may further include a candidate having a zero motion vector (hereinafter referred to as a zero candidate).
  • the zero candidate may be added to the candidate list after rearrangement of affine candidates, which will be described later. Alternatively, rearrangement of affine candidates may be performed after the candidate having the zero motion vector is added to the candidate list.
  • the plurality of affine candidates described above may be arranged in the candidate list based on a predetermined priority order. For example, a plurality of affine candidates may be arranged in the candidate list in the order of an inherited candidate, a constructed candidate, an average candidate, and a zero candidate. Alternatively, the plurality of affine candidates may be arranged in the candidate list in the order of inherited candidates, average candidates, constructed candidates, and zero candidates. Alternatively, inherited candidates may be divided into spatial candidates and temporal candidates. In this case, the plurality of affine candidates may be arranged in the candidate list in the order of a spatial candidate, an average candidate, a temporal candidate, a constructed candidate, and a zero candidate.
  • the method of configuring the inherited and/or configured candidates and the number of the candidates may be changed prior to generation of the average candidate. That is, a configuration method and a change in the number of inherited candidates and/or configured candidates can be applied when an average candidate is added to the candidate list. However, it is not limited thereto, and may be applied independently regardless of whether the average candidate is used or not.
  • one block coded in the affine mode among the left neighboring blocks adjacent to the current block and one block encoded in the affine mode among the upper neighboring blocks adjacent to the current block are selected.
  • two or more blocks encoded in an affine mode from each of the left neighboring blocks and the upper neighboring blocks may be used to configure more diverse inherited candidates and increase the number of average candidates that can be generated.
  • the check order of neighboring blocks may be changed as shown in Table 2 below.
  • the check order of the neighboring blocks may mean an order in which inherited candidates derived based on the neighboring blocks are inserted into the candidate list.
  • Case 1 A1, B1, B0, A0, B2 Case 2: B1, A1, B0, A0, B2 Case 3: A1, B1, B0, A0, (if Num ⁇ 4, B2) Case 4: B1, A1, B0, A0, (if Num ⁇ 4, B2) Case 5: w > h ? A1, B1, B0, A0, B2 : B1, A1, B0, A0, B2 Case 6: w > h ? A1, B1, B0, A0, (if Num ⁇ 4, B2) : B1, A1, B0, A0, (if Num ⁇ 4, B2) : B1, A1, B0, A0, (if Num ⁇ 4, B2)
  • A0 may mean a lower left neighboring block
  • A1 may mean a left neighboring block
  • B0 may mean an upper right neighboring block
  • B1 may mean an upper neighboring block
  • B2 may mean an upper left neighboring block.
  • A0 is (x-1, y + h) Neighboring block
  • A1 is a neighboring block containing samples at coordinates (x-1, y+h-1)
  • B0 is a neighboring block containing samples at coordinates (x+w, y-1)
  • B1 is (x+w- 1, y-1) may mean a neighboring block including a sample
  • B2 may mean a neighboring block including a sample of (x-1, y-1) coordinates, respectively.
  • the candidate list may be configured by checking whether the corresponding neighboring block is a block encoded in the affine mode in the order according to Case 3 or Case 4. However, whether to check B2 may be determined in consideration of the number of affine candidates previously included in the candidate list.
  • a different check order may be applied according to the shape of the current block. For example, if the width of the current block is greater than the height, the check order of A1, B1, B0, A0, B2 is applied, and if the width of the current block is less than or equal to the height, B1, A1, B0, A0, B2 A check order of can be applied. Conversely, if the width of the current block is greater than the height, the check order of B1, A1, B0, A0, B2 is applied, and if the width of the current block is less than or equal to the height, A1, B1, B0, A0, B2 A check order of may be applied.
  • the check order of A1, B1, B0, A0, B2 is applied, and the ratio of the width to height of the current block is less than the predetermined threshold.
  • the check order of B1, A1, B0, A0, and B2 may be applied.
  • the check order of B1, A1, B0, A0, B2 is applied, and the ratio of the width to height of the current block is greater than the predetermined threshold. If less than or equal to, the check order of A1, B1, B0, A0, B2 may be applied.
  • the foregoing embodiment assumes that up to five neighboring blocks are available. However, depending on the shape of the current block or the width-to-height ratio, at least one of the five neighboring blocks may be set to be unavailable. In this case, the above-described check process may be omitted for an unavailable neighboring block.
  • B2 means an upper left neighboring block
  • B3 means an upper neighboring block
  • A2 means a left neighboring block, respectively
  • B2 B3 and A2 may be adjacent to the first control point corresponding to the upper left corner of the current block.
  • B1 denotes an upper-side neighboring block and B0 denotes an upper-right neighboring block, respectively, and B1 and B0 may be adjacent to the second control point corresponding to the upper-right corner of the current block.
  • A1 denotes a left neighboring block and A0 denotes a lower left neighboring block, respectively, and A1 and A0 may be adjacent to the third control point corresponding to the lower left corner of the current block.
  • T0 is a temporal neighboring block, and may be temporally adjacent to the fourth control point corresponding to the lower right corner of the current block.
  • B2 is a neighboring block containing samples of coordinates (x-1, y-1)
  • B3 is a neighboring block containing samples of coordinates (x, y-1)
  • A2 is a neighboring block containing samples of coordinates (x-1, y).
  • Each of the neighboring blocks including the sample may mean.
  • B1 may mean a neighboring block including a sample at coordinates of (x+w-1, y-1)
  • B0 may mean a neighboring block including a sample at coordinates of (x+w-1, y-1).
  • A1 may mean a neighboring block including a sample at coordinates of (x-1, y+h-1)
  • A0 may mean a neighboring block including a sample at coordinates of (x-1, y+h).
  • T0 may mean a neighboring block including samples of (x+w, y+h) coordinates.
  • the spatial/temporal neighboring blocks of the current block may be divided into a plurality of groups. For convenience of description, it is assumed that spatial/temporal neighboring blocks of the current block are divided into four groups.
  • the four groups mean a first group consisting of at least one of B2, B3, or A2, a second group consisting of at least one of B1 or B0, a third group consisting of at least one of A1 or A0, and a fourth group consisting of T0. can do. However, it is not limited thereto, and can be divided into two or three groups among the four groups described later.
  • a control point vector corresponding to each control point may be derived by sequentially searching one or more neighboring blocks belonging to each group according to a predetermined search order.
  • the search may be performed until one available motion vector is derived. That is, when even one usable motion vector is derived in the search process, the search may be prematurely terminated.
  • the search may be performed until all available motion vectors are derived from all neighboring blocks belonging to each group. In this case, one or more motion vectors may be derived for each group.
  • one or more usable motion vectors may be derived by performing a search in the order of B2, B3, and A2.
  • one or more usable motion vectors may be derived by performing search in the order of B1 and B0.
  • one or more usable motion vectors may be derived by performing search in the order of A1 and A0.
  • a candidate configured using motion vectors derived for each group may be generated.
  • a candidate constructed using only motion vectors having the same reference picture index among motion vectors derived for each group may be generated.
  • a candidate constructed using only motion vectors having the same reference picture may be generated.
  • Neighboring blocks having the same reference picture index may be extracted from each group, and a candidate configured based on motion vectors of the extracted neighboring blocks may be generated.
  • a neighboring block having a reference picture index of 0 and a neighboring block having a reference picture index of 1 may be derived according to a predetermined search order for each group.
  • the constructed candidates are the motion vector of B2 (cpMV[0]), the motion vector of B1 (cpMV[1]), and the motion vector of A1. (cpMV[2]) or a motion vector of T0 (cpMV[3]).
  • the configured candidates are (cpMV[0], cpMV[1], cpMV[2]), (cpMV[0], cpMV[1], cpMV[3]), (cpMV[0], cpMV[2] , cpMV[3]), (cpMV[1], cpMV[2], cpMV[3]), (cpMV[0], cpMV[1]), (cpMV[0], cpMV[2]), (cpMV [0], cpMV[3]), (cpMV[1], cpMV[2]), (cpMV[1], cpMV[3]), or (cpMV[2], cpMV[3]) can include Through the same method, a configured candidate may be generated based on a motion vector of a neighboring block having a reference picture index of 1, and a detailed description thereof will be omitted.
  • a candidate configured based on motion vectors of neighboring blocks having the same reference picture index may be generated, and each candidate configured may be generated for each reference picture index.
  • Case 1 The order of (B2->B3->A2), (B1->B0), (A1->A0), and T0 of neighboring blocks having a reference picture index of 0 among spatial/temporal neighboring blocks adjacent to each control point , and in the same way, a neighboring block having a reference picture index of 1 is checked to construct a candidate configured for each reference picture index.
  • Case 2 The order of (B2->A2->B3), (B0->B1), (A0->A1), and T0 of neighboring blocks with reference picture index 0 among spatial/temporal neighboring blocks adjacent to each control point , and in the same way, a neighboring block having a reference picture index of 1 is checked to construct a candidate configured for each reference picture index.
  • the reference picture index is also considered for T0, which is a temporal neighboring block, but this is only an example.
  • T0 can be used to generate a constructed candidate, regardless of whether the reference picture index is the same. Alternatively, it may be restricted not to be used to generate configured candidates for T0.
  • candidates configured based on neighboring blocks having the same reference picture index are generated, but this is only an example.
  • Neighboring blocks having the same reference picture are derived from each group, and a candidate configured through a combination of motion vectors for the derived neighboring blocks may be generated.
  • a duplicate check process may be performed in the process of constructing the candidate list. At least one of an inherited candidate, a constructed candidate, an average candidate, and a zero candidate may be added to the candidate list, and each affine candidate may be selectively added through overlapping checks with affine candidates previously included in the candidate list. there is.
  • affine mode various types of affine models such as 4-parameter and 6-parameter may be used, and thus, the number of control point vectors may be different for each affine candidate. Therefore, it is necessary to equalize the number of control point vectors in order to check redundancy between affine candidates.
  • the duplication check may be performed between an affine candidate previously included in the candidate list (hereinafter referred to as a previous affine candidate) and an affine candidate to be added to the candidate list (hereinafter referred to as a current affine candidate).
  • the type of the affine model of the current affine candidate when the type of the affine model of the current affine candidate is 4-parameter, it may include only the first control point vector cpMV[0] and the second control point vector cpMV[1]. In this case, by deriving a motion vector from the third control point and assigning it to the third control point vector (cpMV[2]), it is possible to check overlap with the previous affine candidate having a 6-parameter affine model type. can do Alternatively, when the type of the affine model of the current affine candidate is 4-parameter, it may include only the first control point vector (cpMV[0]) and the third control point vector (cpMV[2]). In this case, by deriving a motion vector at the second control point and assigning it to the second control point vector (cpMV[1]), it is possible to check overlapping with the previous affine candidate having a 6-parameter affine model type. can do
  • the duplication check may be limited to being performed between affine candidates having the same affine model type.
  • an affine candidate whose affine model type is 4-parameter is checked for duplication. can do.
  • a duplicate check may be performed with an affine candidate whose affine model type is 6-parameter among affine candidates previously included in the candidate list.
  • the overlapping check may be performed based on whether the control point vectors are identical.
  • whether the control point vectors are the same can be checked for each of the first to third control point vectors.
  • whether the control point vectors are identical to any one of the first to third control point vectors may be checked, and the other control point vectors may not be checked.
  • any one of the first to third control point vectors may be limited to the first control point vector.
  • whether the control point vectors are identical for any two of the first to third control point vectors may be checked, and the other control point vectors may not be checked.
  • any two of the first to third control point vectors may be limited to the first control point vector and the second control point vector.
  • the duplicate check based on whether the control point vectors are identical may be performed as shown in Table 5 below.
  • Cand[i][0], Cand[i][1], and Cand[i][2] mean the first control point vector, the second control point vector, and the third control point vector of the previous affine candidate, respectively. can do.
  • CurrCand[0], CurrCand[1], and CurrCand[2] may mean the first control point vector, the second control point vector, and the third control point vector of the current affine candidate, respectively.
  • the redundancy check may be performed based on whether control point vector differences are identical.
  • the control point vector difference is the difference between the second control point vector and the first control point vector (hereinafter referred to as the first control point vector difference) or the difference between the third control point vector and the first control point vector (hereinafter referred to as the second control point vector difference). ) may include at least one of them.
  • the control point vector difference may further include a third control point vector difference that is a difference between the third control point vector and the second control point vector.
  • the overlapping check based on whether control point vector differences are identical may be performed as shown in Table 6 below.
  • diffMV0[i] may mean a first control point vector difference between previous affine candidates
  • diffMV1[i] may mean a second control point vector difference between previous affine candidates
  • diffCurrMV0 may mean the first control point vector difference between the current affine candidates
  • diffCurrMV1 may mean the second control point vector difference.
  • the redundancy check may be performed based on whether control point vector differences are identical and whether control point vectors are identical.
  • the control point vector difference may include at least one of a first control point vector difference and a second control point vector difference.
  • the control point vector difference may further include a third control point vector difference that is a difference between the third control point vector and the second control point vector.
  • whether the control point vectors are the same can be checked for each of the first to third control point vectors.
  • whether the control point vectors are identical to any one of the first to third control point vectors may be checked, and the other control point vectors may not be checked.
  • any one of the first to third control point vectors may be limited to the first control point vector.
  • control point vectors are identical for any two of the first to third control point vectors may be checked, and the other control point vectors may not be checked.
  • any two of the first to third control point vectors may be limited to the first control point vector and the second control point vector.
  • the duplication check may be performed as shown in Table 7 below.
  • (diffMV1[i] ! diffCurrMV1)
  • diffMV0[i] may mean a first control point vector difference between previous affine candidates
  • diffMV1[i] may mean a second control point vector difference between previous affine candidates
  • diffCurrMV0 may mean the first control point vector difference between the current affine candidates
  • diffCurrMV1 may mean the second control point vector difference.
  • Cand[i][0] may mean a first control point vector of a previous affine candidate
  • CurrCand[0] may mean a first control point vector of a current affine candidate.
  • the current affine Candidates can be added to the candidate list.
  • the first control point vector of the previous affine candidate and the second control point vector of the current affine candidate are the same.
  • the current affine candidate may be added to the candidate list.
  • the difference between the first and second control point vectors of the previous affine candidate and the difference between the first and second control point vectors of the current affine candidate are the same, and the first control point vector of the previous affine candidate and the second control point vector of the current affine candidate are the same. 1 If the control point vectors are the same, the current affine candidate may not be added to the candidate list.
  • the redundancy check is whether the difference between the control point vectors between the previous affine candidate and the current affine candidate is smaller than a predetermined threshold value or whether the difference between the control point vectors between the previous affine candidate and the current affine candidate is a predetermined threshold value. It may be performed based on at least one of whether it is less than or not.
  • the difference between the control point vectors is the difference between the first control point vector difference between the previous affine candidate and the first control point vector difference between the current affine candidate and the second control point vector difference between the previous affine candidate and the current affine candidate. It may include at least one of a difference between the control point vector difference and a difference between the third control point vector difference between the previous affine candidate and the third control point vector difference between the current affine candidate.
  • the difference between the control point vectors between the previous affine candidate and the current affine candidate is the difference between the first control point vector of the previous affine candidate and the first control point vector of the current affine candidate, the second control point vector of the previous affine candidate and the current affine candidate. It may include at least one of a difference between the second control point vector of the affine candidate and a difference between the third control point vector of the previous affine candidate and the third control point vector of the current affine candidate.
  • the threshold values may be partly the same or different depending on comparison targets.
  • the threshold value may be variably determined in consideration of the shape, size, or picture resolution of a block, or may be a pre-defined value identically to an encoding device and a decoding device.
  • the duplication check may be performed as shown in Table 8 below.
  • diffMV0[i] may mean a first control point vector difference between previous affine candidates
  • diffMV1[i] may mean a second control point vector difference between previous affine candidates
  • diffCurrMV0 may mean the first control point vector difference between the current affine candidates
  • diffCurrMV1 may mean the second control point vector difference.
  • Cand[i][0] may mean a first control point vector of a previous affine candidate
  • CurrCand[0] may mean a first control point vector of a current affine candidate.
  • the difference between the first control point vector difference of the previous affine candidate and the first control point vector difference of the current affine candidate (
  • ) between the difference between the 2 control point vectors and the second control point vector of the current affine candidate is less than threshold_1, or the first control point vector of the previous affine candidate and the second control point vector of the current affine candidate
  • the difference between the two control point vectors (
  • the current affine candidate may be added to the candidate list.
  • the current affine candidate may not be added to the candidate list.
  • compression efficiency can be increased by generating various affine candidates.
  • various candidates are configured by increasing the number of merge candidates for merge mode without changing the number of affine candidates for affine mode, the selection probability of merge mode rather than affine mode is increased. Therefore, it is necessary to increase the number of affine candidates for affine modes.
  • by deriving the same number of affine candidates as merge candidates for merge mode ease of implementation and understanding can be improved. For example, when the number of merge candidates for merge mode is 10, the number of affine candidates for affine mode may be determined as 10 or (10-1). Alternatively, when the number of merge candidates for merge mode is 15, the number of affine candidates for affine mode may be determined as 15 or (15-1).
  • a plurality of affine candidates belonging to the candidate list may be rearranged.
  • the rearrangement may be performed based on a template-based cost.
  • the plurality of affine candidates may be grouped into one or more subgroups.
  • a candidate with a zero motion vector belonging to the candidate list may be excluded from the grouping target.
  • the candidate list may include a first subgroup composed of inherited candidates, a second subgroup composed of composed candidates, a third subgroup composed of history-based candidates, a fourth subgroup composed of average candidates, or other candidates. Can be divided into at least one of the fifth subgroup consisting of.
  • the size of the subgroup may be 2, 3, 4, 5 or more as pre-defined equally in the encoding device and the decoding device.
  • the size of the subgroup may be determined/defined differently according to the inter prediction mode of the current block. For example, when the current block is coded in the merge mode, the size of the subgroup may be 5, and when the current block is coded in the affine mode, the size of the subgroup may be 3.
  • the rearrangement may be performed for each subgroup of the candidate list. In this case, it may be performed for all subgroups of the candidate list or only for some subgroups. For example, rearrangement may not be performed on a subgroup including a candidate having a zero motion vector. Alternatively, rearrangement may not be performed on a subgroup to which a candidate having the smallest index (ie, 0) among the indexes of the candidate list belongs, and rearrangement may be performed on a subgroup to which a candidate having the largest index belongs. .
  • rearrangement may be performed on a subgroup to which a candidate having the smallest index among indexes of the candidate list belongs, and rearrangement may not be performed on a subgroup to which a candidate having the largest index belongs.
  • rearrangement may not be performed on a subblock to which a candidate with the smallest index and a candidate with the largest index among the indexes of the candidate list belong, respectively, and rearrangement may be performed on the remaining subgroups.
  • rearrangement may not be performed on a subgroup to which a candidate having the largest index among indexes of the candidate list belongs, and rearrangement may be performed on the remaining subgroups.
  • the template-based cost for rearrangement may be calculated as a sum of absolute difference (SAD) between samples belonging to the template area of the current block and samples belonging to the reference template area.
  • the SAD may be calculated for each of a plurality of affine candidates belonging to the subgroup.
  • a plurality of affine candidates belonging to the subgroup may be rearranged in ascending order of the calculated SAD.
  • the template area of the current block may be a peripheral area adjacent to the current block.
  • the peripheral area may include at least one of an upper peripheral area, a left peripheral area, and an upper left peripheral area of the current block.
  • the upper peripheral area may have a size of W ⁇ N
  • the left peripheral area may have a size of N ⁇ H
  • the upper left peripheral area may have a size of N ⁇ N.
  • N can be an integer of 1, 2, 3 or more.
  • N may be the same pre-defined value for the encoding device and the decoding device, and may be variably determined based on at least one of the size, shape, and inter prediction mode of the current block.
  • the range of the neighboring area may be determined differently according to the inter prediction mode of the current block.
  • the neighboring area may be composed of an upper neighboring area and a left neighboring area, or the neighboring area may be composed of either an upper neighboring area or a left neighboring area.
  • the reference template region may be determined based on motion information of an affine candidate belonging to the candidate list. For example, a motion vector in sub-block units for the current block may be derived based on the control point vector of the affine candidate. Based on the motion vectors of each subblock belonging to the top row and left column of the current block, reference subblocks corresponding to the motion vectors may be determined.
  • the reference template area may be configured based on the sub-template area of each reference sub-block.
  • a sub-template region of a reference sub-block corresponding to a sub-block belonging to an upper row in the current block may be configured as an upper peripheral region of the reference sub-block.
  • the width and height of the reference subblock are subW and subH
  • the area around the top of the reference subblock may have a size of subW x N.
  • the sub-template region of the reference sub-block corresponding to the sub-block belonging to the left column of the current block may be composed of the left peripheral region of the reference sub-block.
  • the left peripheral area of the reference subblock may have a size of N x subH.
  • the sub-template region of the reference sub-block corresponding to the upper-left sub-block may consist of the upper and left peripheral regions of the reference sub-block.
  • the upper peripheral area of the reference subblock may have a size of subW x N
  • the left peripheral area may have a size of N x subH.
  • the template area of the current block is composed of an upper edge area and a left edge area. Accordingly, the range of the reference template area may be determined to correspond to the range of the template area of the current block.
  • each reference subblock may be determined using only the motion vector of each subblock belonging to the upper row in the current block, and the subtemplate region of each reference subblock Based on this, the reference template area can be configured.
  • each reference subblock may be determined using only the motion vector of each subblock belonging to the left column in the current block, and based on the subtemplate region of each reference subblock The reference template area can be configured.
  • Rearrangement may be performed for each type of affine candidate. For example, rearrangement may be performed on each of the inherited and configured candidates. Alternatively, rearrangement may be performed on average candidates.
  • a plurality of affine candidates belonging to the candidate list may be divided into subgroups, and rearrangement may be performed for each subgroup. For example, when the size of a subgroup for rearrangement is 5, rearrangement may be performed targeting 5 affine candidates belonging to each subgroup.
  • Rearrangement may be performed for all affine candidates belonging to the candidate list without considering the type of the affine candidate and/or the size of the subgroup.
  • a candidate index may be signaled for all candidates belonging to the rearranged candidate list.
  • the candidate index may be information specifying any one of all candidates belonging to the rearranged candidate list.
  • a candidate index may be signaled targeting a predetermined number of candidates independently of the total number of candidates belonging to the rearranged candidate list.
  • the candidate index may be information specifying any one of some candidates rather than all candidates belonging to the rearranged candidate list.
  • the partial candidates may mean affine candidates belonging to one or more rearranged subgroups.
  • the partial candidates may mean affine candidates belonging to a first subgroup among a plurality of subgroups of the candidate list.
  • the first subgroup may refer to a subgroup to which an affine candidate having the smallest index among indexes of the candidate list belongs.
  • the predetermined number is a pre-defined number identically to the encoding device and the decoding device, and may be smaller than the total number of candidates belonging to the candidate list. If the predetermined number is two, the partial candidates may mean the top two candidates of the candidate list. At this time, the candidate index may be information specifying any one of the top two candidates. However, the predetermined number is not limited to two, and may be defined as three, four, five or more.
  • control point vector of the current block may be derived based on the candidate list and the candidate index (S410).
  • the candidate index may mean an index coded to derive a control point vector of the current block.
  • the candidate index may specify any one of a plurality of candidates belonging to the candidate list.
  • the control point vector of the current block may be derived using the control point vector of the corresponding affine candidate.
  • the type of the affine model of the current block is 4-parameter (ie, it is determined that the current block uses two control point vectors).
  • the affine candidate specified by the candidate index has three control point vectors
  • three control point vectors of the specified affine candidates may be set as control point vectors of the current block.
  • the type of the affine model of the current block may be updated as a 6-parameter.
  • the type of the affine model of the current block is 6-parameter (ie, it is determined that the current block uses 3 control point vectors).
  • the affine candidate specified by the candidate index has two control point vectors
  • one additional control point vector is generated, and the two control point vectors and the additional control point vector of the affine candidate are used as the control point vector of the current block.
  • the additional control point vector may be derived based on at least one of two control point vectors of affine candidates, a location of a block corner corresponding to the two control point vectors, and size/position information of a current/neighboring block.
  • two control point vectors of the specified affine candidate may be set as control point vectors of the current block.
  • the type of the affine model of the current block may be updated as a 4-parameter.
  • a motion vector of the current block can be derived (S420).
  • the motion vector may be derived in units of subblocks of the current block.
  • the current block may be divided into a plurality of NxM sub-blocks.
  • the N and M values may be 2, 4, 8, 16, 32 or more.
  • inter prediction may be performed on the current block using the derived motion vector (S430).
  • the reference block can be specified using the motion vector of the current block.
  • the reference block may be specified for each sub-block of the current block.
  • a reference block of each sub-block may belong to one reference picture. That is, subblocks belonging to the current block can share one reference picture.
  • reference picture indexes may be independently set for each subblock of the current block.
  • the specified reference block may be set as a prediction block of the current block.
  • Steps S400 to S430 reviewed with reference to FIG. 4 may be performed identically/similarly in the affine inter mode, and duplicate descriptions will be omitted here.
  • the predictive value of the control point vector of the current block is derived based on the above-described candidate list and the candidate index, and the control point vector of the current block may be derived based on the predicted value of the control point vector and the control point vector difference value signaled from the bitstream.
  • This control point vector derivation method can be regarded as a distinguishing feature from the affine merge mode.
  • the maximum number of affine candidates that can be included in the candidate list may be limited to two.
  • the candidate index for the affine inter mode has a value of 0 or 1, and can specify any one of two affine candidates belonging to the candidate list.
  • the candidate list for the affine inter mode may include a plurality of affine candidates, and the plurality of affine candidates select at least one of an inherited candidate, a constructed candidate, a history-based candidate, and an average candidate.
  • the candidate list may further include one or more translational candidates.
  • the translational candidate may mean a candidate having one motion vector.
  • the translational candidate may be derived from a spatial or temporal neighboring block of the current block, and may further include a candidate having a zero motion vector.
  • candidate indices may be signaled for all or some candidates belonging to the candidate list.
  • the number of all or some of the candidates may be 2, 3, 4, 5, or more.
  • the method for determining some candidates is as described above.
  • the average candidate of the candidate list for the aforementioned affine mode may be applied to the merge mode.
  • merge mode one of a plurality of merge candidates belonging to the candidate list is specified based on a candidate index, motion information of a current block is derived based on the specified merge candidate, and motion information of a current block is derived based on the derived motion information. can make predictions.
  • the plurality of merge candidates may include at least one of a spatial candidate, a temporal candidate, a non-adjacent candidate, a history-based candidate, an average candidate, and a zero candidate.
  • the average candidate may be generated based on two target candidates in the candidate list, and a method for selecting the two target candidates is described with reference to Table 1. Also, the two target candidates may be selected from the candidate list in consideration of the candidate types. For example, at least one of the non-adjacent candidate and the history-based candidate may be excluded as a target candidate for generating the average candidate. However, it is not limited thereto, and it goes without saying that the method for selecting a target candidate in the affine mode described above can be equally/similarly applied to the merge mode.
  • the motion vector of the average candidate may be calculated as an average value between a motion vector of the first target candidate Cand0 and a motion vector of the second target candidate Cand1. At least one of the reference picture index of the average candidate, inter-prediction direction information, weight information for bi-directional weighted prediction, or a flag indicating whether local illumination compensation is applied may be inherited from one of the two target candidates.
  • the average candidate may be adaptively added/stored to the candidate list in consideration of a pre-defined specific condition.
  • the specific condition is whether a target candidate is available (condition 1), whether two target candidates have the same reference picture index (condition 2), or whether two target candidates have the same reference picture (condition 2). 3) may include at least one of them.
  • an average candidate may be generated based on the two target candidates.
  • the motion vector in the L0 prediction direction for the average candidate may be calculated as an average value between the motion vector in the L0 prediction direction for the first target candidate and the motion vector in the L0 prediction direction for the second target candidate.
  • an average candidate may not be generated based on the two target candidates. That is, the motion vector in the L0 prediction direction for the average candidate may not be calculated as an average value between the motion vector in the L0 prediction direction for the first target candidate and the motion vector in the L0 prediction direction for the second target candidate. Instead, the motion vector of the L0 prediction direction for the average candidate may be set/saved as the motion vector of the L0 prediction direction of the available target candidate for the L0 prediction direction.
  • an average candidate may be generated based on the two target candidates.
  • the motion vector in the L0 prediction direction for the average candidate may be calculated as an average value between the motion vector in the L0 prediction direction for the first target candidate and the motion vector in the L0 prediction direction for the second target candidate.
  • an average candidate may not be generated based on the two target candidates. That is, the motion vector in the L0 prediction direction for the average candidate may not be calculated as an average value between the motion vector in the L0 prediction direction for the first target candidate and the motion vector in the L0 prediction direction for the second target candidate. Instead, the motion vector of the L0 prediction direction for the average candidate may be set/stored as a motion vector of the L0 prediction direction for either one of the two target candidates.
  • any one of the two target candidates may mean a candidate having a smaller candidate list index (ie, Cando) among the two target candidates.
  • one of the two target candidates may mean a candidate having a smaller reference picture index among the two target candidates.
  • output order information (POC ref0 [Cand0]) of a reference picture in the L0 prediction direction for the first target candidate and output order information (POC ref0 [Cand0]) of a reference picture in the L0 prediction direction for the second target candidate If the difference between ref0 [Cand1]) is 0, an average candidate may be generated based on the two target candidates.
  • the motion vector in the L0 prediction direction for the average candidate may be calculated as an average value between the motion vector in the L0 prediction direction for the first target candidate and the motion vector in the L0 prediction direction for the second target candidate.
  • an average candidate may not be generated based on the two target candidates. That is, the motion vector in the L0 prediction direction for the average candidate may not be calculated as an average value between the motion vector in the L0 prediction direction for the first target candidate and the motion vector in the L0 prediction direction for the second target candidate. Instead, the motion vector of the L0 prediction direction for the average candidate may be set/stored as a motion vector of the L0 prediction direction for either one of the two target candidates.
  • any one of the two target candidates may mean a candidate having a smaller candidate list index (ie, Cando) among the two target candidates.
  • one of the two target candidates may mean a candidate having a smaller reference picture index among the two target candidates.
  • one of the two target candidates may mean a candidate having a smaller difference from output order information of the current picture to which the current block belongs among the two target candidates.
  • the average candidate may be restricted to be added to the candidate list.
  • the average candidate may be added to the candidate list. That is, when each of the motion vectors of the L0 and L1 prediction directions for the average candidate is set/stored as a motion vector of any one of two target candidates, the average candidate may not be added to the candidate list.
  • the average candidate may be adaptively added based on a duplicate check with a merge candidate previously included in the candidate list.
  • the redundancy check may be performed based on whether the motion vector of the average candidate and the motion vector of the previously included merge candidate are the same.
  • the redundancy check may be performed based on whether a difference between the motion vector of the average candidate and the motion vector of the previously included merge candidate is smaller than a predetermined threshold value. If the difference between the motion vector of the average candidate and the motion vector of the previously included merge candidate is smaller than a predetermined threshold, the average candidate may not be added to the candidate list. Otherwise, the average candidate Can be added to the candidate list.
  • embodiments of a method for constructing a candidate list for affine mode may be equally/similarly applied to construct a candidate list for merge mode, and a detailed description thereof will be omitted.
  • FIG. 5 relates to a method of deriving an affine candidate from a control point vector of a spatial/temporal neighboring block as an embodiment according to the present disclosure.
  • the width and height of the current block 500 are cbW and cbH, respectively, and the position of the current block is (xCb, yCb).
  • the width and height of the spatial neighboring blocks 510 to 550 are nbW and nbH, respectively, and the positions of the spatial neighboring blocks are (xNb, yNb).
  • the spatial neighboring block may include at least one of the left block 510, the lower left block 540, the upper right block 430, the upper block 420, or the upper left block 450 of the current block. Alternatively, the spatial neighboring block may further include at least one of a block adjacent to the right side or a block adjacent to the bottom side of the upper left block 450 .
  • a spatial candidate may have N control point vectors (cpMV).
  • N may be an integer of 1, 2, 3, or greater.
  • the N value is based on at least one of information about whether or not the subblock is decoded, information about whether the block is coded in an affine mode, or information about the type (4-parameter or 6-parameter) of the affine model. can be determined
  • the corresponding block may have two control point vectors.
  • the corresponding block may not perform prediction based on the affine model.
  • the corresponding block may have three control point vectors.
  • the corresponding block may not perform prediction based on the affine model.
  • the aforementioned information may be encoded and signaled in an encoding device.
  • all or part of the above information may be derived by the decoding device based on the attribute of the block.
  • the block may mean the current block or spatial/temporal neighboring blocks of the current block.
  • the attribute may mean size, shape, location, segmentation type, inter mode, parameters related to residual coefficients, and the like.
  • the inter mode is a mode pre-defined in the decoding device, and may mean merge mode, skip mode, AMVP mode, affine mode, intra/inter combination mode, current picture reference mode, and the like.
  • the value of N may be derived by the decoding device based on the above-described attribute of the block.
  • the N control point vectors include the first control point vector (cpMV[0]), the second control point vector (cpMV[1]), the third control point vector (cpMV[2]), ... the Nth control point It can be expressed as a vector (cpMV[N-1]).
  • the first control point vector cpMV[0], the second control point vector cpMV[1], the third control point vector cpMV[2], and the fourth control point vector cpMV[3] are block It may be a vector respectively corresponding to the positions of the upper left corner, upper right corner, lower left corner, and lower right corner of .
  • the spatial candidates have three control point vectors, and the three control point vectors may be arbitrary control point vectors selected from the first to Nth control point vectors.
  • the spatial candidate may have two control point vectors, and the two control point vectors may be arbitrary control point vectors selected from the first to Nth control point vectors.
  • the first control point vector is derived based on at least one of the first control point vector of the spatial neighboring block, a predetermined difference value, location information (xCb, yCb) of the current block, or location information (xNb, yNb) of the spatial neighboring block It can be.
  • the number of difference values may be 1, 2, 3 or more.
  • the number of difference values may be variably determined in consideration of the above-described attribute of the block, or may be the same pre-defined value in the encoding/decoding device.
  • the difference value may be defined as a difference value between one of a plurality of control point vectors and another one.
  • the difference value may include a first difference between the second control point vector and the first control point vector, a second difference between the third control point vector and the first control point vector, and a fourth control point vector between the fourth control point vector and the third control point vector. It may include at least one of a third difference value or a fourth difference value between the fourth control point vector and the second control point vector.
  • the first control point vector may be derived as in Equation 1 below.
  • cpMvLX[ 0 ][ 0 ] ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb - yNb ) )
  • the variables mvScaleHor and mvScaleVer may mean the first control point vector of the spatial neighboring block or may mean values derived by applying a shift operation by k to the first control point vector.
  • k may be 1, 2, 3, 4, 5, 6, 7, 8, 9 or more integers.
  • the variables dHorX and dVerX respectively correspond to the x-component and y-component of the first difference between the second control point vector and the first control point vector.
  • the variables dHorY and dVerY respectively correspond to the x-component and y-component of the second difference between the third control point vector and the first control point vector.
  • the above variables may be derived as in Equation 2 below.
  • dHorX ( CpMvLX[ xNb + nNbW - 1 ][ yNb ][ 1 ][ 0 ] - CpMvLX[ xNb ][ yNb ][ 0 ][ 0 ] ) ⁇ ( 7 - log2NbW )
  • the second control point vector is the first control point vector of the spatial neighboring block, a predetermined difference value, position information (xCb, yCb) of the current block, block size (width or height), or position information (xNb, yNb) of the spatial neighboring block It can be derived based on at least one of Here, the block size may mean the size of the current block and/or spatial neighboring blocks. Since the difference value is the same as the first control point vector, a detailed description thereof will be omitted. However, the range and/or number of difference values used in the process of deriving the second control point vector may be different from that of the first control point vector.
  • the second control point vector may be derived as in Equation 3 below.
  • cpMvLX[ 1 ][ 0 ] ( mvScaleHor + dHorX * ( xCb + cbWidth - xNb ) + dHorY * ( yCb - yNb ) )
  • cpMvLX[ 1 ][ 1 ] ( mvScaleVer + dVerX * ( xCb + cbWidth - xNb ) + dVerY * ( yCb - yNb ) )
  • Equation 3 the variables mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, and dVerY are the same as in Equation 1, and a detailed description thereof will be omitted.
  • the third control point vector is the first control point vector of the spatial neighboring block, a predetermined difference value, position information (xCb, yCb) of the current block, block size (width or height), or position information (xNb, yNb) of the spatial neighboring block It can be derived based on at least one of Here, the block size may mean the size of the current block and/or spatial neighboring blocks. Since the difference value is the same as the first control point vector, a detailed description thereof will be omitted. However, the range and/or number of difference values used in the process of deriving the third control point vector may be different from those of the first control point vector or the second control point vector.
  • the third control point vector may be derived as in Equation 4 below.
  • cpMvLX[ 2 ][ 0 ] ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb + cbHeight - yNb ) )
  • cpMvLX[ 2 ][ 1 ] ( mvScaleVer + dVerX * ( xCb - xNb ) + dVerY * ( yCb + cbHeight - yNb ) )
  • Equation 4 the variables mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, and dVerY are the same as in Equation 1, and a detailed description thereof will be omitted. Meanwhile, through the above process, the first to N th control point vectors of spatial candidates may be derived.
  • the first control point vector is based on at least one of a motion vector (MV) of a spatial neighboring block, a predetermined difference value, position information (xCb, yCb) of a current block, or position information (xNb, yNb) of a spatial neighboring block. can be induced.
  • MV motion vector
  • xCb, yCb position information
  • xNb, yNb position information
  • the motion vector may be a motion vector of a sub-block located at the lowermost end of a spatial neighboring block.
  • the sub-block may be located at the leftmost, center or rightmost among a plurality of subblocks located at the lowest end of the spatial neighboring blocks.
  • the motion vector may mean an average value, a maximum value, or a minimum value of motion vectors of subblocks.
  • the number of difference values may be 1, 2, 3 or more.
  • the number of difference values may be variably determined in consideration of the above-described attribute of the block, or may be the same pre-defined value in the encoding/decoding device.
  • the difference value may be defined as a difference value between one of a plurality of motion vectors stored in units of sub blocks in a spatial neighboring block and another one.
  • the difference value may mean a difference value between a motion vector of a lower right subblock of a spatial neighboring block and a motion vector of a lower left subblock.
  • the first control point vector may be derived as in Equation 5 below.
  • cpMvLX[ 0 ][ 0 ] ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb - yNb ) )
  • the variables mvScaleHor and mvScaleVer may mean the motion vector (MV) of the above-described spatial neighboring block or a value derived by applying a shift operation by k to the motion vector.
  • k may be 1, 2, 3, 4, 5, 6, 7, 8, 9 or more integers.
  • the variables dHorX and dVerX correspond to the x-component and y-component of a predetermined difference value, respectively.
  • the difference value means a difference value between a motion vector of a lower right subblock and a motion vector of a lower left subblock in a spatial neighboring block.
  • the variables dHorY and dVerY can be derived based on the variables dHorX and dVerX. The aforementioned variables may be derived as in Equation 6 below.
  • mvScaleHor MvLX[ xNb ][ yNb + nNbH - 1 ][ 0 ] ⁇ 7
  • dHorX ( MvLX[ xNb + nNbW - 1 ][ yNb + nNbH - 1 ][ 0 ] - MvLX[ xNb ][ yNb + nNbH - 1 ][ 0 ] ) ⁇ ( 7 - log2NbW )
  • the second control point vector is a motion vector (MV) of the spatial neighboring block, a predetermined difference value, position information (xCb, yCb) of the current block, block size (width or height), or position information (xNb, yNb) of the spatial neighboring block ) It can be derived based on at least one of.
  • the block size may mean the size of the current block and/or spatial neighboring blocks.
  • the motion vector and the difference value as described in the first control point vector, a detailed description thereof will be omitted.
  • the position of the motion vector used in the process of deriving the second control point vector, and the range and/or number of difference values may be different from those of the first control point vector.
  • the second control point vector may be derived as in Equation 7 below.
  • cpMvLX[ 1 ][ 0 ] ( mvScaleHor + dHorX * ( xCb + cbWidth - xNb ) + dHorY * ( yCb - yNb ) )
  • cpMvLX[ 1 ][ 1 ] ( mvScaleVer + dVerX * ( xCb + cbWidth - xNb ) + dVerY * ( yCb - yNb ) )
  • Equation 7 the variables mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, and dVerY are the same as in Equation 5, and a detailed description thereof will be omitted.
  • the third control point vector is a motion vector (MV) of the spatial neighboring block, a predetermined difference value, position information (xCb, yCb) of the current block, block size (width or height), or position information (xNb, yNb) of the spatial neighboring block ) It can be derived based on at least one of.
  • the block size may mean the size of the current block and/or spatial neighboring blocks. Since the motion vector and the difference value are the same as the first control point vector, a detailed description thereof will be omitted. However, the position of the motion vector used in the process of deriving the third control point vector, and the range and/or number of difference values may be different from those of the first control point vector or the second control point vector.
  • the third control point vector may be derived as in Equation 8 below.
  • cpMvLX[ 2 ][ 0 ] ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb + cbHeight - yNb ) )
  • cpMvLX[ 2 ][ 1 ] ( mvScaleVer + dVerX * ( xCb - xNb ) + dVerY * ( yCb + cbHeight - yNb ) )
  • Equation 8 the variables mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, and dVerY are the same as in Equation 5, and a detailed description thereof will be omitted. Meanwhile, through the above process, the first to N th control point vectors of spatial candidates may be derived.
  • the process of deriving the affine candidates described above may be performed for each of the pre-defined spatial neighboring blocks.
  • the pre-defined spatial neighboring block may include at least one of a left block, a lower left block, an upper right block, an upper block, or an upper left block of the current block.
  • the process of deriving the affine candidates may be performed for each group of the spatial neighboring blocks.
  • the spatial neighboring blocks may be classified into a first group including a left block and a lower left block and a second group including an upper right block, an upper block, and an upper left block.
  • one affine candidate may be derived from a spatial neighboring block belonging to the first group.
  • the derivation may be performed based on a predetermined priority order until an available affine candidate is found.
  • the priority order may be in the order of the left block -> the lower left block or vice versa. According to the priority order, it may be determined whether a corresponding spatial neighboring block in the first group is a block decoded through affine model-based prediction, and a block decoded by first affine model-based prediction may be selected as an affine candidate. there is.
  • one affine candidate may be derived from a spatial neighboring block belonging to the second group.
  • the derivation may be performed based on a predetermined priority order until an available affine candidate is found.
  • the priority order may be in the order of upper right block -> upper block -> upper left block, or vice versa. According to the priority order, it may be determined whether a corresponding spatial neighboring block in the second group is a block decoded through affine model-based prediction, and a block decoded by first affine model-based prediction may be selected as an affine candidate. there is.
  • the temporal neighboring block belongs to a different picture from the current block, but may be a block co-located with the current block.
  • a co-located block may be a block including the location of the top left sample of the current block, the center location, or the location of a sample adjacent to the bottom right sample of the current block.
  • the temporal neighboring block may refer to a block at a position shifted by a predetermined disparity vector from the block at the same position.
  • the disparity vector may be determined based on a motion vector of any one of spatial neighboring blocks of the above-described current block.
  • FIG. 6 illustrates a method of deriving a candidate constructed based on a combination of motion vectors of spatial/temporal neighboring blocks as an embodiment according to the present disclosure.
  • Candidates configured according to the present disclosure may be derived based on a combination of at least two of control point vectors (hereinafter referred to as control point vectors (cpMVCorner[n])) corresponding to each corner position of the current block.
  • cpMVCorner[n] control point vectors
  • n may be 0, 1, 2, or 3.
  • the control point vector cpMVCorner[n] may be derived based on a motion vector of a spatial neighboring block and/or a temporal neighboring block.
  • the spatial neighboring block is a first neighboring block (A, B or C) adjacent to the upper left sample of the current block, a second neighboring block (D or E) adjacent to the upper right sample of the current block, or a lower left sample of the current block. At least one of adjacent third neighboring blocks F or G may be included.
  • a temporal neighboring block is a block belonging to a picture different from the current block, and may refer to a block (hereinafter referred to as a fourth neighboring block Col) co-located with the current block.
  • the fourth neighboring block may mean a block (H, I, or J) including the position of the upper left sample, upper right sample, or lower left sample of the current block, or a block adjacent to the position of the lower right sample of the current block. there is.
  • the first neighboring block may refer to an upper left (A), upper (B), or left (C) neighboring block of the current block. It is determined whether the motion vectors of the neighboring blocks A, B, and C are available according to a predetermined priority order, and the control point vector may be determined using the motion vectors of the available neighboring blocks. The availability determination may be performed until a neighboring block having an available motion vector is found.
  • the order of priority may be A -> B -> C. However, it is not limited thereto, and may be in the order of A -> C -> B, C -> A -> B, B -> A -> C.
  • the second neighboring block may mean an upper (D) or upper right (E) neighboring block of the current block.
  • D upper
  • E upper right
  • the availability determination may be performed until a neighboring block having an available motion vector is found.
  • the order of priority may be in the order of D -> E or E -> D.
  • the third neighboring block may refer to a left (F) or lower left (G) neighboring block of the current block.
  • the availability determination may be performed until a neighboring block having an available motion vector is found.
  • the order of priority may be in the order of G -> F or F -> G.
  • the first control point vector cpMVCorner[0] may be set as a motion vector of a first neighboring block
  • the second control point vector (cpMVCorner[1]) may be set as a motion vector of a second neighboring block
  • the third control point vector (cpMVCorner[2]) may be set as a motion vector of a third neighboring block
  • the fourth control point vector (cpMVCorner[3]) may be set as a motion vector of a fourth neighboring block.
  • the first control point vector may be derived using a motion vector of at least one of the first neighboring block and the fourth neighboring block, wherein the fourth neighboring block is a block (H) including the position of the upper left sample.
  • the second control point vector may be derived using a motion vector of at least one of the second neighboring block and the fourth neighboring block.
  • the fourth neighboring block may be a block (I) including the position of the top right sample.
  • the third control point vector may be derived using a motion vector of at least one of the third neighboring block and the fourth neighboring block.
  • the fourth neighboring block may be a block (J) including the position of the lower left sample.
  • any one of the first to fourth control point vectors may be derived based on the other one.
  • the second control point vector may be derived by applying a predetermined offset vector to the first control point vector.
  • the offset vector may be a difference vector between the third control point vector and the first control point vector or may be derived by applying a predetermined scaling factor to the difference vector.
  • the scaling factor may be determined based on at least one of the width or height of the current block and/or neighboring blocks.
  • K configured candidates (ConstK) may be determined through a combination of at least two of the aforementioned first to fourth control point vectors.
  • the K value may be an integer of 1, 2, 3, 4, 5, 6, 7 or more.
  • the K value may be derived based on information signaled by the encoding device, or may be a pre-defined value identical to the encoding/decoding device.
  • the information may include information indicating the maximum number of configured candidates included in the candidate list.
  • the first configured candidate Const1 may be derived by combining the first to third control point vectors.
  • the first configured candidate Const1 may have a control point vector as shown in Table 9 below.
  • the control point vector may be configured as shown in Table 9.
  • the reference picture information may mean a reference picture index indicating a position of a corresponding reference picture in a reference picture list, or may mean a picture order count (POC) value indicating an output order.
  • POC picture order count
  • the second constructed candidate Const2 may be derived by combining the first, second and fourth control point vectors.
  • the second configured candidate Const2 may have a control point vector as shown in Table 10 below.
  • the control point vector may be configured as shown in Table 10.
  • reference picture information is as described above.
  • the third configured candidate Const3 may be derived by combining the first, third, and fourth control point vectors.
  • the third configured candidate Const3 may have a control point vector as shown in Table 11 below.
  • the control point vector may be configured as shown in Table 11.
  • reference picture information is as described above.
  • the fourth configured candidate Const4 may be derived by combining the second, third, and fourth control point vectors.
  • the fourth configured candidate Const4 may have a control point vector as shown in Table 12 below.
  • Table 12 shows the reference picture information of the second neighboring block.
  • reference picture information is as described above.
  • the fifth configured candidate Const5 may be derived by combining the first and second control point vectors.
  • the fifth configured candidate Const5 may have control point vectors as shown in Table 13 below.
  • the control point vector may be configured as shown in Table 13.
  • reference picture information is as described above.
  • Idx control point vector One cpMvCorner[ 0 ] 2 cpMvCorner[ 1 ]
  • the sixth configured candidate Const6 may be derived by combining the first and third control point vectors.
  • the sixth configured candidate Const6 may have control point vectors as shown in Table 14 below.
  • the control point vector may be configured as shown in Table 14.
  • reference picture information is as described above.
  • Idx control point vector control point vector One cpMvCorner[ 0 ] cpMvCorner[ 0 ] 2 cpMvCorner[ 2 ] cpMvCorner[ 1 ]
  • cpMvCorner[ 1 ] may be a second control point vector derived based on the first and third control point vectors.
  • the second control point vector may be derived based on at least one of the first control point vector, a predetermined difference value, and the size of the current/neighboring block.
  • the second control point vector may be derived as in Equation 9 below.
  • cpMvCorner[ 1 ][ 0 ] ( cpMvCorner[ 0 ][ 0 ] ⁇ 7 ) + ( ( cpMvCorner[ 2 ][ 1 ] - cpMvCorner[ 0 ][ 1 ] ) ⁇ ( 7 + Log2( cbHeight / cbWidth ) ) )
  • cpMvCorner[ 1 ][ 1 ] ( cpMvCorner[ 0 ][ 1 ] ⁇ 7 ) + ( ( cpMvCorner[ 2 ][ 0 ] - cpMvCorner[ 0 ][ 0 ] ) ⁇ ( 7 + Log2( cbHeight / cbWidth ) ) )
  • the K value does not limit the positions or priorities of the configured candidates arranged in the candidate list.
  • all of the first through sixth configured candidates may be included in the candidate list, or only some of them may be included in the candidate list.
  • the current block uses three control point vectors
  • only candidates configured through a combination of three of the first to fourth control point vectors may be used.
  • a configured candidate generated through a combination of at least two of the first to fourth control point vectors is used, or a candidate generated through a combination of only two of the first to fourth control point vectors Constructed candidates may also be used.
  • only partially configured candidates may be included in the candidate list in consideration of the maximum number of affine candidates included in the candidate list.
  • the maximum number may be determined based on maximum number information signaled by an encoding device or may be variably determined in consideration of the above-described attribute of the current block.
  • the K value of the configured candidate may mean a priority inserted into the candidate list.
  • the configured candidate may not be inserted into the candidate list.
  • the difference value between the two control point vectors may be divided into a difference value in a horizontal direction and a difference value in a vertical direction.
  • the difference value in the horizontal direction means the difference value between the first control point vector 610 and the second control point vector 620
  • the difference value in the vertical direction means the difference value between the first control point vector 610 and the third control point vector 630.
  • the threshold value may mean 0 or a vector having a pre-defined size equal to the encoding/decoding device.
  • FIG. 7 illustrates a method of deriving a motion vector in units of sub-blocks of a current block as an embodiment of the present disclosure.
  • the motion vector of the current block can be derived using the control point vector of the current block.
  • the current block may be divided into a plurality of sub-blocks, and motion vectors may be derived in units of sub-blocks.
  • the sub-block may have the same pre-defined size/shape in an encoding/decoding device.
  • the size/shape of a sub-block may be a square such as 4x4, 8x8, or 16x16 or a non-square such as 2x4, 2x8, 4x8, or 4x16.
  • the sub-block may be defined as a block in which the sum of width and height is an integer of 8, 12, 16, or more.
  • the sub-block may be defined as a block in which the product of width and height is an integer of 16, 32, 64 or more.
  • the size/shape of the sub-block may be variably derived based on the aforementioned attribute of the block. For example, if the size of the current block is greater than or equal to a predetermined threshold size, the current block is divided into first sub-block units (e.g., 8x8, 16x16); otherwise, the current block is divided into second sub-blocks. Can be divided into units of (e.g., 4x4).
  • first sub-block units e.g., 8x8, 16x16
  • second sub-blocks e.g., 4x4
  • information about the size/shape of a subblock may be encoded in an encoding device and signaled to a decoding device.
  • the information indicates at least one of the size or shape of the subblock, which may be signaled at the level of at least one of a sequence, picture, slice, tile, coding tree unit (CTU), or coding unit.
  • CTU coding tree unit
  • the control point representing the current block 700 may include an upper left corner position 711 and an upper right corner position 721 .
  • the control point may include three points of the upper left corner location 711, the upper right corner location 721, and the lower left corner location 731, or may further include a plurality of additional points.
  • a first control point vector 710 corresponding to the first control point 711, a second control point vector 720 corresponding to the second control point 721, and A motion vector for each sub-block of the current block may be derived using at least one of the position (x, y) or the size (w or h) of the current block.
  • a motion vector of each sub-block may be derived as in Equation 10 below.
  • the motion vectors for each subblock in the current block include the corrected first control point vector, the third control point vector, the position of the corresponding subblock, Alternatively, it may be derived using at least one of the heights of the current block.
  • the first control point vector 710 corresponding to the first control point 711, the second control point vector 720 corresponding to the second control point 721, and the third control point 731 A motion vector for each subblock of the current block may be derived using at least one of the third control point vector 730, the location (x, y) of the subblock, or the size (w or h) of the current block. .
  • a difference vector between the second control point vector and the first control point vector may be used, and a difference vector between the third control point vector and the first control point vector may be used.
  • the difference vector may be calculated for each of a horizontal direction (x-axis direction) and a vertical direction (y-axis direction).
  • FIG 8 illustrates a schematic configuration of an inter prediction unit 332 performing inter prediction based on an affine model as an embodiment according to the present disclosure.
  • the affine model-based inter prediction method performed in the decoding device has been reviewed, and this may be equally performed in the inter prediction unit 332 of the decoding device. Therefore, redundant descriptions will be omitted here.
  • the inter prediction unit 332 may include a control point vector derivation unit 800, a motion vector derivation unit 810, and a predicted sample generation unit 820.
  • the control point vector deriving unit 800 may generate a candidate list for predicting motion information of a current block.
  • the candidate list includes at least one of an affine candidate and a zero candidate, and the affine candidate may include at least one of an inherited candidate, a constructed candidate, a history-based candidate, an average candidate, and a combined candidate. .
  • a method of configuring and rearranging the candidate list has been reviewed in detail with reference to FIG. 4, and a detailed description thereof will be omitted.
  • the control point vector deriving unit 800 may derive the control point vector of the current block based on the candidate list and the candidate index.
  • the motion vector deriving unit 810 may derive a motion vector of the current block based on the derived control point vector.
  • the motion vector may be derived in units of subblocks of the current block, and for this purpose, the current block may be divided into a plurality of subblocks.
  • a method of deriving a motion vector in units of sub-blocks of the current block has been reviewed with reference to FIG. 7, and a detailed description thereof will be omitted herein.
  • the prediction sample generator 820 may perform inter prediction on the current block using the derived motion vector.
  • the inter prediction unit 332 may include a motion information derivation unit and a prediction sample generation unit.
  • the motion information derivation unit may generate a candidate list for the merge mode of the current block and rearrange it if necessary. This has been reviewed in detail with reference to FIG. 4 , and a detailed description thereof will be omitted.
  • the motion information derivation unit may derive motion information of a current block based on the candidate list and the candidate index.
  • the prediction sample generation unit may perform inter prediction on the current block using the derived motion information.
  • FIG 9 illustrates an affine model-based inter prediction method performed by an encoding apparatus as an embodiment according to the present disclosure.
  • an affine model-based inter prediction method performed by a decoding device has been described with reference to FIG. 4 , and this may be equally/similarly applied to an affine model-based inter prediction method performed by an encoding device. Therefore, redundant descriptions will be omitted here.
  • a candidate list for predicting motion information of a current block may be generated (S900).
  • the candidate list includes at least one of an affine candidate and a zero candidate, and the affine candidate may include at least one of an inherited candidate, a constructed candidate, a history-based candidate, an average candidate, and a combined candidate.
  • a method of configuring and rearranging the candidate list has been reviewed in detail with reference to FIG. 4, and a detailed description thereof will be omitted.
  • a control point vector of the current block may be determined based on the candidate list (S910).
  • An optimal candidate may be determined from among a plurality of candidates belonging to the candidate list, and a control point vector of the current block may be determined based on the determined optimal candidate.
  • a candidate index specifying an optimal candidate used to determine the control point vector of the current block may be encoded.
  • a motion vector of a current block may be derived based on the derived control point vector (S920).
  • the motion vector may be derived in units of subblocks of the current block, and for this purpose, the current block may be divided into a plurality of subblocks.
  • a method of deriving a motion vector in units of sub-blocks of the current block has been reviewed with reference to FIG. 7, and a detailed description thereof will be omitted herein.
  • inter prediction may be performed on the current block using the derived motion vector (S930).
  • a reference subblock corresponding to each subblock of the current block may be specified, and a prediction block of a corresponding subblock within the current block may be obtained based on the specified reference subblock.
  • a bitstream may be generated by encoding the current block based on a predicted block of the current block.
  • motion information of the current block may be determined based on one of a plurality of merge candidates included in the candidate list, and inter prediction may be performed based on the determined motion information.
  • a candidate index specifying any one of the plurality of merge candidates may be encoded.
  • FIG. 10 illustrates a schematic configuration of an inter prediction unit 221 performing inter prediction based on an affine model as an embodiment according to the present disclosure.
  • the inter prediction unit 221 may include a control point vector determination unit 1000, a motion vector derivation unit 1010, and a predicted sample generation unit 1020.
  • the control point vector determiner 1000 may generate a candidate list for predicting motion information of a current block.
  • the candidate list includes at least one of an affine candidate and a zero candidate, and the affine candidate may include at least one of an inherited candidate, a constructed candidate, a history-based candidate, an average candidate, and a combined candidate. .
  • a method of configuring and rearranging the candidate list has been reviewed in detail with reference to FIG. 4, and a detailed description thereof will be omitted.
  • the control point vector determiner 1000 may determine an optimal candidate from among a plurality of candidates belonging to the candidate list, and determine a control point vector of the current block based on the determined optimal candidate.
  • the entropy encoding unit 240 may encode a candidate index specifying an optimal candidate used to determine the control point vector of the current block and insert it into the bitstream.
  • the motion vector deriving unit 1010 may derive a motion vector of the current block based on the derived control point vector.
  • the motion vector may be derived in units of subblocks of the current block, and for this purpose, the current block may be divided into a plurality of subblocks.
  • a method of deriving a motion vector in units of sub-blocks of the current block has been reviewed with reference to FIG. 7, and a detailed description thereof will be omitted herein.
  • the prediction sample generator 1020 may perform inter prediction on the current block using the derived motion vector.
  • the inter prediction unit 221 may include a motion information determination unit and a prediction sample generation unit.
  • the motion information determiner may determine motion information of a current block based on one of a plurality of merge candidates included in a candidate list, and perform inter prediction based on the determined motion information.
  • the entropy encoding unit 240 may encode a candidate index specifying an optimal merge candidate used to determine the motion vector of the current block and insert it into the bitstream.
  • the prediction sample generation unit may perform inter prediction on the current block using the determined motion information.
  • the above-described method according to the embodiments of this document may be implemented in the form of software, and the encoding device and/or decoding device according to this document may be used to display images of, for example, a TV, computer, smartphone, set-top box, display device, etc. It can be included in the device that performs the processing.
  • a module can be stored in memory and executed by a processor.
  • the memory may be internal or external to the processor, and may be coupled with the processor in a variety of well-known means.
  • a processor may include an application-specific integrated circuit (ASIC), other chipsets, logic circuits, and/or data processing devices.
  • Memory may include read-only memory (ROM), random access memory (RAM), flash memory, memory cards, storage media, and/or other storage devices. That is, the embodiments described in this document may be implemented and performed on a processor, microprocessor, controller, or chip. For example, functional units shown in each drawing may be implemented and performed on a computer, processor, microprocessor, controller, or chip. In this case, information for implementation (eg, information on instructions) or an algorithm may be stored in a digital storage medium.
  • a decoding device and an encoding device to which the embodiment(s) of the present specification are applied may be used in a multimedia broadcasting transceiving device, a mobile communication terminal, a home cinema video device, a digital cinema video device, a surveillance camera, a video conversation device, and a video communication device.
  • Real-time communication device mobile streaming device, storage medium, camcorder, video-on-demand (VoD) service providing device, OTT video (Over the top video) device, Internet streaming service providing device, 3D (3D) video device, VR (virtual reality) ) device, AR (argumente reality) device, video phone video device, transportation terminal (ex.
  • OTT video devices may include game consoles, Blu-ray players, Internet-connected TVs, home theater systems, smart phones, tablet PCs, digital video recorders (DVRs), and the like.
  • the processing method to which the embodiment (s) of the present specification is applied may be produced in the form of a program executed by a computer and stored in a computer-readable recording medium.
  • Multimedia data having a data structure according to the embodiment(s) of the present specification may also be stored in a computer-readable recording medium.
  • the computer-readable recording medium includes all types of storage devices and distributed storage devices in which computer-readable data is stored.
  • the computer-readable recording medium includes, for example, Blu-ray Disc (BD), Universal Serial Bus (USB), ROM, PROM, EPROM, EEPROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical A data storage device may be included.
  • the computer-readable recording medium includes media implemented in the form of a carrier wave (eg, transmission through the Internet).
  • the bitstream generated by the encoding method may be stored in a computer-readable recording medium or transmitted through a wired or wireless communication network.
  • embodiment(s) of the present specification may be implemented as a computer program product using program codes, and the program code may be executed on a computer by the embodiment(s) of the present specification.
  • the program code may be stored on a carrier readable by a computer.
  • FIG. 11 shows an example of a content streaming system to which embodiments of the present disclosure may be applied.
  • a content streaming system to which the embodiment(s) of the present specification is applied may largely include an encoding server, a streaming server, a web server, a media storage, a user device, and a multimedia input device.
  • the encoding server compresses content input from multimedia input devices such as smart phones, cameras, camcorders, etc. into digital data to generate a bitstream and transmits it to the streaming server.
  • multimedia input devices such as smart phones, cameras, and camcorders directly generate bitstreams
  • the encoding server may be omitted.
  • the bitstream may be generated by an encoding method or a bitstream generation method to which the embodiment(s) of the present specification is applied, and the streaming server temporarily stores the bitstream in a process of transmitting or receiving the bitstream.
  • the streaming server transmits multimedia data to a user device based on a user request through a web server, and the web server serves as a medium informing a user of what kind of service is available.
  • the web server transmits the request to the streaming server, and the streaming server transmits multimedia data to the user.
  • the content streaming system may include a separate control server, and in this case, the control server serves to control commands/responses between devices in the content streaming system.
  • the streaming server may receive content from a media storage and/or encoding server. For example, when content is received from the encoding server, the content can be received in real time. In this case, in order to provide smooth streaming service, the streaming server may store the bitstream for a certain period of time.
  • Examples of the user devices include mobile phones, smart phones, laptop computers, digital broadcasting terminals, personal digital assistants (PDAs), portable multimedia players (PMPs), navigation devices, slate PCs, Tablet PC, ultrabook, wearable device (e.g., smartwatch, smart glass, HMD (head mounted display)), digital TV, desktop There may be computers, digital signage, and the like.
  • PDAs personal digital assistants
  • PMPs portable multimedia players
  • navigation devices slate PCs
  • Tablet PC ultrabook
  • wearable device e.g., smartwatch, smart glass, HMD (head mounted display)
  • digital TV desktop There may be computers, digital signage, and the like.
  • Each server in the content streaming system may be operated as a distributed server, and in this case, data received from each server may be distributed and processed.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

본 개시에 따른 영상 디코딩/인코딩 방법 및 장치는, 현재 블록의 후보 리스트를 생성하고, 현재 블록의 후보 리스트에 평균 후보를 추가하고, 평균 후보를 포함한 후보 리스트에 기초하여 현재 블록의 인터 예측을 수행할 수 있다. 여기서, 후보 리스트는 복수의 어파인 후보를 포함하고, 복수의 어파인 후보는 상속된 후보, 구성된 후보, 또는 히스토리 기반의 후보 중 적어도 하나를 포함하고, 평균 후보는 후보 리스트로부터 소정의 우선순서에 따라 선택된 2개의 어파인 후보에 기초하여 생성될 수 있다.

Description

영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체
본 발명은 영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체에 관한 것이다.
최근 HD(High Definition) 영상 및 UHD(Ultra High Definition) 영상과 같은 고해상도, 고품질의 영상에 대한 수요가 다양한 응용 분야에서 증가하고 있고, 이에 따라 고효율의 영상 압축 기술들이 논의되고 있다.
영상 압축 기술로 현재 픽쳐의 이전 또는 이후 픽쳐로부터 현재 픽쳐에 포함된 화소값을 예측하는 인터 예측 기술, 현재 픽쳐 내의 화소 정보를 이용하여 현재 픽쳐에 포함된 화소값을 예측하는 인트라 예측 기술, 출현 빈도가 높은 값에 짧은 부호를 할당하고 출현 빈도가 낮은 값에 긴 부호를 할당하는 엔트로피 부호화 기술 등 다양한 기술이 존재하고 이러한 영상 압축 기술을 이용해 영상 데이터를 효과적으로 압축하여 전송 또는 저장할 수 있다.
본 개시는, 어파인 모드를 위한 다양한 어파인 후보를 생성하는 방법 및 장치를 제공하고자 한다.
본 개시는, 어파인 모드를 위한 후보 리스트를 구성/재배열하는 방법 및 장치를 제공하고자 한다.
본 개시는, 어파인 후보 간의 중복 체크에 기반하여 후보 리스트를 구성하는 방법 및 장치를 제공하고자 한다.
본 개시는, 어파인 후보의 움직임 정보를 유도하는 방법 및 장치를 제공하고자 한다.
본 개시는, 머지 모드를 위한 머지 후보를 생성하는 방법 및 장치를 제공하고자 한다.
본 개시는, 머지 모드를 위한 후보 리스트를 구성/재배열하는 방법 및 장치를 제공하고자 한다.
본 개시에 따른 영상 디코딩 방법 및 장치는, 현재 블록의 후보 리스트를 생성하고, 상기 후보 리스트에 평균 후보를 추가하고, 상기 평균 후보를 포함한 상기 후보 리스트에 기초하여 상기 현재 블록의 인터 예측을 수행할 수 있다. 여기서, 상기 후보 리스트는 복수의 어파인 후보를 포함하고, 상기 복수의 어파인 후보는 상속된 후보, 구성된 후보, 또는 히스토리 기반의 후보 중 적어도 하나를 포함할 수 있다. 상기 평균 후보는, 상기 후보 리스트로부터 소정의 우선순서에 따라 선택된 2개의 어파인 후보에 기초하여 생성될 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 상기 선택된 2개의 어파인 후보가 서로 다른 어파인 모델의 타입을 가지는 경우, 상기 평균 후보에 대한 어파인 모델의 타입은, 상기 선택된 2개의 어파인 후보 중 어느 하나에 대한 어파인 모델의 타입으로 결정될 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치는, 상기 후보 리스트에 속한 복수의 어파인 후보 중 적어도 하나에 대한 어파인 모델의 타입을 고려하여, 상기 후보 리스트에 속한 복수의 어파인 후보를 재배열할 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 상기 평균 후보는, 상기 2개의 어파인 후보의 이용 가능성, 상기 2개의 어파인 후보 간 참조 픽쳐 인덱스의 동일 여부, 또는 상기 2개의 어파인 후보 간 참조 픽쳐의 동일 여부 중 적어도 하나에 기초하여, 상기 후보 리스트에 적응적으로 추가될 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 상기 상속된 후보는, 상기 현재 블록의 공간적 주변 블록이 어파인 모드로 부호화된 블록인지 여부, 상기 후보 리스트에 기-포함된 상속된 후보의 개수, 또는 상기 현재 블록의 형태 중 적어도 하나를 고려하여, 상기 후보 리스트에 구성될 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 상기 평균 후보는, 상기 후보 리스트에 속한 복수의 어파인 후보 중 적어도 하나와의 중복 체크에 기초하여 상기 후보 리스트에 추가되고, 상기 중복 체크는, 상기 평균 후보에 대한 어파인 모델의 타입에 기초하여 수행될 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 템플릿 기반의 코스트에 기초하여, 상기 후보 리스트에 포함된 상기 상속된 후보, 상기 구성된 후보, 또는 상기 평균 후보 중 적어도 하나에 대해서 재배열이 수행될 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 상기 평균 후보의 움직임 정보는, 상기 선택된 2개의 어파인 후보 중 적어도 하나의 움직임 정보에 기초하여 결정되고, 상기 움직임 정보는, 지역 조도 보상이 적용되는지 여부를 나타내는 플래그 또는 양방향 가중 예측을 위한 가중치 정보 중 적어도 하나를 포함할 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치는, 비트스트림으로부터 상기 현재 블록의 후보 인덱스를 획득할 수 있고, 상기 후보 인덱스는, 상기 평균 후보를 포함한 상기 후보 리스트 내에서 일부 어파인 후보 중 어느 하나를 특정할 수 있다.
본 개시에 따른 영상 디코딩 방법 및 장치에 있어서, 상기 후보 리스트는, 조합된 후보를 더 포함하고, 상기 조합된 후보는, 상기 복수의 어파인 후보 중 어느 하나에 대한 L0 예측 방향의 움직임 정보와 상기 복수의 어파인 후보 중 다른 하나에 대한 L1 예측 방향의 움직임 정보를 조합하여 생성될 수 있다.
본 개시에 따른 영상 인코딩 방법 및 장치는, 현재 블록의 후보 리스트를 생성하고, 상기 후보 리스트에 평균 후보를 추가하며, 상기 평균 후보를 포함한 상기 후보 리스트에 기초하여 상기 현재 블록의 인터 예측을 수행할 수 있다. 여기서, 상기 후보 리스트는 복수의 어파인 후보를 포함하고, 상기 복수의 어파인 후보는 상속된 후보, 구성된 후보, 또는 히스토리 기반의 후보 중 적어도 하나를 포함할 수 있다. 상기 평균 후보는, 상기 후보 리스트로부터 소정의 우선순서에 따라 선택된 2개의 어파인 후보에 기초하여 생성될 수 있다.
본 개시에 따른 디코딩 장치에 의하여 영상 디코딩 방법을 수행하도록 야기하는 인코딩된 비디오/영상 정보가 저장된 컴퓨터 판독 가능한 디지털 저장 매체가 제공된다.
본 개시에 따른 영상 인코딩 방법에 따라 생성된 비디오/영상 정보가 저장된 컴퓨터 판독 가능한 디지털 저장 매체가 제공된다.
본 개시에 따른 영상 인코딩 방법에 따라 생성된 비디오/영상 정보를 전송하는 방법 및 장치가 제공된다.
본 개시에 따르면, 다양한 어파인 후보 또는 머지 후보를 생성함으로써, 인터 예측의 정확도 및 신뢰도를 향상시킬 수 있다.
본 개시에 따르면, 중복 체크를 통해 후보 리스트를 구성하고, 이를 재배열함으로써, 후보 인덱스의 부호화 효율을 향상시킬 수 있다.
본 개시에 따르면, 어파인 모델의 타입을 고려하여 평균 후보 또는 조합된 후보의 생성을 위한 대상 후보를 결정함으로써, 평균 후보 또는 조합된 후보의 신뢰도를 높이고, 인터 예측의 정확도를 향상시킬 수 있다.
도 1은 본 개시에 따른 비디오/영상 코딩 시스템을 도시한 것이다.
도 2는 본 개시의 실시예가 적용될 수 있고, 비디오/영상 신호의 인코딩이 수행되는 인코딩 장치의 개략적인 블록도를 나타낸다.
도 3은 본 개시의 실시예가 적용될 수 있고, 비디오/영상 신호의 디코딩이 수행되는 디코딩 장치의 개략적인 블록도를 나타낸다.
도 4는 본 개시에 따른 일실시예로서, 디코딩 장치에 의해 수행되는 어파인 모델 기반의 인터 예측 방법을 도시한 것이다.
도 5는 본 개시에 따른 일실시예로서, 공간적/시간적 주변 블록의 제어점 벡터로부터 어파인 후보를 유도하는 방법에 관한 것이다.
도 6은 본 개시에 따른 일실시예로서, 공간적/시간적 주변 블록의 움직임 벡터의 조합에 기초하여 구성된 후보를 유도하는 방법을 도시한 것이다.
도 7은 본 개시의 일실시예로서, 현재 블록의 서브 블록 단위로 움직임 벡터를 유도하는 방법을 도시한 것이다.
도 8은 본 개시에 따른 일실시예로서, 어파인 모델 기반의 인터 예측을 수행하는 인터 예측부(332)의 개략적인 구성을 도시한 것이다.
도 9는 본 개시에 따른 일실시예로서, 인코딩 장치에 의해 수행되는 어파인 모델 기반의 인터 예측 방법을 도시한 것이다.
도 10은 본 개시에 따른 일실시예로서, 어파인 모델 기반의 인터 예측을 수행하는 인터 예측부(221)의 개략적인 구성을 도시한 것이다.
도 11은 본 개시의 실시예들이 적용될 수 있는 컨텐츠 스트리밍 시스템의 예를 나타낸다.
본 개시는 다양한 변경을 가할 수 있고 여러 가지 실시예를 가질 수 있는 바, 특정 실시예들을 도면에 예시하고 상세한 설명에 상세하게 설명하고자 한다. 그러나, 이는 본 개시를 특정한 실시 형태에 대해 한정하려는 것이 아니며, 본 개시의 사상 및 기술 범위에 포함되는 모든 변경, 균등물 내지 대체물을 포함하는 것으로 이해되어야 한다. 각 도면을 설명하면서 유사한 참조부호를 유사한 구성요소에 대해 사용하였다.
제1, 제2 등의 용어는 다양한 구성요소들을 설명하는데 사용될 수 있지만, 상기 구성요소들은 상기 용어들에 의해 한정되어서는 안 된다. 상기 용어들은 하나의 구성요소를 다른 구성요소로부터 구별하는 목적으로만 사용된다. 예를 들어, 본 개시의 권리 범위를 벗어나지 않으면서 제1 구성요소는 제2 구성요소로 명명될 수 있고, 유사하게 제2 구성요소도 제1 구성요소로 명명될 수 있다. 및/또는 이라는 용어는 복수의 관련된 기재된 항목들의 조합 또는 복수의 관련된 기재된 항목들 중의 어느 항목을 포함한다.
어떤 구성요소가 다른 구성요소에 "연결되어" 있다거나 "접속되어"있다고 언급된 때에는, 그 다른 구성요소에 직접적으로 연결되어 있거나 또는 접속되어 있을 수도 있지만, 중간에 다른 구성요소가 존재할 수도 있다고 이해되어야 할 것이다. 반면에, 어떤 구성요소가 다른 구성요소에 "직접 연결되어"있다거나 "직접 접속되어"있다고 언급된 때에는, 중간에 다른 구성요소가 존재하지 않는 것으로 이해되어야 할 것이다.
본 출원에서 사용한 용어는 단지 특정한 실시예를 설명하기 위해 사용된 것으로, 본 개시를 한정하려는 의도가 아니다. 단수의 표현은 문맥상 명백하게 다르게 뜻하지 않는 한, 복수의 표현을 포함한다. 본 출원에서, "포함하다" 또는 "가지다" 등의 용어는 명세서상에 기재된 특징, 숫자, 단계, 동작, 구성요소, 부품 또는 이들을 조합한 것이 존재함을 지정하려는 것이지, 하나 또는 그 이상의 다른 특징들이나 숫자, 단계, 동작, 구성요소, 부품 또는 이들을 조합한 것들의 존재 또는 부가 가능성을 미리 배제하지 않는 것으로 이해되어야 한다.
본 개시는 비디오/영상 코딩에 관한 것이다. 예를 들어, 본 명세서에서 개시된 방법/실시예는 VVC (versatile video coding) 표준에 개시되는 방법에 적용될 수 있다. 또한, 이 명세서에서 개시된 방법/실시예는 EVC (essential video coding) 표준, AV1 (AOMedia Video 1) 표준, AVS2 (2nd generation of audio video coding standard) 또는 차세대 비디오/영상 코딩 표준(ex. H.267 or H.268 등)에 개시되는 방법에 적용될 수 있다.
이 명세서는 비디오/영상 코딩에 관한 다양한 실시예들을 제시하며, 다른 언급이 없는 한 상기 실시예들은 서로 조합되어 수행될 수도 있다.
이 명세서에서 비디오(video)는 시간의 흐름에 따른 일련의 영상(image)들의 집합을 의미할 수 있다. 픽쳐(picture)는 일반적으로 특정 시간대의 하나의 영상을 나타내는 단위를 의미하며, 슬라이스(slice)/타일(tile)은 코딩에 있어서 픽쳐의 일부를 구성하는 단위이다. 슬라이스/타일은 하나 이상의 CTU(coding tree unit)을 포함할 수 있다. 하나의 픽쳐는 하나 이상의 슬라이스/타일로 구성될 수 있다. 하나의 타일은 하나의 픽쳐의 특정 타일 열과 특정 타일 행 내에 있는 복수의 CTU들로 구성된 직사각형 영역이다. 타일 열은 픽쳐의 높이와 동일한 높이와 픽쳐 파라미터 세트의 신택스 요구에 의해 지정된 너비를 갖는 CTU들의 직사각형 영역이다. 타일 행은 픽쳐 파라미터 세트에 의해 지정된 높이와 픽쳐의 너비와 동일한 너비를 갖는 CTU들의 직사각형 영역이다. 하나의 타일 내에 CTU들은 CTU 래스터 스캔에 따라 연속적으로 배열되는 반면, 하나의 픽쳐 내 타일들은 타일의 래스터 스캔에 따라 연속적으로 배열될 수 있다. 하나의 슬라이스는 단일 NAL 유닛에 배타적으로 포함될 수 있는 픽쳐의 타일 내에서 정수 개수의 완전한 타일 또는 정수 개수의 연속적인 완전한 CTU 행을 포함할 수 있다. 한편, 하나의 픽쳐는 둘 이상의 서브픽쳐로 구분될 수 있다. 서브픽쳐는 픽쳐 내 하나 이상의 슬라이스들의 직사각형 영역일 수 있다.
화소, 픽셀(pixel) 또는 펠(pel)은 하나의 픽쳐(또는 영상)을 구성하는 최소의 단위를 의미할 수 있다. 또한, 픽셀에 대응하는 용어로서 '샘플(sample)'이 사용될 수 있다. 샘플은 일반적으로 픽셀 또는 픽셀의 값을 나타낼 수 있으며, 휘도(luma) 성분의 픽셀/픽셀값만을 나타낼 수도 있고, 색차(chroma) 성분의 픽셀/픽셀 값만을 나타낼 수도 있다.
유닛(unit)은 영상 처리의 기본 단위를 나타낼 수 있다. 유닛은 픽쳐의 특정 영역 및 해당 영역에 관련된 정보 중 적어도 하나를 포함할 수 있다. 하나의 유닛은 하나의 루마 블록 및 두개의 크로마(ex. cb, cr) 블록을 포함할 수 있다. 유닛은 경우에 따라서 블록(block) 또는 영역(area) 등의 용어와 혼용하여 사용될 수 있다. 일반적인 경우, MxN 블록은 M개의 열과 N개의 행으로 이루어진 샘플들(또는 샘플 어레이) 또는 변환 계수(transform coefficient)들의 집합(또는 어레이)을 포함할 수 있다.
본 명세서에서 "A 또는 B(A or B)"는 "오직 A", "오직 B" 또는 "A와 B 모두"를 의미할 수 있다. 달리 표현하면, 본 명세서에서 "A 또는 B(A or B)"는 "A 및/또는 B(A and/or B)"으로 해석될 수 있다. 예를 들어, 본 명세서에서 "A, B 또는 C(A, B or C)"는 "오직 A", "오직 B", "오직 C", 또는 "A, B 및 C의 임의의 모든 조합(any combination of A, B and C)"를 의미할 수 있다.
본 명세서에서 사용되는 슬래쉬(/)나 쉼표(comma)는 "및/또는(and/or)"을 의미할 수 있다. 예를 들어, "A/B"는 "A 및/또는 B"를 의미할 수 있다. 이에 따라 "A/B"는 "오직 A", "오직 B", 또는 "A와 B 모두"를 의미할 수 있다. 예를 들어, "A, B, C"는 "A, B 또는 C"를 의미할 수 있다.
본 명세서에서 "적어도 하나의 A 및 B(at least one of A and B)"는, "오직 A", "오직 B" 또는 "A와 B 모두"를 의미할 수 있다. 또한, 본 명세서에서 "적어도 하나의 A 또는 B(at least one of A or B)"나 "적어도 하나의 A 및/또는 B(at least one of A and/or B)"라는 표현은 "적어도 하나의 A 및 B(at least one of A and B)"와 동일하게 해석될 수 있다.
또한, 본 명세서에서 "적어도 하나의 A, B 및 C(at least one of A, B and C)"는, "오직 A", "오직 B", "오직 C", 또는 "A, B 및 C의 임의의 모든 조합(any combination of A, B and C)"를 의미할 수 있다. 또한, "적어도 하나의 A, B 또는 C(at least one of A, B or C)"나 "적어도 하나의 A, B 및/또는 C(at least one of A, B and/or C)"는 "적어도 하나의 A, B 및 C(at least one of A, B and C)"를 의미할 수 있다.
또한, 본 명세서에서 사용되는 괄호는 "예를 들어(for example)"를 의미할 수 있다. 구체적으로, "예측(인트라 예측)"로 표시된 경우, "예측"의 일례로 "인트라 예측"이 제안된 것일 수 있다. 달리 표현하면 본 명세서의 "예측"은 "인트라 예측"으로 제한(limit)되지 않고, "인트라 예측"이 "예측"의 일례로 제안될 것일 수 있다. 또한, "예측(즉, 인트라 예측)"으로 표시된 경우에도, "예측"의 일례로 "인트라 예측"이 제안된 것일 수 있다.
본 명세서에서 하나의 도면 내에서 개별적으로 설명되는 기술적 특징은, 개별적으로 구현될 수도 있고, 동시에 구현될 수도 있다.
도 1은 본 개시에 따른 비디오/영상 코딩 시스템을 도시한 것이다.
도 1을 참조하면, 비디오/영상 코딩 시스템은 제1 장치(소스 디바이스) 및 제2 장치(수신 디바이스)를 포함할 수 있다.
소스 디바이스는 인코딩된 비디오(video)/영상(image) 정보 또는 데이터를 파일 또는 스트리밍 형태로 디지털 저장매체 또는 네트워크를 통하여 수신 디바이스로 전달할 수 있다. 상기 소스 디바이스는 비디오 소스, 인코딩 장치, 전송부를 포함할 수 있다. 상기 수신 디바이스는 수신부, 디코딩 장치 및 렌더러를 포함할 수 있다. 상기 인코딩 장치는 비디오/영상 인코딩 장치라고 불릴 수 있고, 상기 디코딩 장치는 비디오/영상 디코딩 장치라고 불릴 수 있다. 송신기는 인코딩 장치에 포함될 수 있다. 수신기는 디코딩 장치에 포함될 수 있다. 렌더러는 디스플레이부를 포함할 수도 있고, 디스플레이부는 별개의 디바이스 또는 외부 컴포넌트로 구성될 수도 있다.
비디오 소스는 비디오/영상의 캡쳐, 합성 또는 생성 과정 등을 통하여 비디오/영상을 획득할 수 있다. 비디오 소스는 비디오/영상 캡쳐 디바이스 및/또는 비디오/영상 생성 디바이스를 포함할 수 있다. 비디오/영상 캡쳐 디바이스는 하나 이상의 카메라, 이전에 캡쳐된 비디오/영상을 포함하는 비디오/영상 아카이브 등을 포함할 수 있다. 비디오/영상 생성 디바이스는 컴퓨터, 타블렛 및 스마트폰 등을 포함할 수 있으며 (전자적으로) 비디오/영상을 생성할 수 있다. 예를 들어, 컴퓨터 등을 통하여 가상의 비디오/영상이 생성될 수 있으며, 이 경우 관련 데이터가 생성되는 과정으로 비디오/영상 캡쳐 과정이 갈음될 수 있다.
인코딩 장치는 입력 비디오/영상을 인코딩할 수 있다. 인코딩 장치는 압축 및 코딩 효율을 위하여 예측, 변환, 양자화 등 일련의 절차를 수행할 수 있다. 인코딩된 데이터(인코딩된 비디오/영상 정보)는 비트스트림(bitstream) 형태로 출력될 수 있다.
전송부는 비트스트림 형태로 출력된 인코딩된 비디오/영상 정보 또는 데이터를 파일 또는 스트리밍 형태로 디지털 저장매체 또는 네트워크를 통하여 수신 디바이스의 수신부로 전달할 수 있다. 디지털 저장 매체는 USB, SD, CD, DVD, 블루레이, HDD, SSD 등 다양한 저장 매체를 포함할 수 있다. 전송부는 미리 정해진 파일 포멧을 통하여 미디어 파일을 생성하기 위한 엘리먼트를 포함할 수 있고, 방송/통신 네트워크를 통한 전송을 위한 엘리먼트를 포함할 수 있다. 수신부는 상기 비트스트림을 수신/추출하여 디코딩 장치로 전달할 수 있다.
디코딩 장치는 인코딩 장치의 동작에 대응하는 역양자화, 역변환, 예측 등 일련의 절차를 수행하여 비디오/영상을 디코딩할 수 있다.
렌더러는 디코딩된 비디오/영상을 렌더링할 수 있다. 렌더링된 비디오/영상은 디스플레이부를 통하여 디스플레이될 수 있다.
도 2는 본 개시의 실시예가 적용될 수 있고, 비디오/영상 신호의 인코딩이 수행되는 인코딩 장치의 개략적인 블록도를 나타낸다.
도 2를 참조하면, 인코딩 장치(200)는 영상 분할부(image partitioner, 210), 예측부(predictor, 220), 레지듀얼 처리부(residual processor, 230), 엔트로피 인코딩부(entropy encoder, 240), 가산부(adder, 250), 필터링부(filter, 260) 및 메모리(memory, 270)를 포함하여 구성될 수 있다. 예측부(220)는 인터 예측부(221) 및 인트라 예측부(222)를 포함할 수 있다. 레지듀얼 처리부(230)는 변환부(transformer, 232), 양자화부(quantizer 233), 역양자화부(dequantizer 234), 역변환부(inverse transformer, 235)를 포함할 수 있다. 레지듀얼 처리부(230)은 감산부(subtractor, 231)를 더 포함할 수 있다. 가산부(250)는 복원부(reconstructor) 또는 복원 블록 생성부(recontructged block generator)로 불릴 수 있다. 상술한 영상 분할부(210), 예측부(220), 레지듀얼 처리부(230), 엔트로피 인코딩부(240), 가산부(250) 및 필터링부(260)는 실시예에 따라 하나 이상의 하드웨어 컴포넌트(예를 들어 인코딩 장치 칩셋 또는 프로세서)에 의하여 구성될 수 있다. 또한 메모리(270)는 DPB(decoded picture buffer)를 포함할 수 있고, 디지털 저장 매체에 의하여 구성될 수도 있다. 상기 하드웨어 컴포넌트는 메모리(270)을 내/외부 컴포넌트로 더 포함할 수도 있다.
영상 분할부(210)는 인코딩 장치(200)에 입력된 입력 영상(또는, 픽쳐, 프레임)를 하나 이상의 처리 유닛(processing unit)으로 분할할 수 있다. 일 예로, 상기 처리 유닛은 코딩 유닛(coding unit, CU)이라고 불릴 수 있다. 이 경우 코딩 유닛은 코딩 트리 유닛(coding tree unit, CTU) 또는 최대 코딩 유닛(largest coding unit, LCU)으로부터 QTBTTT (Quad-tree binary-tree ternary-tree) 구조에 따라 재귀적으로(recursively) 분할될 수 있다.
예를 들어, 하나의 코딩 유닛은 쿼드 트리 구조, 바이너리 트리 구조, 및/또는 터너리 구조를 기반으로 하위(deeper) 뎁스를 가진 복수의 코딩 유닛들로 분할될 수 있다. 이 경우, 예를 들어 쿼드 트리 구조가 먼저 적용되고 바이너리 트리 구조 및/또는 터너리 구조가 나중에 적용될 수 있다. 또는, 바이너리 트리 구조가 쿼드 트리 구조보다 먼저 적용될 수도 있다. 더 이상 분할되지 않는 최종 코딩 유닛을 기반으로 본 명세서에 따른 코딩 절차가 수행될 수 있다. 이 경우 영상 특성에 따른 코딩 효율 등을 기반으로, 최대 코딩 유닛이 바로 최종 코딩 유닛으로 사용될 수 있고, 또는 필요에 따라 코딩 유닛은 재귀적으로(recursively) 하위 뎁스의 코딩 유닛들로 분할되어, 최적의 크기를 가진 코딩 유닛이 최종 코딩 유닛으로 사용될 수 있다. 여기서, 코딩 절차라 함은 후술하는 예측, 변환, 및 복원 등의 절차를 포함할 수 있다.
다른 예로, 상기 처리 유닛은 예측 유닛(PU: Prediction Unit) 또는 변환 유닛(TU: Transform Unit)을 더 포함할 수 있다. 이 경우, 상기 예측 유닛 및 상기 변환 유닛은 각각 상술한 최종 코딩 유닛으로부터 분할 또는 파티셔닝될 수 있다. 상기 예측 유닛은 샘플 예측의 단위일 수 있고, 상기 변환 유닛은 변환 계수를 유도하는 단위 및/또는 변환 계수로부터 레지듀얼 신호(residual signal)를 유도하는 단위일 수 있다.
유닛은 경우에 따라서 블록(block) 또는 영역(area) 등의 용어와 혼용하여 사용될 수 있다. 일반적인 경우, MxN 블록은 M개의 열과 N개의 행으로 이루어진 샘플들 또는 변환 계수(transform coefficient)들의 집합을 나타낼 수 있다. 샘플은 일반적으로 픽셀 또는 픽셀의 값을 나타낼 수 있으며, 휘도(luma) 성분의 픽셀/픽셀값만을 나타낼 수도 있고, 색차(chroma) 성분의 픽셀/픽셀 값만을 나타낼 수도 있다. 샘플은 하나의 픽쳐(또는 영상)을 픽셀(pixel) 또는 펠(pel)에 대응하는 용어로서 사용될 수 있다.
인코딩 장치(200)는 입력 영상 신호(원본 블록, 원본 샘플 어레이)에서 인터 예측부(221) 또는 인트라 예측부(222)로부터 출력된 예측 신호(예측 블록, 예측 샘플 어레이)를 감산하여 레지듀얼 신호(residual signal, 잔여 블록, 잔여 샘플 어레이)를 생성할 수 있고, 생성된 레지듀얼 신호는 변환부(232)로 전송된다. 이 경우, 인코딩 장치(200) 내에서 입력 영상 신호(원본 블록, 원본 샘플 어레이)에서 예측 신호(예측 블록, 예측 샘플 어레이)를 감산하는 유닛은 감산부(231)라고 불릴 수 있다.
예측부(220)는 처리 대상 블록(이하, 현재 블록이라 함)에 대한 예측을 수행하고, 상기 현재 블록에 대한 예측 샘플들을 포함하는 예측 블록(predicted block)을 생성할 수 있다. 예측부(220)는 현재 블록 또는 CU 단위로 인트라 예측이 적용되는지 또는 인터 예측이 적용되는지 결정할 수 있다. 예측부(220)는 각 예측 모드에 대한 설명에서 후술하는 바와 같이 예측 모드 정보 등 예측에 관한 다양한 정보를 생성하여 엔트로피 인코딩부(240)로 전달할 수 있다. 예측에 관한 정보는 엔트로피 인코딩부(240)에서 인코딩되어 비트스트림 형태로 출력될 수 있다.
인트라 예측부(222)는 현재 픽쳐 내의 샘플들을 참조하여 현재 블록을 예측할 수 있다. 상기 참조되는 샘플들은 예측 모드에 따라 상기 현재 블록의 주변(neighbor)에 위치할 수 있고, 또는 현재 블록으로부터 일정 거리만큼 떨어져서 위치할 수도 있다. 인트라 예측에서 예측 모드들은 하나 이상의 비방향성 모드와 복수의 방향성 모드를 포함할 수 있다. 비방향성 모드는, DC 모드 또는 플래너 모드(Planar 모드) 중 적어도 하나를 포함할 수 있다. 방향성 모드는 예측 방향의 세밀한 정도에 따라 33개의 방향성 모드 또는 65개의 방향성 모드를 포함할 수 있다. 다만, 이는 예시로서 설정에 따라 그 이상 또는 그 이하의 개수의 방향성 모드들이 사용될 수 있다. 인트라 예측부(222)는 주변 블록에 적용된 예측 모드를 이용하여, 현재 블록에 적용되는 예측 모드를 결정할 수도 있다.
인터 예측부(221)는 참조 픽쳐 상에서 움직임 벡터에 의해 특정되는 참조 블록(참조 샘플 어레이)을 기반으로, 현재 블록에 대한 예측 블록을 유도할 수 있다. 이때, 인터 예측 모드에서 전송되는 움직임 정보의 양을 줄이기 위해, 주변 블록과 현재 블록 간의 움직임 정보의 상관성에 기초하여 움직임 정보를 블록, 서브블록 또는 샘플 단위로 예측할 수 있다. 상기 움직임 정보는 움직임 벡터 및 참조 픽쳐 인덱스를 포함할 수 있다. 상기 움직임 정보는 인터 예측 방향 정보(L0 예측, L1 예측, Bi 예측 등)를 더 포함할 수 있다. 인터 예측의 경우에, 주변 블록은 현재 픽쳐 내에 존재하는 공간적 주변 블록(spatial neighboring block)과 참조 픽쳐에 존재하는 시간적 주변 블록(temporal neighboring block)을 포함할 수 있다. 상기 참조 블록을 포함하는 참조 픽쳐와 상기 시간적 주변 블록을 포함하는 참조 픽쳐는 동일할 수도 있고, 다를 수도 있다. 상기 시간적 주변 블록은 동일 위치 참조 블록(collocated reference block), 동일 위치 CU(colCU) 등의 이름으로 불릴 수 있으며, 상기 시간적 주변 블록을 포함하는 참조 픽쳐는 동일 위치 픽쳐(collocated picture, colPic)라고 불릴 수도 있다. 예를 들어, 인터 예측부(221)는 주변 블록들을 기반으로 움직임 정보 후보 리스트를 구성하고, 상기 현재 블록의 움직임 벡터 및/또는 참조 픽쳐 인덱스를 도출하기 위하여 어떤 후보가 사용되는지를 지시하는 정보를 생성할 수 있다. 다양한 예측 모드를 기반으로 인터 예측이 수행될 수 있으며, 예를 들어, 스킵 모드와 머지 모드의 경우에, 인터 예측부(221)는 주변 블록의 움직임 정보를 현재 블록의 움직임 정보로 이용할 수 있다. 스킵 모드의 경우, 머지 모드와 달리 레지듀얼 신호가 전송되지 않을 수 있다. 움직임 정보 예측(motion vector prediction, MVP) 모드의 경우, 주변 블록의 움직임 벡터를 움직임 벡터 예측자(motion vector predictor)로 이용하고, 움직임 벡터 차분(motion vector difference)을 시그날링함으로써 현재 블록의 움직임 벡터를 지시할 수 있다.
예측부(220)는 후술하는 다양한 예측 방법을 기반으로 예측 신호를 생성할 수 있다. 예를 들어, 예측부는 하나의 블록에 대한 예측을 위하여 인트라 예측 또는 인터 예측을 적용할 수 있을 뿐 아니라, 인트라 예측과 인터 예측을 동시에 적용할 수 있다. 이는 combined inter and intra prediction (CIIP) 모드라고 불릴 수 있다. 또한, 예측부는 블록에 대한 예측을 위하여 인트라 블록 카피(intra block copy, IBC) 예측 모드에 기반할 수도 있고 또는 팔레트 모드(palette mode)에 기반할 수도 있다. 상기 IBC 예측 모드 또는 팔레트 모드는 SCC(screen content coding) 등과 같이 게임 등의 컨텐츠 영상/동영상 코딩을 위하여 사용될 수 있다. IBC는 기본적으로 현재 픽쳐 내에서 예측을 수행하나, 현재 픽쳐 내에서 참조 블록을 도출하는 점에서 인터 예측과 유사하게 수행될 수 있다. 즉, IBC는 본 명세서에서 설명되는 인터 예측 기법들 중 적어도 하나를 이용할 수 있다. 팔레트 모드는 인트라 코딩 또는 인트라 예측의 일 예로 볼 수 있다. 팔레트 모드가 적용되는 경우 팔레트 테이블 및 팔레트 인덱스에 관한 정보를 기반으로 픽쳐 내 샘플 값을 시그날링할 수 있다. 상기 예측부(220)를 통해 생성된 예측 신호는 복원 신호를 생성하기 위해 이용되거나, 레지듀얼 신호를 생성하기 위해 이용될 수 있다.
변환부(232)는 레지듀얼 신호에 변환 기법을 적용하여 변환 계수들(transform coefficients)를 생성할 수 있다. 예를 들어, 변환 기법은 DCT(Discrete Cosine Transform), DST(Discrete Sine Transform), KLT(Karhunen-Loeve Transform), GBT(Graph-Based Transform), 또는 CNT(Conditionally Non-linear Transform) 중 적어도 하나를 포함할 수 있다. 여기서, GBT는 픽셀 간의 관계 정보를 그래프로 표현한다고 할 때, 이 그래프로부터 얻어진 변환을 의미한다. CNT는 이전에 복원된 모든 픽셀를 이용하여 예측 신호를 생성하고, 그에 기초하여 획득되는 변환을 의미한다. 또한, 변환 과정은 정사각형의 동일한 크기를 갖는 픽셀 블록에 적용될 수도 있고, 정사각형이 아닌 가변 크기의 블록에도 적용될 수 있다.
양자화부(233)는 변환 계수들을 양자화하여 엔트로피 인코딩부(240)로 전송되고, 엔트로피 인코딩부(240)는 양자화된 신호(양자화된 변환 계수들에 관한 정보)를 인코딩하여 비트스트림으로 출력할 수 있다. 상기 양자화된 변환 계수들에 관한 정보는 레지듀얼 정보라고 불릴 수 있다. 양자화부(233)는 계수 스캔 순서(scan order)를 기반으로 블록 형태의 양자화된 변환 계수들을 1차원 벡터 형태로 재배열할 수 있고, 상기 1차원 벡터 형태의 양자화된 변환 계수들을 기반으로 상기 양자화된 변환 계수들에 관한 정보를 생성할 수도 있다.
엔트로피 인코딩부(240)는 지수 골롬(exponential Golomb), CAVLC(context-adaptive variable length coding), CABAC(context-adaptive binary arithmetic coding) 등과 같은 다양한 인코딩 방법을 수행할 수 있다. 엔트로피 인코딩부(240)는 양자화된 변환 계수들 외 비디오/이미지 복원에 필요한 정보들(예컨대, 신택스 요소들(syntax elements)의 값 등)을 함께 또는 별도로 인코딩할 수도 있다.
인코딩된 정보(ex. 인코딩된 비디오/영상 정보)는 비트스트림 형태로 NAL(network abstraction layer) 유닛 단위로 전송 또는 저장될 수 있다. 상기 비디오/영상 정보는 어댑테이션 파라미터 세트(APS), 픽쳐 파라미터 세트(PPS), 시퀀스 파라미터 세트(SPS) 또는 비디오 파라미터 세트(VPS) 등 다양한 파라미터 세트에 관한 정보를 더 포함할 수 있다. 또한, 상기 비디오/영상 정보는 일반 제한 정보(general constraint information)을 더 포함할 수 있다. 본 명세서에서 인코딩 장치에서 디코딩 장치로 전달/시그날링되는 정보 및/또는 신택스 요소들은 비디오/영상 정보에 포함될 수 있다. 상기 비디오/영상 정보는 상술한 인코딩 절차를 통하여 인코딩되어 상기 비트스트림에 포함될 수 있다. 상기 비트스트림은 네트워크를 통하여 전송될 수 있고, 또는 디지털 저장매체에 저장될 수 있다. 여기서, 네트워크는 방송망 및/또는 통신망 등을 포함할 수 있고, 디지털 저장매체는 USB, SD, CD, DVD, 블루레이, HDD, SSD 등 다양한 저장매체를 포함할 수 있다. 엔트로피 인코딩부(240)로부터 출력된 신호는 전송하는 전송부(미도시) 및/또는 저장하는 저장부(미도시)가 인코딩 장치(200)의 내/외부 엘리먼트로서 구성될 수 있고, 또는 전송부는 엔트로피 인코딩부(240)에 포함될 수도 있다.
양자화부(233)로부터 출력된 양자화된 변환 계수들은 예측 신호를 생성하기 위해 이용될 수 있다. 예를 들어, 양자화된 변환 계수들에 역양자화부(234) 및 역변환부(235)를 통해 역양자화 및 역변환을 적용함으로써 레지듀얼 신호(레지듀얼 블록 or 레지듀얼 샘플들)를 복원할 수 있다. 가산부(250)는 복원된 레지듀얼 신호를 인터 예측부(221) 또는 인트라 예측부(222)로부터 출력된 예측 신호에 더함으로써 복원(reconstructed) 신호(복원 픽쳐, 복원 블록, 복원 샘플 어레이)가 생성될 수 있다. 스킵 모드가 적용된 경우와 같이 처리 대상 블록에 대한 레지듀얼이 없는 경우, 예측된 블록이 복원 블록으로 사용될 수 있다. 가산부(250)는 복원부 또는 복원 블록 생성부라고 불릴 수 있다. 생성된 복원 신호는 현재 픽쳐 내 다음 처리 대상 블록의 인트라 예측을 위하여 사용될 수 있고, 후술하는 바와 같이 필터링을 거쳐서 다음 픽쳐의 인터 예측을 위하여 사용될 수도 있다. 한편, 픽쳐 인코딩 및/또는 복원 과정에서 LMCS (luma mapping with chroma scaling)가 적용될 수도 있다.
필터링부(260)는 복원 신호에 필터링을 적용하여 주관적/객관적 화질을 향상시킬 수 있다. 예를 들어, 필터링부(260)은 복원 픽쳐에 다양한 필터링 방법을 적용하여 수정된(modified) 복원 픽쳐를 생성할 수 있고, 상기 수정된 복원 픽쳐를 메모리(270), 구체적으로 메모리(270)의 DPB에 저장할 수 있다. 상기 다양한 필터링 방법은, 디블록킹 필터링, 샘플 적응적 오프셋(sample adaptive offset), 적응적 루프 필터(adaptive loop filter), 양방향 필터(bilateral filter) 등을 포함할 수 있다. 필터링부(260)은 필터링에 관한 다양한 정보를 생성하여 엔트로피 인코딩부(240)로 전달할 수 있다. 필터링 관한 정보는 엔트로피 인코딩부(240)에서 인코딩되어 비트스트림 형태로 출력될 수 있다.
메모리(270)에 전송된 수정된 복원 픽쳐는 인터 예측부(221)에서 참조 픽쳐로 사용될 수 있다. 인코딩 장치는 이를 통하여 인터 예측이 적용되는 경우, 인코딩 장치(200)와 디코딩 장치에서의 예측 미스매치를 피할 수 있고, 부호화 효율도 향상시킬 수 있다.
메모리(270)의 DPB는 수정된 복원 픽쳐를 인터 예측부(221)에서의 참조 픽쳐로 사용하기 위해 저장할 수 있다. 메모리(270)는 현재 픽쳐 내 움직임 정보가 도출된(또는 인코딩된) 블록의 움직임 정보 및/또는 이미 복원된 픽쳐 내 블록들의 움직임 정보를 저장할 수 있다. 상기 저장된 움직임 정보는 공간적 주변 블록의 움직임 정보 또는 시간적 주변 블록의 움직임 정보로 활용하기 위하여 인터 예측부(221)에 전달할 수 있다. 메모리(270)는 현재 픽쳐 내 복원된 블록들의 복원 샘플들을 저장할 수 있고, 인트라 예측부(222)에 전달할 수 있다.
도 3은 본 개시의 실시예가 적용될 수 있고, 비디오/영상 신호의 디코딩이 수행되는 디코딩 장치의 개략적인 블록도를 나타낸다.
도 3을 참조하면, 디코딩 장치(300)는 엔트로피 디코딩부(entropy decoder, 310), 레지듀얼 처리부(residual processor, 320), 예측부(predictor, 330), 가산부(adder, 340), 필터링부(filter, 350) 및 메모리(memoery, 360)를 포함하여 구성될 수 있다. 예측부(330)는 인터 예측부(331) 및 인트라 예측부(332)를 포함할 수 있다. 레지듀얼 처리부(320)는 역양자화부(dequantizer, 321) 및 역변환부(inverse transformer, 321)를 포함할 수 있다.
상술한 엔트로피 디코딩부(310), 레지듀얼 처리부(320), 예측부(330), 가산부(340) 및 필터링부(350)는 실시예에 따라 하나의 하드웨어 컴포넌트(예를 들어 디코딩 장치 칩셋 또는 프로세서)에 의하여 구성될 수 있다. 또한 메모리(360)는 DPB(decoded picture buffer)를 포함할 수 있고, 디지털 저장 매체에 의하여 구성될 수도 있다. 상기 하드웨어 컴포넌트는 메모리(360)을 내/외부 컴포넌트로 더 포함할 수도 있다.
비디오/영상 정보를 포함하는 비트스트림이 입력되면, 디코딩 장치(300)는 도 2의 인코딩 장치에서 비디오/영상 정보가 처리된 프로세스에 대응하여 영상을 복원할 수 있다. 예를 들어, 디코딩 장치(300)는 상기 비트스트림으로부터 획득한 블록 분할 관련 정보를 기반으로 유닛들/블록들을 도출할 수 있다. 디코딩 장치(300)는 인코딩 장치에서 적용된 처리 유닛을 이용하여 디코딩을 수행할 수 있다. 따라서, 디코딩의 처리 유닛은 코딩 유닛일 수 있고, 코딩 유닛은 쿼드 트리 구조, 바이너리 트리 구조 및/또는 터너리 트리 구조에 따라서 코딩 트리 유닛 또는 최대 코딩 유닛으로부터 분할된 것일 수 있다. 코딩 유닛으로부터 하나 이상의 변환 유닛이 도출될 수 있다. 그리고, 디코딩 장치(300)를 통해 디코딩 및 출력된 복원 영상 신호는 재생 장치를 통해 재생될 수 있다.
디코딩 장치(300)는 도 2의 인코딩 장치로부터 출력된 신호를 비트스트림 형태로 수신할 수 있고, 수신된 신호는 엔트로피 디코딩부(310)를 통해 디코딩될 수 있다. 예를 들어, 엔트로피 디코딩부(310)는 상기 비트스트림을 파싱하여 영상 복원(또는 픽쳐 복원)에 필요한 정보(ex. 비디오/영상 정보)를 도출할 수 있다. 상기 비디오/영상 정보는 어댑테이션 파라미터 세트(APS), 픽쳐 파라미터 세트(PPS), 시퀀스 파라미터 세트(SPS) 또는 비디오 파라미터 세트(VPS) 등 다양한 파라미터 세트에 관한 정보를 더 포함할 수 있다. 또한 상기 비디오/영상 정보는 일반 제한 정보(general constraint information)을 더 포함할 수 있다. 디코딩 장치는 상기 파라미터 세트에 관한 정보 및/또는 상기 일반 제한 정보를 더 기반으로 픽쳐를 디코딩할 수 있다. 본 명세서에서 후술되는 시그날링/수신되는 정보 및/또는 신택스 요소들은 상기 디코딩 절차를 통하여 디코딩되어 상기 비트스트림으로부터 획득될 수 있다. 예컨대, 엔트로피 디코딩부(310)는 지수 골롬 부호화, CAVLC 또는 CABAC 등의 코딩 방법을 기초로 비트스트림 내 정보를 디코딩하고, 영상 복원에 필요한 신택스 요소의 값, 레지듀얼에 관한 변환 계수의 양자화된 값들을 출력할 수 있다. 보다 상세하게, CABAC 엔트로피 디코딩 방법은, 비트스트림에서 각 구문 요소에 해당하는 빈을 수신하고, 디코딩 대상 구문 요소 정보와 주변 및 디코딩 대상 블록의 디코딩 정보 혹은 이전 단계에서 디코딩된 심볼/빈의 정보를 이용하여 문맥(context) 모델을 결정하고, 결정된 문맥 모델에 따라 빈(bin)의 발생 확률을 예측하여 빈의 산술 디코딩(arithmetic decoding)를 수행하여 각 구문 요소의 값에 해당하는 심볼을 생성할 수 있다. 이때, CABAC 엔트로피 디코딩 방법은 문맥 모델 결정 후 다음 심볼/빈의 문맥 모델을 위해 디코딩된 심볼/빈의 정보를 이용하여 문맥 모델을 업데이트할 수 있다. 엔트로피 디코딩부(310)에서 디코딩된 정보 중 예측에 관한 정보는 예측부(인터 예측부(332) 및 인트라 예측부(331))로 제공되고, 엔트로피 디코딩부(310)에서 엔트로피 디코딩이 수행된 레지듀얼 값, 즉 양자화된 변환 계수들 및 관련 파라미터 정보는 레지듀얼 처리부(320)로 입력될 수 있다. 레지듀얼 처리부(320)는 레지듀얼 신호(레지듀얼 블록, 레지듀얼 샘플들, 레지듀얼 샘플 어레이)를 도출할 수 있다. 또한, 엔트로피 디코딩부(310)에서 디코딩된 정보 중 필터링에 관한 정보는 필터링부(350)으로 제공될 수 있다. 한편, 인코딩 장치로부터 출력된 신호를 수신하는 수신부(미도시)가 디코딩 장치(300)의 내/외부 엘리먼트로서 더 구성될 수 있고, 또는 수신부는 엔트로피 디코딩부(310)의 구성요소일 수도 있다.
한편, 본 명세서에 따른 디코딩 장치는 비디오/영상/픽쳐 디코딩 장치라고 불릴 수 있고, 상기 디코딩 장치는 정보 디코딩 장치(비디오/영상/픽쳐 정보 디코딩 장치) 및 샘플 디코딩 장치(비디오/영상/픽쳐 샘플 디코딩 장치)로 구분할 수도 있다. 상기 정보 디코딩 장치는 상기 엔트로피 디코딩부(310)를 포함할 수 있고, 상기 샘플 디코딩 장치는 상기 역양자화부(321), 역변환부(322), 가산부(340), 필터링부(350), 메모리(360), 인터 예측부(332) 및 인트라 예측부(331) 중 적어도 하나를 포함할 수 있다.
역양자화부(321)에서는 양자화된 변환 계수들을 역양자화하여 변환 계수들을 출력할 수 있다. 역양자화부(321)는 양자화된 변환 계수들을 2차원의 블록 형태로 재배열할 수 있다. 이 경우, 상기 재배열은 인코딩 장치에서 수행된 계수 스캔 순서를 기반하여 재배열을 수행할 수 있다. 역양자화부(321)는 양자화 파라미터(예를 들어 양자화 스텝 사이즈 정보)를 이용하여 양자화된 변환 계수들에 대한 역양자화를 수행하고, 변환 계수들(transform coefficient)를 획득할 수 있다.
역변환부(322)에서는 변환 계수들를 역변환하여 레지듀얼 신호(레지듀얼 블록, 레지듀얼 샘플 어레이)를 획득하게 된다.
예측부(320)는 현재 블록에 대한 예측을 수행하고, 상기 현재 블록에 대한 예측 샘플들을 포함하는 예측 블록(predicted block)을 생성할 수 있다. 예측부(320)는 엔트로피 디코딩부(310)로부터 출력된 상기 예측에 관한 정보를 기반으로 상기 현재 블록에 인트라 예측이 적용되는지 또는 인터 예측이 적용되는지 결정할 수 있고, 구체적인 인트라/인터 예측 모드를 결정할 수 있다.
예측부(320)는 후술하는 다양한 예측 방법을 기반으로 예측 신호를 생성할 수 있다. 예를 들어, 예측부(320)는 하나의 블록에 대한 예측을 위하여 인트라 예측 또는 인터 예측을 적용할 수 있을 뿐 아니라, 인트라 예측과 인터 예측을 동시에 적용할 수 있다. 이는 combined inter and intra prediction (CIIP) 모드라고 불릴 수 있다. 또한, 예측부는 블록에 대한 예측을 위하여 인트라 블록 카피(intra block copy, IBC) 예측 모드에 기반할 수도 있고 또는 팔레트 모드(palette mode)에 기반할 수도 있다. 상기 IBC 예측 모드 또는 팔레트 모드는 SCC(screen content coding) 등과 같이 게임 등의 컨텐츠 영상/동영상 코딩을 위하여 사용될 수 있다. IBC는 기본적으로 현재 픽쳐 내에서 예측을 수행하나 현재 픽쳐 내에서 참조 블록을 도출하는 점에서 인터 예측과 유사하게 수행될 수 있다. 즉, IBC는 본 명세서에서 설명되는 인터 예측 기법들 중 적어도 하나를 이용할 수 있다. 팔레트 모드는 인트라 코딩 또는 인트라 예측의 일 예로 볼 수 있다. 팔레트 모드가 적용되는 경우 팔레트 테이블 및 팔레트 인덱스에 관한 정보가 상기 비디오/영상 정보에 포함되어 시그날링될 수 있다.
인트라 예측부(331)는 현재 픽쳐 내의 샘플들을 참조하여 현재 블록을 예측할 수 있다. 상기 참조되는 샘플들은 예측 모드에 따라 상기 현재 블록의 주변(neighbor)에 위치할 수 있고, 또는 현재 블록으로부터 일정 거리만큼 떨어져서 위치할 수도 있다. 인트라 예측에서 예측 모드들은 하나 이상의 비방향성 모드와 복수의 방향성 모드를 포함할 수 있다. 인트라 예측부(331)는 주변 블록에 적용된 예측 모드를 이용하여, 현재 블록에 적용되는 예측 모드를 결정할 수도 있다.
인터 예측부(332)는 참조 픽쳐 상에서 움직임 벡터에 의해 특정되는 참조 블록(참조 샘플 어레이)을 기반으로, 현재 블록에 대한 예측 블록을 유도할 수 있다. 이때, 인터 예측 모드에서 전송되는 움직임 정보의 양을 줄이기 위해 주변 블록과 현재 블록 간의 움직임 정보의 상관성에 기초하여 움직임 정보를 블록, 서브블록 또는 샘플 단위로 예측할 수 있다. 상기 움직임 정보는 움직임 벡터 및 참조 픽쳐 인덱스를 포함할 수 있다. 상기 움직임 정보는 인터 예측 방향 정보(L0 예측, L1 예측, Bi 예측 등)를 더 포함할 수 있다. 인터 예측의 경우에, 주변 블록은 현재 픽쳐 내에 존재하는 공간적 주변 블록(spatial neighboring block)과 참조 픽쳐에 존재하는 시간적 주변 블록(temporal neighboring block)을 포함할 수 있다. 예를 들어, 인터 예측부(332)는 주변 블록들을 기반으로 움직임 정보 후보 리스트를 구성하고, 수신한 후보 선택 정보를 기반으로 상기 현재 블록의 움직임 벡터 및/또는 참조 픽쳐 인덱스를 도출할 수 있다. 다양한 예측 모드를 기반으로 인터 예측이 수행될 수 있으며, 상기 예측에 관한 정보는 상기 현재 블록에 대한 인터 예측 모드를 지시하는 정보를 포함할 수 있다.
가산부(340)는 획득된 레지듀얼 신호를 예측부(인터 예측부(332) 및/또는 인트라 예측부(331) 포함)로부터 출력된 예측 신호(예측 블록, 예측 샘플 어레이)에 더함으로써 복원 신호(복원 픽쳐, 복원 블록, 복원 샘플 어레이)를 생성할 수 있다. 스킵 모드가 적용된 경우와 같이 처리 대상 블록에 대한 레지듀얼이 없는 경우, 예측 블록이 복원 블록으로 사용될 수 있다.
가산부(340)는 복원부 또는 복원 블록 생성부라고 불릴 수 있다. 생성된 복원 신호는 현재 픽쳐 내 다음 처리 대상 블록의 인트라 예측을 위하여 사용될 수 있고, 후술하는 바와 같이 필터링을 거쳐서 출력될 수도 있고 또는 다음 픽쳐의 인터 예측을 위하여 사용될 수도 있다. 한편, 픽쳐 디코딩 과정에서 LMCS (luma mapping with chroma scaling)가 적용될 수도 있다.
필터링부(350)는 복원 신호에 필터링을 적용하여 주관적/객관적 화질을 향상시킬 수 있다. 예를 들어 필터링부(350)는 복원 픽쳐에 다양한 필터링 방법을 적용하여 수정된(modified) 복원 픽쳐를 생성할 수 있고, 상기 수정된 복원 픽쳐를 메모리(360), 구체적으로 메모리(360)의 DPB에 전송할 수 있다. 상기 다양한 필터링 방법은, 디블록킹 필터링, 샘플 적응적 오프셋(sample adaptive offset), 적응적 루프 필터(adaptive loop filter), 양방향 필터(bilateral filter) 등을 포함할 수 있다.
메모리(360)의 DPB에 저장된 (수정된) 복원 픽쳐는 인터 예측부(332)에서 참조 픽쳐로 사용될 수 있다. 메모리(360)는 현재 픽쳐 내 움직임 정보가 도출된(또는 디코딩된) 블록의 움직임 정보 및/또는 이미 복원된 픽쳐 내 블록들의 움직임 정보를 저장할 수 있다. 상기 저장된 움직임 정보는 공간적 주변 블록의 움직임 정보 또는 시간적 주변 블록의 움직임 정보로 활용하기 위하여 인터 예측부(260)에 전달할 수 있다. 메모리(360)는 현재 픽쳐 내 복원된 블록들의 복원 샘플들을 저장할 수 있고, 인트라 예측부(331)에 전달할 수 있다.
본 명세서에서, 인코딩 장치(200)의 필터링부(260), 인터 예측부(221) 및 인트라 예측부(222)에서 설명된 실시예들은 각각 디코딩 장치(300)의 필터링부(350), 인터 예측부(332) 및 인트라 예측부(331)에도 동일 또는 대응되도록 적용될 수 있다.
도 4는 본 개시에 따른 일실시예로서, 디코딩 장치에 의해 수행되는 어파인 모델 기반의 인터 예측 방법을 도시한 것이다.
도 4를 참조하면, 현재 블록의 움직임 정보를 예측하기 위한 후보 리스트를 생성할 수 있다(S400).
상기 후보 리스트는, 어파인 모델을 이용한 인터 예측에 있어서 어파인 머지 모드(affine merge mode)를 위한 것일 수 있다. 다만, 이에 한정되지 아니하며, 상기 후보 리스트는 어파인 인터 모드(affine inter mode)를 위한 것일 수도 있다. 이하, 어파인 모드는 어파인 머지 모드로 해석될 수도 있고, 어파인 인터 모드로 해석될 수도 있다.
상기 움직임 정보는, 움직임 벡터, 참조 픽쳐 인덱스, 인터 예측 방향 정보, 양방향 가중 예측을 위한 가중치 정보, 또는 지역 조도 보상의 적용 여부를 나타내는 플래그 중 적어도 하나를 포함할 수 있다.
상기 후보 리스트는, 하나 또는 그 이상의 어파인 모델 기반의 후보(이하, 어파인 후보라 함)를 포함할 수 있다. 어파인 후보(Affine Candidate)는, 제어점 벡터(control point vector)를 가진 후보를 의미할 수 있다. 제어점 벡터는, 어파인 모델을 위한 제어점의 움직임 벡터를 의미하고, 블록의 코너 위치(예를 들어, 좌상단, 우상단, 좌하단, 또는 우하단 코너 중 적어도 하나의 위치)에 대해서 각각 정의될 수 있다.
어파인 후보는, 공간적 후보, 시간적 후보 또는 구성된 후보(constructed candidate) 중 적어도 하나를 포함할 수 있다. 여기서, 공간적 후보는, 현재 블록에 공간적으로 인접한 주변 블록(이하, 공간적 주변 블록이라 함)의 벡터로부터 유도될 수 있고, 시간적 후보는, 현재 블록에 시간적으로 인접한 주변 블록(이하, 시간적 주변 블록이라 함)의 벡터로부터 유도될 수 있다. 여기서, 상기 주변 블록은 어파인 모드로 부호화된 블록을 의미할 수 있다. 상기 벡터는, 제어점 벡터를 의미할 수도 있고, 움직임 벡터를 의미할 수도 있다.
공간적/시간적 주변 블록의 벡터를 기반으로 공간적/시간적 후보를 유도하는 방법에 대해서는 도 5를 참조하여 자세히 살펴 보도록 한다.
한편, 상기 구성된 후보는, 현재 블록에 대한 공간적/시간적 주변 블록의 움직임 벡터 간의 조합에 기초하여 유도될 수 있으며, 여기서, 주변 블록은 어파인 모드로 부호화된 블록이 아닐 수 있다. 또는, 적어도 하나의 주변 블록은 어파인 모드로 부호화된 블록일 수도 있다. 이는 도 6을 참조하여 자세히 살펴 보기로 한다.
상기 어파인 후보는 히스토리 기반의 후보(history-based candidate)를 더 포함할 수 있다. 현재 블록 이전에 복호화된 블록(이하, 이전 블록이라 함)의 움직임 정보는 복호화 순서에 따라 소정의 크기를 가진 버퍼에 순차적으로 저장될 수 있다. 상기 이전 블록의 움직임 정보는 현재 블록의 움직임 정보를 유도하기 위한 후보로 이용될 수 있으며, 이와 같이 이전 블록의 움직임 정보를 가진 후보를 히스토리 기반의 후보라 부르기로 한다. 상기 이전 블록은, 현재 블록에 인접한 주변 블록일 수도 있고, 현재 블록에 인접하지 않은 주변 블록일 수 있다.
상기 어파인 후보는, 평균 후보(average candidate)를 더 포함할 수 있다.
상기 후보 리스트에 포함 가능한 어파인 후보의 최대 개수(MaxNumSubblockMergeCand) 또는 상기 후보 리스트에 기-포함된 어파인 후보의 개수 중 적어도 하나에 기초하여 생성 가능한 평균 후보의 개수가 결정될 수 있다. 여기서, 상기 최대 개수는 인코딩 장치와 디코딩 장치에 동일하게 정의된 고정된 개수일 수 있다. 또는, 상기 최대 개수는 후보 리스트의 크기 또는 후보 리스트에 포함 가능한 어파인 후보의 최대 개수를 나타내는 정보에 기초하여 가변적으로 결정될 수도 있다. 상기 최대 개수를 나타내는 정보는, 시퀀스 파라미터 세트, 픽쳐 파라미터 세트, 픽쳐 헤더, 슬라이스 헤더, 또는 코딩 블록 중 적어도 하나의 레벨에서 시그날링될 수 있다.
일예로, 후보 리스트에 기-포함된 후보의 개수가 N개이고, 이들 모두 어파인 모델을 이용하여 생성된 후보인 경우, 후보 리스트에 포함 가능한 평균 후보의 개수는 Min(N*(N-1)/2, MaxNumSubblockMergeCand-N)로 유도될 수 있다. 또는, 후보 리스트에 포함 가능한 평균 후보의 개수는 Min(N*(N-1)/2, MaxNumSubblockMergeCand-N-P)로 유도될 수 있다. 여기서, P는 0, 1, 2, 3 또는 그 이상의 정수일 수 있다. P는 인코딩 장치와 디코딩 장치에 동일하게 정의된 고정된 값일 수도 있고, 후보 리스트에 기-포함된 후보 중에서 어파인 모델을 이용하여 생성되지 않은 후보의 개수에 기초하여 가변적으로 결정될 수도 있다.
후보 리스트 내의 2개의 후보 즉, Cand0와 Cand1을 기반으로 평균 후보가 생성될 수 있다. Cand0는 후보 리스트에 기-포함된 후보 중 어느 하나이고, Cand1은 후보 리스트에 기-포함된 후보 중 다른 어느 하나일 수 있다. 상기 Cand0와 Cand1은 소정의 우선순서를 고려하여 결정될 수 있다. 일예로, 상기 우선순서는 다음 표 1과 같으며, 이는 일예로서 10개의 평균 후보를 생성하기 위한 Cand0와 Cand1의 조합을 나타낸 것이다.
Idx 0 1 2 3 4 5 6 7 8 9
Cand0Idx 0 0 1 0 1 2 0 1 2 3
Cand1Idx 1 2 2 3 3 3 4 4 4 4
표 1을 참조하면, Idx는 우선순서를 나타내며, Idx가 작을수록 높은 우선순서를 가질 수 있다. Cand0Idx와 Cand1Idx는 후보 리스트 내에서 평균 후보를 생성하기 위해 이용되는 후보의 위치를 각각 의미할 수 있다. 예를 들어, 0에서 9의 순서으로, 해당 Idx에 대응하는 Cand0와 Cand1를 기반으로 평균 후보가 생성될 수 있다. Idx가 0인 경우, 후보 리스트에서 0의 인덱스를 가진 후보(Cand0)와 후보 리스트에서 1의 인덱스를 가진 후보(Cand1)를 기반으로 평균 후보가 생성될 수 있다.
전술한 방법을 통해, 평균 후보를 생성하기 위한 2개의 대상 후보를 결정할 수 있다. 다만, 이는 일예에 불과하며, 생성되는 평균 후보의 개수 및/또는 우선순서는 변경될 수 있다.
또한, 2개의 대상 후보는 상속된 후보(inherited candidate), 구성된 후보 및 히스토리 기반의 후보 내에서 선택되도록 제한될 수 있다. 여기서, 상속된 후보는, 전술한 공간적 후보 또는 시간적 후보 중 적어도 하나를 포함할 수 있다. 이 경우, 서로 다른 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 일예로, 2개의 대상 후보 중 어느 하나의 타입은 상속된 후보이고, 다른 하나의 타입은 구성된 후보일 수 있다. 또는, 2개의 대상 후보 중 어느 하나의 타입은 구성된 후보이고, 다른 하나의 타입은 히스토리 기반의 후보일 수 있다. 또는, 동일한 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성되도록 제한될 수도 있다. 즉, 평균 후보는 상속된 후보인 2개의 대상 후보를 기반으로 생성되거나, 구성된 후보인 2개의 대상 후보를 기반으로 생성될 수 있다. 마찬가지로, 평균 후보는 히스토리 기반의 후보인 2개의 대상 후보를 기반으로 생성되도록 제한될 수 있다.
또는, 2개의 대상 후보는 상속된 후보와 구성된 후보 내에서 선택되고, 히스토리 기반의 후보 내에서 선택되지 않도록 제한될 수 있다. 이 경우, 전술한 바와 같이, 서로 다른 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성될 수도 있고, 동일한 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성되도록 제한될 수도 있다.
또는, 2개의 대상 후보는 상속된 후보와 히스토리 기반의 후보 내에서 선택되고, 구성된 후보 내에서 선택되지 않도록 제한될 수 있다. 이 경우, 전술한 바와 같이, 서로 다른 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성될 수도 있고, 동일한 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성되도록 제한될 수도 있다.
또는, 2개의 대상 후보는 구성된 후보와 히스토리 기반의 후보 내에서 선택되고, 상속된 후보 내에서 선택되지 않도록 제한될 수 있다. 이 경우, 전술한 바와 같이, 서로 다른 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성될 수도 있고, 동일한 타입의 2개의 대상 후보를 기반으로 평균 후보가 생성되도록 제한될 수도 있다.
또는, 2개의 대상 후보는 상속된 후보, 구성된 후보 또는 히스토리 기반의 후보 중 어느 하나의 타입 내에서만 선택되도록 제한될 수도 있다. 일예로, 2개의 대상 후보는 상속된 후보 내에서만 선택되도록 제한될 수 있다. 2개의 대상 후보는 구성된 후보 내에서만 선택되도록 제한될 수도 있다. 2개의 대상 후보는 히스토리 기반의 후보 내에서만 선택되도록 제한될 수도 있다.
한편, 상기 2개의 대상 블록 중 어느 하나의 어파인 모델의 타입이 4-parameter이고, 다른 하나의 어파인 모델의 타입이 6-parameter인 경우가 존재할 수 있다. 이러한 경우, 평균 후보를 생성하는 방법에 대해서 살펴보기로 한다.
만일 2개의 대상 후보가 모두 4-parameter인 경우, 2개의 대상 후보의 제1 제어점 벡터들(cpMV[0]) 및 제2 제어점 벡터들(cpMV[1])을 각각 평균하며, 4-parameter로 어파인 모델의 타입을 저장/전파시킬 수 있다. 마찬가지로, 2개의 대상 후보가 모두 6-parameter인 경우, 제1 제어점 벡터들(cpMV[0]), 제2 제어점 벡터들(cpMV[1]) 및 제3 제어점 벡터들(cpMV[2])을 각각 평균하며, 6-parameter로 어파인 모델의 타입을 저장/전파시킬 수 있다.
구체적으로, 2개의 대상 후보에 대한 어파인 모델의 타입이 4-parameter인 경우, 2개의 대상 후보의 제1 제어점 벡터들(cpMV[0]) 및 제2 제어점 벡터들(cpMV[1])을 각각 평균하여 상기 평균 후보의 제1 및 제2 제어점 벡터들을 각각 산출할 수 있다. 이때, 평균 후보에 대한 어파인 모델의 타입은 4-parameter로 설정될 수 있다. 상기 평균 후보가 현재 블록의 제어점 벡터를 유도하기 위해 이용된 경우, 평균 후보에 대한 어파인 모델의 타입은 현재 블록에 저장될 수 있고, 현재 블록 이후의 블록으로 전파될 수도 있다 마찬가지로, 2개의 대상 후보에 대한 어파인 모델의 타입이 모두 6-parameter인 경우, 2개의 대상 후보의 제1 제어점 벡터들(cpMV[0]), 제2 제어점 벡터들(cpMV[1]) 및 제3 제어점 벡터들(cpMV[2])을 각각 평균하여 상기 평균 후보의 제1 내지 제3 제어점 벡터들을 각각 산출할 수 있다. 이때, 평균 후보의 어파인 모델의 타입은 6-parameter로 설정될 수 있다. 상기 평균 후보가 현재 블록의 제어점 벡터를 유도하기 위해 이용된 경우, 평균 후보에 대한 어파인 모델의 타입은 현재 블록에 저장될 수 있고, 현재 블록 이후의 블록으로 전파될 수도 있다.
또는, 2개의 대상 후보에 대한 어파인 모델의 타입이 4-parameter인 경우라도, 2개의 대상 후보 중 어느 하나(Cand0)는 제1 및 제2 제어점 벡터들을 가지고, 다른 하나(Cand1)는 제1 및 제3 제어점 벡터들을 가질 수 있다. 이 경우, 평균 후보의 제1 제어점 벡터는, Cand0의 제1 제어점 벡터와 Cand1의 제1 제어점 벡터의 평균값으로 산출될 수 있다. 평균 후보의 제2 제어점 벡터는 Cand0의 제2 제어점 벡터로 설정되거나, Cand0의 제2 제어점 벡터와 Cand1의 제3 제어점 벡터의 평균값으로 산출될 수 있다. 또는, Cand1의 제1 제어점 벡터와 제3 제어점 벡터를 기반으로 제2 제어점 벡터가 유도되고, Cand0의 제2 제어점 벡터와 Cand1의 기-유도된 제2 제어점 벡터의 평균값으로 평균 후보의 제2 제어점 벡터가 산출될 수도 있다.
또는, 전술한 바와 같이, 2개의 대상 후보에 대한 어파인 모델의 타입이 4-parameter인 경우라도, 2개의 대상 후보 중 어느 하나(Cand0)는 제1 및 제2 제어점 벡터들을 가지고, 다른 하나(Cand1)는 제1 및 제3 제어점 벡터들을 가지는 경우, 이러한 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않도록 제한될 수도 있다.
2개의 대상 후보에 대한 어파인 모델의 타입이 서로 다른 경우, 4-parameter 또는 6-parameter 중 어느 하나로 어파인 모델의 타입을 저장/전파할 수 있다.
일예로, 4-parameter인 대상 블록이 제1 제어점 벡터와 제2 제어점 벡터를 가지는 경우, 제1 제어점 벡터와 제2 제어점 벡터를 이용하여 제3 제어점 벡터를 유도할 수 있다. 또는, 4-parameter인 대상 블록이 제1 제어점 벡터와 제3 제어점 벡터를 가지는 경우, 제1 제어점 벡터와 제3 제어점 벡터를 이용하여 제2 제어점 벡터를 유도할 수 있다. 그런 다음, 6-parameter인 다른 대상 후보의 제1 내지 제3 제어점 벡터들과 각각 평균하여 평균 후보의 제1 내지 제3 제어점 벡터를 산출할 수 있다. 이때, 평균 후보의 어파인 모델의 타입은 6-parameter로 설정될 수 있다. 상기 평균 후보가 현재 블록의 제어점 벡터를 유도하기 위해 이용된 경우, 평균 후보에 대한 어파인 모델의 타입은 현재 블록에 저장될 수 있고, 현재 블록 이후의 블록으로 전파될 수도 있다.
또는, 2개의 대상 후보에 대한 어파인 모델의 타입이 서로 다른 경우, parameter의 개수가 적은 타입을 저장/전파할 수 있다. 즉, Cand0에 대한 어파인 모델의 타입이 4-parameter이고, Cand1에 대한 어파인 모델의 타입이 6-parameter인 경우, 평균 후보의 어파인 모델의 타입은 Cand0의 어파인 모델의 타입과 동일하게 4-parameter로 설정될 수 있다. 상기 평균 후보가 현재 블록의 제어점 벡터를 유도하기 위해 이용된 경우, 평균 후보에 대한 어파인 모델의 타입은 현재 블록에 저장될 수 있고, 현재 블록 이후의 블록으로 전파될 수도 있다.
또는, 2개의 대상 후보에 대한 어파인 모델의 타입이 서로 다른 경우, 2개의 대상 후보 간 제어점 벡터들의 차이가 소정의 문턱값보다 작은 경우, 4-parameter로 저장/전파하고, 그렇지 않은 경우, 6-parameter로 저장/전파할 수 있다. 여기서, 제어점 벡터들의 차이는, Cand0의 제1 제어점 벡터와 Cand1의 제1 제어점 벡터 간의 제1 차이를 의미할 수 있다. 상기 제어점 벡터들의 차이는, Cand0의 제2 제어점 벡터와 Cand1의 제2 제어점 벡터 간의 제2 차이를 의미할 수 있다. 상기 제어점 벡터들의 차이는, Cand0의 제3 제어점 벡터와 Cand1의 제3 제어점 벡터 간의 제3 차이를 의미할 수 있다. 상기 제어점 벡터들의 차이는, Cand0에 대한 차분 제어점 벡터와 Cand1에 대한 차분 제어점 벡터 간의 제4 차이를 의미할 수 있다. 여기서, 차분 제어점 벡터는, (cpMV[0]-cpMV[1]), (cpMV[0]-cpMV[2]), (cpMV[1]-cpMV[2]), (cpMV[1]-cpMV[0]), (cpMV[2]-cpMV[0]), 또는 (cpMV[2]-cpMV[1]) 중 적어도 하나로 정의될 수 있다.
Cand0와 Cand1가 제2 제어점 벡터를 가지지 않을 수 있다. 이러한 경우, 전술한 제어점 벡터들의 차이를 산출하기 위해, 해당 대상 후보의 제1 제어점 벡터와 제3 제어점 벡터를 기반으로 제2 제어점 벡터가 유도될 수 있다. 마찬가지로, Cand0와 Cand1가 제3 제어점 벡터를 가지지 않을 수 있으며, 이러한 경우, 해당 대상 후보의 제1 제어점 벡터와 제2 제어점 벡터를 기반으로 제3 제어점 벡터가 유도될 수 있다. 상기 제어점 벡터들의 차이는, 상기 제1 차이, 제2 차이, 제3 차이, 또는 제4 차이 중 어느 하나를 의미할 수도 있고, 제1 내지 제4 차이 중 적어도 둘을 포함할 수도 있다. 상기 문턱값은 인코딩 장치와 디코딩 장치에 동일하게 기-정의된 값일 수 있다.
또는, 어파인 모델의 타입이 서로 동일한 2개의 대상 후보만을 이용하도록 제한될 수 있다. 예를 들어, 전술한 우선순서에 따라 순차적으로 2개의 대상 후보를 선택하되, 선택된 2개의 대상 후보가 서로 상이한 어파인 모델의 타입을 가지는 경우, 해당 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 반면, 선택된 2개의 대상 후보가 동일한 어파인 모델의 타입을 가지는 경우, 해당 대상 후보를 기반으로 평균 후보가 생성될 수 있다.
또한, 어파인 모델의 타입을 고려하여 후보 리스트 내의 어파인 후보들을 재배열한 뒤, 평균 후보를 생성할 수도 있다. 후보 리스트에 서로 다른 어파인 모델의 타입이 혼재되어 있을 경우, 어파인 모델의 타입을 고려하지 않고 전술한 우선순서에 따라 순차적으로 평균 후보를 생성하면 해당 후보의 정확도가 낮아질 수 있다. 따라서, 평균 후보를 생성하기 전에 후보 리스트의 어파인 후보들을 재배열하고, 재배열된 후보 리스트를 기반으로 평균 후보를 생성할 수 있다.
일예로, 6-parameter인 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수 있다. 또는, 4-parameter인 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수도 있다. 또는, 후보 리스트의 첫번째 어파인 후보의 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수 있다. 여기서, 첫번째 어파인 후보는 후보 리스트 내에서 가장 작은 인덱스가 할당된 후보를 의미할 수 있다. 또는, 후보 리스트 내에 존재하는 어파인 후보들 중에서 가장 빈도수가 높은 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수도 있다.
상기 재배열된 후보 리스트에서 어파인 모델의 타입이 동일한 2개의 대상 후보만을 이용하여 평균 후보가 생성될 수 있다. 또는, 2개의 대상 후보가 동일한 어파인 모델의 타입을 가지는지 여부와 관계없이, 재배열된 후보 리스트에서 소정의 우선순서에 따라 선택된 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다.
후보 리스트를 구성하는 과정에 있어서, 툴(tool) 별 다양한 상속 규칙을 살펴보기로 한다.
현재 블록에 대해서 지역 조도 보상(local illumination compensation, LIC)이 적용될 때, 평균 후보에 대한 지역 조도 보상의 적용 여부를 나타내는 플래그(LIC_flag)는 다음과 같이 유도될 수 있다.
평균 후보를 생성하기 위한 2개의 대상 후보 중 하나라도 LIC_flag가 TRUE이면 평균 후보에 대한 LIC_flag는 TRUE로 저장/전파될 수 있다. 만약, 지역 조도 보상이 단방향 예측의 블록에만 적용되도록 제한된 경우, 전술한 상속 규칙에 따라 TRUE로 저장되었더라도 이는 FALSE로 변경되어 저장/전파될 수 있다. 2개의 대상 후보를 평균하는 과정에서, 후술하는 바와 같이 2개의 대상 후보가 특정 조건을 만족하지 않는 경우(일예로, 대상 후보가 이용 가능하지 않은 경우, 2개의 대상 후보 간의 참조 픽쳐 인덱스 또는 참조 픽쳐가 서로 상이한 경우), 2개의 대상 후보 중 어느 하나의 제어점 벡터가 저장될 수 있다. 이러한 경우, 평균 후보의 LIC_flag는 상기 2개의 대상 후보 중 어느 하나의 움직임 정보에 포함된 LIC flag를 따를 수 있다.
현재 블록에 대해서 양방향 가중 예측(Bi-prediction with CU-level Weights, BCW)이 적용될 때, 평균 후보에 대한 양방향 가중 예측을 위한 가중치 정보(BCW_index)는 다음과 같이 유도될 수 있다.
2개의 대상 후보가 양방향 예측을 위한 움직임 정보를 가지고, 각 대상 후보의 BCW_index가 동일한 경우, 평균 후보에 대한 BCW_index는 대상 후보의 BCW_index로 저장/전파될 수 있다. 그렇지 않으면, 평균 후보에 대한 BCW_index는 디폴트 값(예를 들어, 0)으로 저장/전파될 수 있다. BCW_index가 디폴트 값인 경우, L0 예측 블록과 L1 예측 블록의 평균을 통해 최종 예측 블록이 생성될 수 있다.
또는, 평균 후보는 2개의 대상 후보의 평균을 통해 새롭게 생성된 후보이므로, 대상 후보로부터의 상속없이, 평균 후보의 LIC_flag와 BCW_index는 FALSE와 디폴트 값으로 각각 설정될 수 있다.
평균 후보는, 기-정의된 특정 조건을 고려하여 후보 리스트에 적응적으로 추가/저장될 수 있다. 여기서, 특정 조건은, 대상 후보의 이용 가능 여부(조건 1), 2개의 대상 후보가 동일한 참조 픽쳐 인덱스를 가지는지 여부(조건 2), 또는 2개의 대상 후보가 동일한 참조 픽쳐를 가지는지 여부(조건 3) 중 적어도 하나를 포함할 수 있다.
이하, 설명의 편의를 위해, 평균 후보를 생성하기 위한 2개의 대상 후보를 제1 대상 후보(Cand0[list][idx])와 제2 대상 후보(Cand1[list][idx])라 한다. 여기서, 변수 list는 0 또는 1의 값을 가지며, L0 예측 방향 또는 L1 예측 방향 중 어느 하나를 지시할 수 있다. 또한, 2개의 대상 후보에 대한 어파인 모델의 타입이 6-parameter인 경우, 변수 idx는 0 내지 2의 값을 가지며, 이는 제1 내지 제3 제어점 벡터들을 각각 지시할 수 있다. 2개의 대상 후보에 대한 어파인 모델의 타입이 4-parameter인 경우, 변수 idx는 0 및 1의 값을 가지며, 이는 제1 및 제2 제어점 벡터들을 각각 지시할 수 있다.
또한, Avail0[list]는 각 예측 방향 별 제1 대상 후보의 이용 가능성을 나타내고, Avail1[list]는 각 예측 방향 별 제2 대상 후보의 이용 가능성을 나타낼 수 있다. 일예로, AvailX[0]이 1인 경우(여기서, X는 0 또는 1임), 이는 대상 후보가 L0 예측 방향에 대해서 이용 가능함을 나타낼 수 있다. 다시 말해, 이는 대상 후보가 L0 예측 방향의 움직임 정보를 가지고 있음을 의미할 수 있다. 반대로, AvailX[0]이 0인 경우, 이는 대상 후보가 L0 예측 방향에 대해서 이용 가능하지 않음을 나타낼 수 있다. 다시 말해, 이는 대상 후보가 L0 예측 방향의 움직임 정보를 가지고 있지 않음을 의미할 수 있다. 마찬가지로, AvailX[1]이 1인 경우, 이는 대상 후보가 L1 예측 방향에 대해서 이용 가능함을 나타낼 수 있다. 다시 말해, 이는 대상 후보가 L1 예측 방향의 움직임 정보를 가지고 있음을 의미할 수 있다. 반대로, AvailX[1]이 0인 경우, 이는 대상 후보가 L1 예측 방향에 대해서 이용 가능하지 않음을 나타낼 수 있다. 다시 말해, 이는 대상 후보가 L1 예측 방향의 움직임 정보를 가지고 있지 않음을 의미할 수 있다.
또한, ref_idx0[list]는 각 예측 방향 별 제1 대상 후보의 참조 픽쳐 인덱스를 나타내고, ref_idx1[list]는 각 예측 방향 별 제2 대상 후보의 참조 픽쳐 인덱스를 나타낼 수 있다.
(조건 1) 대상 후보의 이용 가능성
2개의 대상 후보의 이용 가능성을 고려하여 2개의 대상 후보의 평균 여부를 결정할 수 있다. 상기 평균 여부는 L0 예측 방향 또는 L1 예측 방향 중 적어도 하나에 대해서 결정될 수 있다.
구체적으로, 2개의 대상 후보가 L0 예측 방향에 대해서 이용 가능한 경우(즉, Avail0[0]=Avail1[0]=1), 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L0 예측 방향의 제어점 벡터 간의 평균값으로 산출될 수 있다.
2개의 대상 후보 중 어느 하나의 대상 후보만이 L0 예측 방향에 대해서 이용 가능한 경우(즉, Avail0[0]=1 & Avail1[0]=0인 경우 또는 Avail0[0]=0 & Avail1[0]=1인 경우), 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 즉, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L0 예측 방향의 제어점 벡터 간의 평균값으로 산출되지 않을 수 있다. 대신에, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, L0 예측 방향에 대해서 이용 가능한 대상 후보의 L0 예측 방향의 제어점 벡터로 설정/저장될 수 있다.
전술한 방법은 L1 예측 방향에 대해서도 동일하게 적용될 수 있다. 즉, L1 예측 방향에 대해서 2개의 대상 후보가 이용 가능한지 여부에 기초하여, 2개의 대상 후보의 평균 여부를 결정할 수 있다.
2개의 대상 후보가 L1 예측 방향에 대해서 이용 가능한 경우(즉, Avail0[1]=Avail1[1]=1), 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 반면, 2개의 대상 후보 중 어느 하나의 대상 후보만이 L1 예측 방향에 대해서 이용 가능한 경우(즉, Avail0[1]=1 & Avail1[1]=0인 경우 또는 Avail0[1]=0 & Avail1[1]=1인 경우), 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 대신에, 평균 후보에 대한 L1 예측 방향의 제어점 벡터는 L1 예측 방향에 대해서 이용 가능한 대상 후보의 L1 예측 방향의 제어점 벡터로 설정/저장될 수 있다.
(조건 2) 2개의 대상 후보가 동일한 참조 픽쳐 인덱스를 가지는지 여부
각 예측 방향에 대해서, 2개의 대상 후보가 동일한 참조 픽쳐 인덱스를 가지는지 여부에 기초하여, 2개의 대상 후보의 평균 여부를 결정할 수 있다. 상기 평균 여부는 L0 예측 방향 또는 L1 예측 방향 중 적어도 하나에 대해서 결정될 수 있다.
구체적으로, 2개의 대상 후보에 대한 L0 예측 방향의 참조 픽쳐 인덱스가 서로 동일한 경우(즉, ref_idx0[0]=ref_idx1[0]), 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L0 예측 방향의 제어점 벡터 간의 평균값으로 산출될 수 있다.
2개의 대상 후보에 대한 L0 예측 방향의 참조 픽쳐 인덱스가 서로 상이한 경우(즉, ref_idx0[0]≠ref_idx1[0]), 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 즉, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L0 예측 방향의 제어점 벡터 간의 평균값으로 산출되지 않을 수 있다. 대신에, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 2개의 대상 후보 중 어느 하나에 대한 L0 예측 방향의 제어점 벡터로 설정/저장될 수 있다. 여기서, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 후보 리스트의 인덱스가 더 작은 후보(즉, Cand0)를 의미할 수 있다. 또는, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 참조 픽쳐 인덱스가 더 작은 후보를 의미할 수도 있다.
전술한 방법은 L1 예측 방향에 대해서도 동일하게 적용될 수 있다. 즉, 제1 대상 후보에 대한 L1 예측 방향의 참조 픽쳐 인덱스가 제2 대상 후보에 대한 L1 예측 방향의 참조 픽쳐 인덱스와 동일한지 여부에 기초하여, 2개의 대상 후보의 평균 여부를 결정할 수 있다.
2개의 대상 후보에 대한 L1 예측 방향의 참조 픽쳐 인덱스가 서로 동일한 경우(즉, ref_idx0[1]=ref_idx1[1]), 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 반면, 2개의 대상 후보에 대한 L1 예측 방향의 참조 픽쳐 인덱스가 서로 상이한 경우(즉, ref_idx0[1]≠ref_idx1[1]), 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 대신에, 평균 후보에 대한 L1 예측 방향의 제어점 벡터는, 2개의 대상 후보 중 어느 하나에 대한 L1 예측 방향의 제어점 벡터로 설정/저장될 수 있다.
(조건 3) 2개의 대상 후보가 동일한 참조 픽쳐를 가지는지 여부
각 예측 방향에 대해서, 2개의 대상 후보가 동일한 참조 픽쳐를 가지는지 여부에 기초하여, 2개의 대상 후보의 평균 여부를 결정할 수 있다. 2개의 참조 픽쳐가 동일한지 여부는 2개의 참조 픽쳐에 각각 할당된 출력 순서 정보(picture order count, POC)를 통해 확인할 수 있다. 즉, 2개의 참조 픽쳐 간 출력 순서 정보의 차이가 0이면, 2개의 참조 픽쳐는 동일한 참조 픽쳐이고, 그렇지 않은 경우, 2개의 참조 픽쳐는 동일한 참조 픽쳐가 아닐 수 있다. 상기 평균 여부는 L0 예측 방향 또는 L1 예측 방향 중 적어도 하나에 대해서 결정될 수 있다.
구체적으로, 제1 대상 후보에 대한 L0 예측 방향의 참조 픽쳐의 출력 순서 정보(POCref0[Cand0])와 제2 대상 후보에 대한 L0 예측 방향의 참조 픽쳐의 출력 순서 정보(POCref0[Cand1])의 차이가 0인 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는 제1 대상 후보에 대한 L0 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L0 예측 방향의 제어점 벡터 간의 평균값으로 산출될 수 있다.
상기 POCref0[Cand0]와 POCref0[Cand1]의 차이가 0이 아닌 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 즉, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L0 예측 방향의 제어점 벡터 간의 평균값으로 산출되지 않을 수 있다. 대신에, 평균 후보에 대한 L0 예측 방향의 제어점 벡터는, 2개의 대상 후보 중 어느 하나에 대한 L0 예측 방향의 제어점 벡터로 설정/저장될 수 있다. 여기서, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 후보 리스트의 인덱스가 더 작은 후보(즉, Cand0)를 의미할 수 있다. 또는, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 참조 픽쳐 인덱스가 더 작은 후보를 의미할 수도 있다. 또는, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 현재 블록이 속한 현재 픽쳐의 출력 순서 정보와의 차이가 더 작은 후보를 의미할 수도 있다.
전술한 방법은 L1 예측 방향에 대해서도 동일하게 적용될 수 있다. 즉, 제1 대상 후보에 대한 L1 예측 방향의 참조 픽쳐의 출력 순서 정보(POCref1[Cand0])와 제2 대상 후보에 대한 L1 예측 방향의 참조 픽쳐의 출력 순서 정보(POCref1[Cand1])의 차이가 0인지 여부에 기초하여, 2개의 대상 후보의 평균 여부를 결정할 수 있다.
구체적으로, POCref1[Cand0]와 POCref1[Cand1]의 차이가 0인 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L1 예측 방향의 제어점 벡터는 제1 대상 후보에 대한 L1 예측 방향의 제어점 벡터와 제2 대상 후보에 대한 L1 예측 방향의 제어점 벡터 간의 평균값으로 산출될 수 있다. 반면, POCref1[Cand0]와 POCref1[Cand1]의 차이가 0이 아닌 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 대신에, 평균 후보에 대한 L1 예측 방향의 제어점 벡터는, 2개의 대상 후보 중 어느 하나에 대한 L1 예측 방향의 제어점 벡터로 설정/저장될 수 있다.
디코딩 장치에 기-정의된 조건을 만족하는 경우에 한하여, 평균 후보는 2개의 대상 후보를 기반으로 생성되어 후보 리스트에 추가될 수 있다. 여기서, 기-정의된 조건은 전술한 (조건 1) 내지 (조건 3) 중 어느 하나일 수 있다. 또는 기-정의된 조건은 전술한 (조건 1) 내지 (조건 3) 중 적어도 둘을 포함할 수 있다.
디코딩 장치에 복수의 조건들이 정의된 경우, 소정의 우선순서에 따라 순차적으로 해당 조건을 만족하는지 여부를 결정할 수 있다. 일예로, 디코딩 장치에 (조건 1)과 (조건 2)가 정의된 경우, 2개의 대상 후보가 (조건 1)을 만족하는지 여부를 먼저 결정하고, (조건 1)을 만족하는 경우에 한하여 2개의 대상 후보가 (조건 2)를 만족하는지 여부를 결정할 수 있다. 또는, 2개의 대상 후보에 대해서 (조건 1)과 (조건 2)를 만족하는지 여부를 동시에 결정할 수도 있다.
각 예측 방향에 대해서, 2개의 대상 후보를 기반으로 평균 후보가 생성된 경우에 한하여 해당 평균 후보가 후보 리스트에 추가될 수 있다. 또는, L0 및 L1 예측 방향 중 어느 하나에 대해서라도 전술한 조건이 만족되어 2개의 대상 후보를 기반으로 평균 후보가 생성된 경우라면, 해당 평균 후보가 후보 리스트에 추가될 수 있다.
상기 어파인 후보는 조합된 후보(combined candidate)를 포함할 수 있다. 조합된 후보는, 후보 리스트 내에 포함된 2개의 대상 후보(Cand0, Cand1)를 조합하여 생성될 수 있다. 조합된 후보는, Cand0의 제1 예측 방향의 움직임 정보와 Cand1의 제2 예측 방향의 움직임 정보를 조합하여 생성될 수 있다. 여기서, 제1 예측 방향은 L0 예측 방향 또는 L1 예측 방향 중 어느 하나이고, 제2 예측 방향은 L0 예측 방향 또는 L1 예측 방향 중 다른 하나일 수 있다.
상기 조합된 후보는 어파인 모드를 위한 어파인 후보로서 독립적으로 정의될 수 있고, 전술한 평균 부호와 함께 이용될 수 있다. 또는, 후보 리스트를 구성함에 있어서, 조합된 후보 또는 평균 후보 중 어느 하나를 선택적으로 이용하는 형태로 구현될 수도 있고, 상기 조합된 후보가 평균 후보를 대체하는 형태로 구현될 수도 있다.
Idx 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
CandL0[candNum] 0 1 0 2 1 2 0 3 1 3 2 3 0 4 1
CandL1[candNum] 1 0 2 0 2 1 3 0 3 1 3 2 4 0 4
Idx 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
CandL0[candNum] 4 2 4 3 4 0 5 1 5 2 5 3 5 4 5
CandL1[candNum] 1 4 2 4 3 5 0 5 1 5 2 5 3 5 4
표 2를 참조하면, Idx는 조합의 우선순서를 나타내며, Idx에 대응하는 2개의 대상 후보를 기반으로 조합된 후보가 생성될 수 있다. 예를 들어, Idx가 0인 경우, 후보 리스트에서 0의 인덱스를 가진 머지 후보의 L0 예측 방향의 움직임 정보와 후보 리스트에서 1의 인덱스를 가진 머지 후보의 L1 예측 방향의 움직임 정보를 조합하여 조합된 후보를 생성할 수 있다.
상기 조합된 후보는, 상기 특정된 2개의 대상 후보에 대한 어파인 모델의 타입이 서로 동일한 경우에 생성될 수 있다.
또는, 2개의 대상 후보에 대한 어파인 모델의 타입이 서로 다른 경우, 4-parameter 또는 6-parameter 중 어느 하나로 어파인 모델의 타입을 저장/전파할 수 있다.
일예로, 4-parameter인 대상 블록이 제1 제어점 벡터와 제2 제어점 벡터를 가지는 경우, 제1 제어점 벡터와 제2 제어점 벡터를 이용하여 제3 제어점 벡터를 유도할 수 있다. 또는, 4-parameter인 대상 블록이 제1 제어점 벡터와 제3 제어점 벡터를 가지는 경우, 제1 제어점 벡터와 제3 제어점 벡터를 이용하여 제2 제어점 벡터를 유도할 수 있다. 그런 다음, 6-parameter인 다른 대상 후보와의 조합을 통해 조합된 후보를 생성할 수 있다. 이때, 조합된 후보에 대한 어파인 모델의 타입은 6-parameter로 설정될 수 있다. 상기 조합된 후보가 현재 블록의 제어점 벡터를 유도하기 위해 이용된 경우, 조합된 후보에 대한 어파인 모델의 타입은 현재 블록에 저장될 수 있고, 현재 블록 이후의 블록으로 전파될 수도 있다.
또는, 2개의 대상 후보에 대한 어파인 모델의 타입이 서로 다른 경우, parameter의 개수가 적은 타입을 저장/전파할 수 있다. 즉, Cand0에 대한 어파인 모델의 타입이 4-parameter이고, Cand1에 대한 어파인 모델의 타입이 6-parameter인 경우, 조합된 후보에 대한 어파인 모델의 타입은 Cand0의 어파인 모델의 타입과 동일하게 4-parameter로 설정될 수 있다. 상기 조합된 후보가 현재 블록의 제어점 벡터를 유도하기 위해 이용된 경우, 조합된 후보에 대한 어파인 모델의 타입은 현재 블록에 저장될 수 있고, 현재 블록 이후의 블록으로 전파될 수도 있다.
또는, 2개의 대상 후보에 대한 어파인 모델의 타입이 서로 다른 경우, 2개의 대상 후보 간 제어점 벡터들의 차이가 소정의 문턱값보다 작은 경우, 4-parameter로 저장/전파하고, 그렇지 않은 경우, 6-parameter로 저장/전파할 수 있다. 여기서, 제어점 벡터들의 차이는, Cand0의 제1 제어점 벡터와 Cand1의 제1 제어점 벡터 간의 제1 차이를 의미할 수 있다. 상기 제어점 벡터들의 차이는, Cand0의 제2 제어점 벡터와 Cand1의 제2 제어점 벡터 간의 제2 차이를 의미할 수 있다. 상기 제어점 벡터들의 차이는, Cand0의 제3 제어점 벡터와 Cand1의 제3 제어점 벡터 간의 제3 차이를 의미할 수 있다. 상기 제어점 벡터들의 차이는, Cand0에 대한 차분 제어점 벡터와 Cand1에 대한 차분 제어점 벡터 간의 제4 차이를 의미할 수 있다. 여기서, 차분 제어점 벡터는, (cpMV[0]-cpMV[1]), (cpMV[0]-cpMV[2]), (cpMV[1]-cpMV[2]), (cpMV[1]-cpMV[0]), (cpMV[2]-cpMV[0]), 또는 (cpMV[2]-cpMV[1]) 중 적어도 하나로 정의될 수 있다.
Cand0와 Cand1가 제2 제어점 벡터를 가지지 않을 수 있다. 이러한 경우, 전술한 제어점 벡터들의 차이를 산출하기 위해, 해당 대상 후보의 제1 제어점 벡터와 제3 제어점 벡터를 기반으로 제2 제어점 벡터가 유도될 수 있다. 마찬가지로, Cand0와 Cand1가 제3 제어점 벡터를 가지지 않을 수 있으며, 이러한 경우, 해당 대상 후보의 제1 제어점 벡터와 제2 제어점 벡터를 기반으로 제3 제어점 벡터가 유도될 수 있다. 상기 제어점 벡터들의 차이는, 상기 제1 차이, 제2 차이, 제3 차이, 또는 제4 차이 중 어느 하나를 의미할 수도 있고, 제1 내지 제4 차이 중 적어도 둘을 포함할 수도 있다. 상기 문턱값은 인코딩 장치와 디코딩 장치에 동일하게 기-정의된 값일 수 있다.
또는, 어파인 모델의 타입이 서로 동일한 2개의 대상 후보만을 이용하도록 제한될 수 있다. 예를 들어, 전술한 우선순서에 따라 순차적으로 2개의 대상 후보를 선택하되, 선택된 2개의 대상 후보가 서로 상이한 어파인 모델의 타입을 가지는 경우, 해당 대상 후보를 기반으로 조합된 후보가 생성되지 않을 수 있다. 반면, 선택된 2개의 대상 후보가 동일한 어파인 모델의 타입을 가지는 경우, 해당 대상 후보를 기반으로 조합된 후보가 생성될 수 있다.
또한, 어파인 모델의 타입을 고려하여 후보 리스트 내의 어파인 후보들을 재배열한 뒤, 조합된 후보를 생성할 수도 있다. 후보 리스트에 서로 다른 어파인 모델의 타입이 혼재되어 있을 경우, 어파인 모델의 타입을 고려하지 않고 전술한 우선순서에 따라 순차적으로 조합된 후보를 생성하면 해당 후보의 정확도가 낮아질 수 있다. 따라서, 조합된 후보를 생성하기 전에 후보 리스트의 어파인 후보들을 재배열하고, 재배열된 후보 리스트를 기반으로 조합된 후보를 생성할 수 있다.
일예로, 6-parameter인 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수 있다. 또는, 4-parameter인 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수도 있다. 또는, 후보 리스트의 첫번째 어파인 후보에 대한 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수 있다. 여기서, 첫번째 어파인 후보는 후보 리스트에 속한 복수의 어파인 후보 중에서 가장 작은 인덱스가 할당된 후보를 의미할 수 있다. 또는, 후보 리스트 내에 존재하는 어파인 후보들 중에서 가장 빈도수가 높은 어파인 모델의 타입을 우선으로 하여 후보 리스트를 재배열할 수도 있다.
또한, 평균 후보의 생성 방법에 대한 실시예들은 조합된 후보를 생성하기 위해 동일/유사하게 적용될 수 있으며, 자세한 설명은 생략하기로 한다.
상기 후보 리스트는, 제로 움직임 벡터를 가진 후보(이하, 제로 후보라 함)를 더 포함할 수 있다. 상기 제로 후보는, 후술할 어파인 후보의 재배열 이후에 상기 후보 리스트에 추가될 수 있다. 또는, 상기 후보 리스트에 상기 제로 움직임 벡터를 가진 후보가 추가된 이후에 어파인 후보의 재배열이 수행될 수도 있다.
전술한 복수의 어파인 후보는, 소정의 우선 순위에 기초하여 상기 후보 리스트에 배열될 수 있다. 일예로, 복수의 어파인 후보는, 상속된 후보, 구성된 후보, 평균 후보, 제로 후보의 순서로 후보 리스트에 배열될 수 있다. 또는, 복수의 어파인 후보는, 상속된 후보, 평균 후보, 구성된 후보, 제로 후보의 순서로 후보 리스트에 배열될 수도 있다. 또는, 상속된 후보는 공간적 후보와 시간적 후보로 구분될 수 있다. 이때, 복수의 어파인 후보는, 공간적 후보, 평균 후보, 시간적 후보, 구성된 후보, 제로 후보의 순서로 후보 리스트에 배열될 수 있다.
다양한 평균 후보를 생성하기 위해서는 후보 리스트에 다양한 어파인 후보가 포함될 필요가 있다. 이를 위해, 평균 후보의 생성에 앞서, 상속된 후보 및/또는 구성된 후보의 구성 방법과 그 후보의 개수가 변경될 수 있다. 즉, 상속된 후보 및/또는 구성된 후보의 구성 방법 및 그 개수의 변경은 평균 후보가 후보 리스트에 추가되는 경우에 적용될 수 있다. 다만 이에 한정되지 아니하며, 평균 후보의 이용 여부와는 별개로 독립적으로 적용될 수도 있다.
상속된 후보를 구성하는 과정에 있어서, 현재 블록에 인접한 좌측 주변 블록들 중에서 어파인 모드로 부호화된 어느 하나의 블록과 현재 블록에 인접한 상단 주변 블록들 중 어파인 모드로 부호화된 어느 하나의 블록을 이용하여 2개의 상속된 후보를 구성할 수 있다. 또는, 보다 다양한 상속된 후보를 구성하고, 생성 가능한 평균 후보의 개수를 늘리기 위해, 좌측 주변 블록들과 상단 주변 블록들 각각으로부터 어파인 모드로 부호화된 둘 이상의 블록들을 이용할 수도 있다. 이를 위해, 주변 블록의 체크 순서가 다음 표 2와 같이 변경될 수 있다. 상기 주변 블록의 체크 순서는, 해당 주변 블록을 기반으로 유도된 상속된 후보가 후보 리스트에 삽입되는 순서를 의미할 수도 있다.
Case 1: A1, B1, B0, A0, B2
Case 2: B1, A1, B0, A0, B2
Case 3: A1, B1, B0, A0, (if Num < 4, B2)
Case 4: B1, A1, B0, A0, (if Num < 4, B2)
Case 5: w > h ? A1, B1, B0, A0, B2 : B1, A1, B0, A0, B2
Case 6: w > h ? A1, B1, B0, A0, (if Num < 4, B2) : B1, A1, B0, A0, (if Num < 4, B2)
표 3에서, A0은 좌하단 주변 블록, A1은 좌측 주변 블록, B0는 우상단 주변 블록, B1은 상단 주변 블록, B2은 좌상단 주변 블록을 각각 의미할 수 있다. 일예로, 현재 블록의 좌상단 샘플의 좌표가 (x, y)이고, 현재 블록의 너비와 높이가 각각 w와 h라 가정할 경우, A0은 (x-1, y+h) 좌표의 샘플을 포함한 주변 블록, A1은 (x-1, y+h-1) 좌표의 샘플을 포함한 주변 블록, B0은 (x+w, y-1) 좌표의 샘플을 포함한 주변 블록, B1은 (x+w-1, y-1) 좌표의 샘플을 포함한 주변 블록, B2은 (x-1, y-1) 좌표의 샘플을 포함한 주변 블록을 각각 의미할 수 있다.
상기 5개의 주변 블록들에 대해서, Case 1 또는 Case 2에 따른 순서대로 해당 주변 블록이 어파인 모드로 부호화된 블록인지 여부를 체크하고, 어파인 모드로 부호화된 블록을 기반으로 후보 리스트를 구성할 수 있다.
또는, Case 3 또는 Case 4에 따른 순서대로 해당 주변 블록이 어파인 모드로 부호화된 블록인지 여부를 체크하여 후보 리스트를 구성할 수 있다. 다만, 후보 리스트에 기-포함된 어파인 후보의 개수를 고려하여, B2의 체크 여부가 결정될 수 있다.
또는, Case 5 또는 Case 6과 같이, 현재 블록의 형태에 따라 다른 체크 순서가 적용될 수도 있다. 일예로, 현재 블록의 너비가 높이보다 큰 경우, A1, B1, B0, A0, B2의 체크 순서가 적용되고, 현재 블록의 너비가 높이보다 작거나 같은 경우, B1, A1, B0, A0, B2의 체크 순서가 적용될 수 있다. 역으로, 현재 블록의 너비가 높이보다 큰 경우, B1, A1, B0, A0, B2의 체크 순서가 적용되고, 현재 블록의 너비가 높이보다 작거나 같은 경우, A1, B1, B0, A0, B2의 체크 순서가 적용될 수도 있다.
또는, 현재 블록의 너비와 높이의 비율이 소정의 문턱값보다 큰 경우, A1, B1, B0, A0, B2의 체크 순서가 적용되고, 현재 블록의 너비와 높이의 비율이 소정의 문턱값보다 작거나 같은 경우, B1, A1, B0, A0, B2의 체크 순서가 적용될 수 있다. 역으로, 현재 블록의 너비와 높이의 비율이 소정의 문턱값보다 큰 경우, B1, A1, B0, A0, B2의 체크 순서가 적용되고, 현재 블록의 너비와 높이의 비율이 소정의 문턱값보다 작거나 같은 경우, A1, B1, B0, A0, B2의 체크 순서가 적용될 수도 있다.
전술한 실시예는 최대 5개의 주변 블록들이 이용 가능한 경우를 전제로 한다. 그러나, 현재 블록의 형태 또는 너비와 높이의 비율에 따라, 5개의 주변 블록들 중 적어도 하나가 이용 가능하지 않은 것으로 설정될 수 있다. 이 경우, 이용 가능하지 않은 주변 블록에 대해서는 전술한 체크 과정이 생략될 수 있다.
구성된 후보를 구성하는 과정에 있어서, 현재 블록의 각 제어점에 인접한 공간적/시간적 주변 블록을 이용하는 방법에 대해서 살펴보기로 한다.
이하, B2는 좌상단 주변 블록, B3는 상단 주변 블록, A2는 좌측 주변 블록을 각각 의미하고, B2, B3 및 A2는 현재 블록의 좌상단 코너에 대응하는 제1 제어점에 인접한 것일 수 있다. B1은 상단 주변 블록, B0은 우상단 주변 블록을 각각 의미하고, B1 및 B0는 현재 블록의 우상단 코너에 대응하는 제2 제어점에 인접한 것일 수 있다. A1은 좌측 주변 블록, A0은 좌하단 주변 블록을 각각 의미하고, A1 및 A0은 현재 블록의 좌하단 코너에 대응하는 제3 제어점에 인접한 것일 수 있다. T0은 시간적 주변 블록으로서, 현재 블록의 우하단 코너에 대응하는 제4 제어점에 시간적으로 인접한 것일 수 있다.
일예로, 현재 블록의 좌상단 샘플의 좌표가 (x, y)이고, 현재 블록의 너비와 높이가 각각 w와 h라 가정한다. 이 경우, B2는 (x-1, y-1) 좌표의 샘플을 포함한 주변 블록, B3는 (x, y-1) 좌표의 샘플을 포함한 주변 블록, A2는 (x-1, y) 좌표의 샘플을 포함한 주변 블록을 각각 의미할 수 있다. 또한, B1은 (x+w-1, y-1) 좌표의 샘플을 포함한 주변 블록, B0은 (x+w, y-1) 좌표의 샘플을 포함한 주변 블록을 각각 의미할 수 있다. 또한, A1은 (x-1, y+h-1) 좌표의 샘플을 포함한 주변 블록, A0은 (x-1, y+h) 좌표의 샘플을 포함한 주변 블록을 각각 의미할 수 있다. T0는 (x+w, y+h) 좌표의 샘플을 포함한 주변 블록을 의미할 수 있다.
상기 현재 블록의 공간적/시간적 주변 블록은 복수의 그룹으로 구분될 수 있다. 설명의 편의를 위해, 현재 블록의 공간적/시간적 주변 블록은 4개의 그룹으로 구분됨을 가정한다. 상기 4개의 그룹은, B2, B3, 또는 A2 중 적어도 하나로 구성된 제1 그룹, B1 또는 B0 중 적어도 하나로 구성된 제2 그룹, A1 또는 A0 중 적어도 하나로 구성된 제3 그룹, T0로 구성된 제4 그룹을 의미할 수 있다. 다만, 이에 한정되지 아니하며, 후술하는 4개의 그룹 중 2개 또는 3개의 그룹으로 구분될 수 있음은 물론이다.
각 그룹에 대하여, 각 그룹에 속한 하나 또는 그 이상의 주변 블록들을 소정의 탐색 순서에 따라 순차적으로 탐색하여 각 제어점에 대응하는 제어점 벡터를 유도할 수 있다. 상기 탐색은, 이용 가능한 하나의 움직임 벡터가 유도될 때까지 수행될 수 있다. 즉, 탐색 과정에서 이용 가능한 움직임 벡터가 하나라도 유도될 경우, 상기 탐색은 조기 종료될 수 있다. 또는, 상기 탐색은, 각 그룹에 속한 모든 주변 블록들로부터 이용 가능한 모든 움직임 벡터가 유도될 때까지 수행될 수 있다. 이 경우, 각 그룹 별로 하나 또는 그 이상의 움직임 벡터가 유도될 수 있다.
일예로, 제1 그룹의 경우, B2, B3, A2의 순서로 탐색을 수행하여 이용 가능한 하나 또는 그 이상의 움직임 벡터를 유도할 수 있다. 제2 그룹의 경우, B1, B0의 순서로 탐색을 수행하여 이용 가능한 하나 또는 그 이상의 움직임 벡터를 유도할 수 있다. 제3 그룹의 경우, A1, A0의 순서로 탐색을 수행하여 이용 가능한 하나 또는 그 이상의 움직임 벡터를 유도할 수 있다. 각 그룹 별로 유도된 움직임 벡터들을 이용하여 구성된 후보가 생성될 수 있다. 또는, 각 그룹 별로 유도된 움직임 벡터들 중에서, 동일한 참조 픽쳐 인덱스를 가진 움직임 벡터들만을 이용하여 구성된 후보가 생성될 수 있다. 또는, 각 그룹 별로 유 도된 움직임 벡터들 중에서, 동일한 참조 픽쳐를 가진 움직임 벡터들만을 이용하여 구성된 후보가 생성될 수 있다.
각 그룹으로부터 동일한 참조 픽쳐 인덱스를 가진 주변 블록들을 추출하고, 추출된 주변 블록들의 움직임 벡터들을 기반으로 구성된 후보가 생성될 수 있다.
일예로, 표 4의 Case1 또는 Case 2와 같이, 각 그룹 별로 소정의 탐색 순서에 따라 참조 픽쳐 인덱스가 0인 주변 블록과 참조 픽쳐 인덱스가 1인 주변 블록을 각각 도출할 수 있다. 이때, 참조 픽쳐 인덱스가 0인 주변 블록이 B2, B1, A1, TO인 경우, 구성된 후보는 B2의 움직임 벡터(cpMV[0]), B1의 움직임 벡터(cpMV[1]), A1의 움직임 벡터(cpMV[2]), 또는 T0의 움직임 벡터(cpMV[3]) 중 적어도 둘의 조합을 기반으로 생성될 수 있다. 즉, 구성된 후보는, (cpMV[0], cpMV[1], cpMV[2]), (cpMV[0], cpMV[1], cpMV[3]), (cpMV[0], cpMV[2], cpMV[3]), (cpMV[1], cpMV[2], cpMV[3]), (cpMV[0], cpMV[1]), (cpMV[0], cpMV[2]), (cpMV[0], cpMV[3]), (cpMV[1], cpMV[2]), (cpMV[1], cpMV[3]), 또는 (cpMV[2], cpMV[3]) 중 적어도 하나를 포함할 수 있다. 동일한 방법을 통해, 참조 픽쳐 인덱스가 1인 주변 블록의 움직임 벡터를 기반으로, 구성된 후보가 생성될 수 있으며, 자세한 설명은 생략하기로 한다.
이와 같이, 동일한 참조 픽쳐 인덱스를 가진 주변 블록들의 움직임 벡터들을 기반으로 구성된 후보가 생성될 수 있고, 이러한 구성된 후보는 참조 픽쳐 인덱스 별로 각각 생성될 수 있다.
Case 1: 각 제어점에 인접한 공간적/시간적 주변 블록들 중에서 참조 픽쳐 인덱스가 0인 주변 블록을 (B2->B3->A2), (B1->B0), (A1->A0), T0의 순서로 체크하고, 동일한 방법으로 참조 픽쳐 인덱스가 1인 주변 블록을 체크하여, 참조 픽쳐 인덱스 별로 구성된 후보를 구성함
Case 2: 각 제어점에 인접한 공간적/시간적 주변 블록들 중에서 참조 픽쳐 인덱스가 0인 주변 블록을 (B2->A2->B3), (B0->B1), (A0->A1), T0의 순서로 체크하고, 동일한 방법으로 참조 픽쳐 인덱스가 1인 주변 블록을 체크하여, 참조 픽쳐 인덱스 별로 구성된 후보를 구성함
상기 실시예에서는, 시간적 주변 블록인 T0에 대해서도 참조 픽쳐 인덱스를 고려하였으나, 이는 일예에 불과하다. T0는, 참조 픽쳐 인덱스의 동일 여부와 관계없이, 구성된 후보를 생성하기 위해 이용될 수 있다. 또는, T0에 대해서는 구성된 후보를 생성하기 위해 이용되지 않도록 제한될 수도 있다.
또한, 상기 실시예에서는 동일한 참조 픽쳐 인덱스를 가진 주변 블록을 기반으로 구성된 후보가 생성되나, 이는 일예에 불과하다. 각 그룹으로부터 동일한 참조 픽쳐를 가진 주변 블록들이 도출되고, 도출된 주변 블록들에 대한 움직임 벡터들의 조합을 통해 구성된 후보가 생성될 수도 있다.
후보 리스트를 구성하는 과정에서 중복 체크 과정이 수행될 수 있다. 후보 리스트에는 상속된 후보, 구성된 후보, 평균 후보, 또는 제로 후보 중 적어도 하나가 추가될 수 있으며, 각 어파인 후보은 후보 리스트에 기-포함된 어파인 후보와의 중복 체크를 통해 선택적으로 추가될 수 있다.
어파인 모드의 경우, 4-parameter, 6-parameter 등의 다양한 어파인 모델의 타입이 이용될 수 있으며, 이로 인해 어파인 후보마다 제어점 벡터의 개수가 서로 다를 수 있다. 따라서, 어파인 후보들 간의 중복 체크를 위해 제어점 벡터의 개수를 동일하게 맞출 필요가 있다. 중복 체크는, 후보 리스트에 기-포함된 어파인 후보(이하, 이전 어파인 후보라 함)와 후보 리스트에 추가 대상인 어파인 후보(이하, 현재 어파인 후보라 함) 간에 수행될 수 있다.
일예로, 현재 어파인 후보의 어파인 모델의 타입이 4-parameter인 경우, 제1 제어점 벡터(cpMV[0])와 제2 제어점 벡터(cpMV[1])만을 포함할 수 있다. 이러한 경우, 제3 제어점에서의 움직임 벡터를 유도하고, 이를 제3 제어점 벡터(cpMV[2])에 할당함으로써, 6-parameter인 어파인 모델의 타입을 가진 이전 어파인 후보와의 중복 체크를 가능하게 할 수 있다. 또는, 현재 어파인 후보의 어파인 모델의 타입이 4-parameter인 경우, 제1 제어점 벡터(cpMV[0])와 제3 제어점 벡터(cpMV[2])만을 포함할 수 있다. 이러한 경우, 제2 제어점에서의 움직임 벡터를 유도하고, 이를 제2 제어점 벡터(cpMV[1])에 할당함으로써, 6-parameter인 어파인 모델의 타입을 가진 이전 어파인 후보와의 중복 체크를 가능하게 할 수 있다.
또는, 상기 중복 체크는, 동일한 어파인 모델의 타입을 가진 어파인 후보들 간에 수행되는 것으로 제한될 수 있다.
일예로, 현재 어파인 후보의 어파인 모델의 타입이 4-parameter인 경우, 후보 리스트에 기-포함된 어파인 후보들 중에서, 어파인 모델의 타입이 4-parameter인 어파인 후보와 중복 체크를 수행할 수 있다. 또는, 현재 어파인 후보의 어파인 모델의 타입이 6-parameter인 경우, 후보 리스트에 기-포함된 어파인 후보들 중에서, 어파인 모델의 타입이 6-parameter인 어파인 후보와 중복 체크를 수행할 수 있다.
상기 중복 체크는, 제어점 벡터의 동일 여부에 기초하여 수행될 수 있다. 여기서, 제어점 벡터의 동일 여부는, 제1 내지 제3 제어점 벡터 각각에 대해서 확인할 수 있다. 또는, 제1 내지 제3 제어점 벡터 중 어느 하나에 대해서 제어점 벡터의 동일 여부를 확인하고, 나머지 제어점 벡터에 대해서는 확인하지 않을 수 있다. 여기서, 제1 내지 제3 제어점 벡터 중 어느 하나는, 제1 제어점 벡터로 제한될 수 있다. 또는, 제1 내지 제3 제어점 벡터 중 어느 둘에 대해서 제어점 벡터의 동일 여부를 확인하고, 나머지 제어점 벡터에 대해서는 확인하지 않을 수 있다. 여기서, 제1 내지 제3 제어점 벡터 중 어느 둘은, 제1 제어점 벡터와 제2 제어점 벡터로 제한될 수 있다.
일예로, 제어점 벡터의 동일 여부에 기초한 중복 체크는 다음 표 5와 같이 수행될 수 있다.
if( (Cand[i][0] != CurrCand[0]) || (Cand[i][1] != CurrCand[1]) || (Cand[i][2] != CurrCand[2]) )
bPruned = false;
else
bPruned = true;
표 5에서, Cand[i][0], Cand[i][1] 및 Cand[i][2]는 이전 어파인 후보의 제1 제어점 벡터, 제2 제어점 벡터 및 제3 제어점 벡터를 각각 의미할 수 있다. CurrCand[0], CurrCand[1] 및 CurrCand[2]는 현재 어파인 후보의 제1 제어점 벡터, 제2 제어점 벡터 및 제3 제어점 벡터를 각각 의미할 수 있다.
표 5와 같이, 이전 어파인 후보와 현재 어파인 후보 간에 각 제어점 벡터의 동일 여부를 확인할 수 있다. 만일 현재 어파인 후보의 제1 내지 제3 제어점 벡터 중 적어도 하나가 이전 어파인 후보와 상이한 경우((Cand[i][0] != CurrCand[0]), (Cand[i][1] != CurrCand[1]), 또는 (Cand[i][2] != CurrCand[2])), 현재 어파인 후보는 후보 리스트에 추가될 수 있다. 반면, 현재 어파인 후보의 제1 내지 제3 제어점 벡터와 이전 어파인 후보의 제1 내지 제3 제어점 벡터가 모두 동일한 경우, 현재 어파인 후보는 후보 리스트에 추가되지 않을 수 있다.
또는, 중복 체크는, 제어점 벡터 차이의 동일 여부에 기초하여 수행될 수 있다. 여기서, 제어점 벡터 차이는, 제2 제어점 벡터와 제1 제어점 벡터의 차이(이하, 제1 제어점 벡터 차이라 함) 또는 제3 제어점 벡터와 제1 제어점 벡터의 차이(이하, 제2 제어점 벡터 차이라 함) 중 적어도 하나를 포함할 수 있다. 또는, 상기 제어점 벡터 차이는 제3 제어점 벡터와 제2 제어점 벡터의 차이인 제3 제어점 벡터 차이를 더 포함할 수도 있다.
일예로, 제어점 벡터 차이의 동일 여부에 기초한 중복 체크는 다음 표 6과 같이 수행될 수 있다.
diffMV0[i] = Cand[i][1] - Cand[i][0];
diffMV1[i] = Cand[i][2] - Cand[i][0];
diffCurrMV0 = CurrCand[1] - CurrCand[0];
diffCurrMV1 = CurrCand[2] - CurrCand[0];

if( (diffMV0[i] != diffCurrMV0) || (diffMV1[i] != diffCurrMV1) )
bPruned = false;
else
bPruned = true;
표 6에서, diffMV0[i]는 이전 어파인 후보의 제1 제어점 벡터 차이를, diffMV1[i]는 이전 어파인 후보의 제2 제어점 벡터 차이를 각각 의미할 수 있다. 또한, diffCurrMV0는 현재 어파인 후보의 제1 제어점 벡터 차이를, diffCurrMV1는 제2 제어점 벡터 차이를 각각 의미할 수 있다.
표 6과 같이, 이전 어파인 후보와 현재 어파인 후보 간에 각 제어점 벡터 차이의 동일 여부를 확인할 수 있다. 만일 현재 어파인 후보의 제1 및 제2 제어점 벡터 차이 중 적어도 하나가 이전 어파인 후보와 상이한 경우((diffMV0[i] != diffCurrMV0) 또는 (diffMV1[i] != diffCurrMV1)), 현재 어파인 후보는 후보 리스트에 추가될 수 있다. 반면, 현재 어파인 후보의 제1 및 제2 제어점 벡터 차이와 이전 어파인 후보의 제1 및 제2 제어점 벡터 차이가 모두 동일한 경우, 현재 어파인 후보는 후보 리스트에 추가되지 않을 수 있다.
또는, 중복 체크는, 제어점 벡터 차이의 동일 여부 및 제어점 벡터의 동일 여부에 기초하여 수행될 수 있다. 전술한 바와 같이, 상기 제어점 벡터 차이는, 제1 제어점 벡터 차이 또는 제2 제어점 벡터 차이 중 적어도 하나를 포함할 수 있다. 상기 제어점 벡터 차이는 제3 제어점 벡터와 제2 제어점 벡터의 차이인 제3 제어점 벡터 차이를 더 포함할 수도 있다. 또한, 제어점 벡터의 동일 여부는, 제1 내지 제3 제어점 벡터 각각에 대해서 확인할 수 있다. 또는, 제1 내지 제3 제어점 벡터 중 어느 하나에 대해서 제어점 벡터의 동일 여부를 확인하고, 나머지 제어점 벡터에 대해서는 확인하지 않을 수 있다. 여기서, 제1 내지 제3 제어점 벡터 중 어느 하나는, 제1 제어점 벡터로 제한될 수 있다. 또는, 제1 내지 제3 제어점 벡터 중 어느 둘에 대해서 제어점 벡터의 동일 여부를 확인하고, 나머지 제어점 벡터에 대해서는 확인하지 않을 수 있다. 여기서, 제1 내지 제3 제어점 벡터 중 어느 둘은, 제1 제어점 벡터와 제2 제어점 벡터로 제한될 수 있다.
일예로, 중복 체크는 다음 표 7과 같이 수행될 수 있다.
diffMV0[i] = Cand[i][1] - Cand[i][0];
diffMV1[i] = Cand[i][2] - Cand[i][0];
diffCurrMV0 = CurrCand[1] - CurrCand[0];
diffCurrMV1 = CurrCand[2] - CurrCand[0];

if( (diffMV0[i] != diffCurrMV0) || (diffMV1[i] != diffCurrMV1) || (Cand[i][0] != CurrCand[0]) )
bPruned = false;
else
bPruned = true;
표 7에서, diffMV0[i]는 이전 어파인 후보의 제1 제어점 벡터 차이를, diffMV1[i]는 이전 어파인 후보의 제2 제어점 벡터 차이를 각각 의미할 수 있다. 또한, diffCurrMV0는 현재 어파인 후보의 제1 제어점 벡터 차이를, diffCurrMV1는 제2 제어점 벡터 차이를 각각 의미할 수 있다. Cand[i][0]는 이전 어파인 후보의 제1 제어점 벡터를, CurrCand[0]는 현재 어파인 후보의 제1 제어점 벡터를 각각 의미할 수 있다.
표 7과 같이, 이전 어파인 후보와 현재 어파인 후보 간에 각 제어점 벡터 차이의 동일 여부 및 이전 어파인 후보와 현재 어파인 후보 간에 제1 제어점 벡터의 동일 여부를 확인할 수 있다.
만일 현재 어파인 후보의 제1 및 제2 제어점 벡터 차이 중 적어도 하나가 이전 어파인 후보와 상이한 경우((diffMV0[i] != diffCurrMV0) 또는 (diffMV1[i] != diffCurrMV1)), 현재 어파인 후보는 후보 리스트에 추가될 수 있다. 또는, 현재 어파인 후보의 제1 및 제2 제어점 벡터 차이와 이전 어파인 후보의 제1 및 제2 제어점 벡터 차이가 모두 동일하더라도, 이전 어파인 후보의 제1 제어점 벡터와 현재 어파인 후보의 제1 제어점 벡터가 상이한 경우(Cand[i][0] != CurrCand[0]), 현재 어파인 후보는 후보 리스트에 추가될 수 있다. 반면, 이전 어파인 후보의 제1 및 제2 제어점 벡터 차이와 현재 어파인 후보의 제1 및 제2 제어점 벡터 차이가 모두 동일하고, 이전 어파인 후보의 제1 제어점 벡터와 현재 어파인 후보의 제1 제어점 벡터가 동일한 경우, 현재 어파인 후보는 후보 리스트에 추가되지 않을 수 있다.
또는, 중복 체크는, 이전 어파인 후보와 현재 어파인 후보 간 제어점 벡터 차이의 차이가 소정의 문턱값보다 작은지 여부 또는 이전 어파인 후보와 현재 어파인 후보 간 제어점 벡터의 차이가 소정의 문턱값보다 작은지 여부 중 적어도 하나에 기초하여 수행될 수 있다.
상기 제어점 벡터 차이의 차이는, 이전 어파인 후보의 제1 제어점 벡터 차이와 현재 어파인 후보의 제1 제어점 벡터 차이의 차이, 이전 어파인 후보의 제2 제어점 벡터 차이와 현재 어파인 후보의 제2 제어점 벡터 차이의 차이, 또는 이전 어파인 후보의 제3 제어점 벡터 차이와 현재 어파인 후보의 제3 제어점 벡터 차이의 차이 중 적어도 하나를 포함할 수 있다.
상기 이전 어파인 후보와 현재 어파인 후보 간 제어점 벡터의 차이는, 이전 어파인 후보의 제1 제어점 벡터와 현재 어파인 후보의 제1 제어점 벡터의 차이, 이전 어파인 후보의 제2 제어점 벡터와 현재 어파인 후보의 제2 제어점 벡터의 차이, 또는 이전 어파인 후보의 제3 제어점 벡터와 현재 어파인 후보의 제3 제어점 벡터의 차이 중 적어도 하나를 포함할 수 있다.
상기 문턱값은 비교 대상에 따라 일부 동일하거나 상이할 수 있다. 상기 문턱값은 블록의 형태, 크기, 또는 픽쳐 해상도 등을 고려하여 가변적으로 결정되거나, 인코딩 장치와 디코딩 장치에 동일하게 기-정의된 값일 수 있다.
일예로, 상기 중복 체크는 다음 표 8과 같이 수행될 수 있다.
diffMV0[i] = Cand[i][1] - Cand[i][0];
diffMV1[i] = Cand[i][2] - Cand[i][0];
diffCurrMV0 = CurrCand[1] - CurrCand[0];
diffCurrMV1 = CurrCand[2] - CurrCand[0];

if( (|diffMV0[i] - diffCurrMV0| < threshold_0) || (|diffMV1[i] - diffCurrMV1| < threshold_1)) || (|Cand[i][0] - CurrCand[0]| < threshold_2) )
bPruned = false;
else
bPruned = true;
표 8에서, diffMV0[i]는 이전 어파인 후보의 제1 제어점 벡터 차이를, diffMV1[i]는 이전 어파인 후보의 제2 제어점 벡터 차이를 각각 의미할 수 있다. 또한, diffCurrMV0는 현재 어파인 후보의 제1 제어점 벡터 차이를, diffCurrMV1는 제2 제어점 벡터 차이를 각각 의미할 수 있다. Cand[i][0]는 이전 어파인 후보의 제1 제어점 벡터를, CurrCand[0]는 현재 어파인 후보의 제1 제어점 벡터를 각각 의미할 수 있다.
표 8과 같이, 이전 어파인 후보의 제1 제어점 벡터 차이와 현재 어파인 후보의 제1 제어점 벡터 차이의 차이(|diffMV0[i] - diffCurrMV0|)가 threshold_0보다 작거나, 이전 어파인 후보의 제2 제어점 벡터 차이와 현재 어파인 후보의 제2 제어점 벡터 차이의 차이(|diffMV1[i] - diffCurrMV1|)가 threshold_1보다 작거나, 또는 이전 어파인 후보의 제1 제어점 벡터와 현재 어파인 후보의 제2 제어점 벡터의 차이(|Cand[i][0] - CurrCand[0]|)가 threshold_2보다 작은 경우, 현재 어파인 후보는 후보 리스트에 추가될 수 있다. 반면, 그렇지 않은 경우, 현재 어파인 후보는 후보 리스트에 추가되지 않을 수 있다.
전술한 실시예들은 설명의 편의를 위해 간략하게 기술한 것이며, 양방향 예측의 경우, 양방향에 대해서 전술한 방법이 각각 적용될 수 있다. 이때, 양방향 중 적어도 하나에 대해서 전술한 실시예의 조건을 만족하는 경우, 현재 어파인 후보가 후보 리스트에 추가될 수 있다. 반면, 양방향에 대해서 전술한 실시예의 조건을 만족하지 않는 경우, 현재 어파인 후보는 후보 리스트에 추가되지 않을 수 있다.
상기 후보 리스트를 구성함에 있어서, 다양한 어파인 후보를 생성함으로써 압축 효율을 높일 수 있다. 특히, 어파인 모드를 위한 어파인 후보의 개수를 변경하지 않고, 머지 모드를 위한 머지 후보의 개수를 증가시켜서 다양한 후보를 구성하게 되면, 어파인 모드보다는 머지 모드의 선택 확률을 높이게 된다. 따라서, 어파인 모드를 위한 어파인 후보의 개수를 증가시킬 필요가 있다. 또한, 머지 모드를 위한 머지 후보와 동일한 개수의 어파인 후보를 유도함으로써, 구현의 용이성 및 이해도를 높일 수 있다. 일예로, 머지 모드를 위한 머지 후보의 개수가 10일 때, 어파인 모드를 위한 어파인 후보의 개수를 10 또는 (10-1)로 결정할 수 있다. 또는, 머지 모드를 위한 머지 후보의 개수가 15일 때, 어파인 모드를 위한 어파인 후보의 개수를 15 또는 (15-1)로 결정할 수 있다.
추가적으로, 상기 후보 리스트에 속한 복수의 어파인 후보들은 재배열될 수 있다. 상기 재배열은, 템플릿 기반의 코스트(cost)를 기반으로 수행될 수 있다. 상기 복수의 어파인 후보들은, 하나 또는 그 이상의 서브 그룹으로 그룹핑될 수 있다. 상기 후보 리스트에 속한 제로 움직임 벡터를 가진 후보는 상기 그룹핑 대상에서 제외될 수 있다.
어파인 후보의 타입 별로 그룹핑될 수도 있다. 일예로, 후보 리스트는, 상속된 후보로 구성된 제1 서브 그룹, 구성된 후보로 구성된 제2 서브 그룹, 히스토리 기반의 후보로 구성된 제3 서브 그룹, 평균 후보로 구성된 제4 서브 그룹, 또는 그외 나머지 후보로 구성된 제5 서브 그룹 중 적어도 하나로 나누어질 수 있다.
또는, 상기 서브 그룹의 크기는 인코딩 장치와 디코딩 장치에 동일하게 기-정의된 것으로서, 2, 3, 4, 5 또는 그 이상일 수 있다. 상기 서브 그룹의 크기는 현재 블록의 인터 예측 모드에 따라 상이하게 결정/정의될 수 있다. 일예로, 현재 블록이 머지 모드로 부호화된 경우, 서브 그룹의 크기는 5이고, 현재 블록이 어파인 모드로 부호화된 경우, 서브 그룹의 크기는 3일 수 있다.
상기 재배열은, 후보 리스트의 서브 그룹 별로 수행될 수 있다. 이때, 후보 리스트의 서브 그룹 전부에 대해서 수행될 수도 있고, 일부 서브 그룹에 대해서만 수행될 수도 있다. 예를 들어, 제로 움직임 벡터를 가진 후보가 포함된 서브 그룹에 대해서는 재배열이 수행되지 않을 수 있다. 또는, 후보 리스트의 인덱스 중 가장 작은 인덱스(즉, 0)를 가진 후보가 속한 서브 그룹에 대해서는 재배열이 수행되지 않고, 가장 큰 인덱스를 가진 후보가 속한 서브 그룹에 대해서 재배열이 수행될 수 있다. 반대로, 후보 리스트의 인덱스 중 가장 작은 인덱스를 가진 후보가 속한 서브 그룹에 대해서는 재배열이 수행되고, 가장 큰 인덱스를 가진 후보가 속한 서브 그룹에 대해서 재배열이 수행되지 않을 수 있다. 또는, 후보 리스트의 인덱스 중 가장 작은 인덱스를 가진 후보와 가장 큰 인덱스를 가진 후보가 각각 속한 서브 블록에 대해서는 재배열이 수행되지 않고, 나머지 서브 그룹에 대해서 재배열이 수행될 수 있다. 또는, 후보 리스트의 인덱스 중 가장 큰 인덱스를 가진 후보가 속한 서브 그룹에 대해서는 재배열이 수행되지 않고, 나머지 서브 그룹에 대해서 재배열이 수행될 수 있다.
상기 재배열을 위한 템플릿 기반의 코스트는, 현재 블록의 템플릿 영역에 속한 샘플들과 참조 템플릿 영역에 속한 샘플들 간의 SAD(sum of absolute difference)로 산출될 수 있다. 상기 SAD는 서브 그룹에 속한 복수의 어파인 후보 각각에 대해서 산출될 수 있다. 산출된 SAD의 오름차순으로 서브 그룹에 속한 복수의 어파인 후보를 재배열할 수 있다.
상기 현재 블록의 템플릿 영역은 현재 블록에 인접한 주변 영역일 수 있다. 여기서, 주변 영역은, 현재 블록의 상단 주변 영역, 좌측 주변 영역, 또는 좌상단 주변 영역 중 적어도 하나를 포함할 수 있다. 현재 블록의 너비와 높이가 W와 H라 가정할 경우, 상단 주변 영역은 WxN의 크기를, 좌측 주변 영역은 N x H의 크기를, 좌상단 주변 영역은 N x N의 크기를 각각 가질 수 있다. N은 1, 2, 3 또는 그 이상의 정수일 수 있다. N은 인코딩 장치와 디코딩 장치에 동일하게 기-정의된 값일 수도 있고, 현재 블록의 크기, 형태, 또는 인터 예측 모드 중 적어도 하나에 기초하여 가변적으로 결정될 수 있다.
상기 주변 영역의 범위는 현재 블록의 인터 예측 모드에 따라 상이하게 결정될 수도 있다. 일예로, 현재 블록의 인터 예측 모드에 따라, 상기 주변 영역은 상단 주변 영역과 좌측 주변 영역으로 구성될 수도 있고, 또는 상기 주변 영역은 상단 주변 영역 또는 좌측 주변 영역 중 어느 하나로 구성될 수도 있다.
상기 참조 템플릿 영역은, 후보 리스트에 속한 어파인 후보의 움직임 정보에 기초하여 결정될 수 있다. 일예로, 어파인 후보의 제어점 벡터를 기반으로 현재 블록에 대한 서브 블록 단위의 움직임 벡터가 유도될 수 있다. 현재 블록 내 상단 행과 좌측 열에 속한 각 서브 블록의 움직임 벡터를 기반으로 이에 대응하는 참조 서브 블록을 각각 결정할 수 있다. 각 참조 서브 블록의 서브 템플릿 영역을 기반으로 상기 참조 템플릿 영역을 구성할 수 있다.
이때, 현재 블록 내 상단 행에 속한 서브 블록에 대응하는 참조 서브 블록의 서브 템플릿 영역은, 상기 참조 서브 블록의 상단 주변 영역으로 구성될 수 있다. 참조 서브 블록의 너비와 높이가 subW와 subH라 가정할 경우, 상기 참조 서브 블록의 상단 주변 영역은 subW x N의 크기를 가질 수 있다. 반면, 현재 블록 내 좌측 열에 속한 서브 블록에 대응하는 참조 서브 블록의 서브 템플릿 영역은, 상기 참조 서브 블록의 좌측 주변 영역으로 구성될 수 있다. 마찬가지로, 상기 참조 서브 블록의 좌측 주변 영역은 N x subH의 크기를 가질 수 있다. 다만, 현재 블록 내 좌상단 서브 블록은 현재 블록 내 상단 행에 속하는 동시에 좌측 열에 속하므로, 상기 좌상단 서브 블록에 대응하는 참조 서브 블록의 서브 템플릿 영역은, 상기 참조 서브 블록의 상단 및 좌측 주변 영역으로 구성될 수 있다. 여기서, 상기 참조 서브 블록의 상단 주변 영역은 subW x N의 크기를, 좌측 주변 영역은 N x subH의 크기를 각각 가질 수 있다.
상기 참조 템플릿 영역에 대한 실시예는, 현재 블록의 템플릿 영역이 상단 주변 영역과 좌측 주변 영역으로 구성된 경우를 가정한 것이다. 따라서, 상기 참조 템플릿 영역의 범위는, 현재 블록의 템플릿 영역의 범위에 대응되도록 결정될 수 있다.
일예로, 현재 블록의 템플릿 영역이 상단 주변 영역으로 구성된 경우, 현재 블록 내 상단 행에 속한 각 서브 블록의 움직임 벡터만을 이용하여 참조 서브 블록을 각각 결정할 수 있고, 각 참조 서브 블록의 서브 템플릿 영역을 기반으로 상기 참조 템플릿 영역을 구성할 수 있다. 또는, 현재 블록의 템플릿 영역이 좌측 주변 영역으로 구성된 경우, 현재 블록 내 좌측 열에 속한 각 서브 블록의 움직임 벡터만을 이용하여 참조 서브 블록을 각각 결정할 수 있고, 각 참조 서브 블록의 서브 템플릿 영역을 기반으로 상기 참조 템플릿 영역을 구성할 수 있다.
상기 템플릿 기반의 코스트를 기반으로 재배열을 수행함에 있어서, 추가적으로 다음의 요소가 고려될 수 있다.
어파인 후보의 타입 별로 재배열이 수행될 수 있다. 일예로, 상속된 후보와 구성된 후보를 대상으로 각각 재배열이 수행될 수 있다. 또는 평균 후보를 대상으로 재배열이 수행될 수 있다.
후보 리스트에 속한 복수의 어파인 후보를 서브 그룹으로 나누고, 서브 그룹 별로 재배열이 수행될 수 있다. 일예로, 재배열을 위한 서브 그룹의 크기가 5인 경우, 각 서브 그룹에 속한 5개의 어파인 후보를 대상으로 재배열이 수행될 수 있다.
어파인 후보의 타입 및/또는 서브 그룹의 크기를 고려하지 않고, 후보 리스트에 속한 어파인 후보 모두를 대상으로 하여 재배열이 수행될 수 있다.
재배열된 후보 리스트에 속한 모든 후보를 대상으로 하여 후보 인덱스가 시그날링될 수 있다. 여기서, 후보 인덱스는 재배열된 후보 리스트에 속한 모든 후보 중에서 어느 하나를 특정하는 정보일 수 있다.
또는, 재배열된 후보 리스트에 속한 후보의 총 개수와 별개로 소정의 개수의 후보를 대상으로 하여 후보 인덱스가 시그날링될 수 있다. 여기서, 후보 인덱스는 재배열된 후보 리스트에 속한 모든 후보가 아닌 일부 후보 중 어느 하나를 특정하는 정보일 수 있다. 일예로, 상기 일부 후보는 재배열된 하나 또는 그 이상의 서브 그룹에 속한 어파인 후보를 의미할 수 있다. 상기 일부 후보는, 후보 리스트의 복수의 서브 그룹 중에서 첫번째 서브 그룹에 속한 어파인 후보를 의미할 수 있다. 여기서, 첫번째 서브 그룹은, 후보 리스트의 인덱스 중 가장 작은 인덱스를 가진 어파인 후보가 속한 서브 그룹을 의미할 수 있다. 또는, 상기 소정의 개수는 인코딩 장치와 디코딩 장치에 동일하게 기-정의된 개수로서, 후보 리스트에 속한 후보의 총 개수보다 작을 수 있다. 만일 소정의 개수가 2개인 경우, 상기 일부 후보는 후보 리스트의 상위 2개의 후보를 의미할 수 있다. 이떄, 후보 인덱스는 상위 2개의 후보 중 어느 하나를 특정하는 정보일 수 있다. 다만, 소정의 개수는 2개로 제한되지 않으며, 3개, 4개, 5개 또는 그 이상으로 정의될 수 있다.
도 4를 참조하면, 상기 후보 리스트와 후보 인덱스에 기반하여, 현재 블록의 제어점 벡터가 유도될 수 있다(S410).
후보 인덱스는, 현재 블록의 제어점 벡터를 유도하기 위해 부호화된 인덱스를 의미할 수 있다. 상기 후보 인덱스는, 후보 리스트에 속한 복수의 후보들 중 어느 하나를 특정할 수 있다. 상기 후보 인덱스에 의해 어파인 후보가 특정되는 경우, 해당 어파인 후보의 제어점 벡터를 이용하여, 현재 블록의 제어점 벡터가 유도될 수 있다.
예를 들어, 현재 블록의 어파인 모델의 타입이 4-parameter인 경우(즉, 현재 블록이 2개의 제어점 벡터를 사용하는 것으로 결정된 경우)를 가정한다. 이때, 상기 후보 인덱스에 의해 특정된 어파인 후보가 3개의 제어점 벡터를 가진 경우, 상기 3개의 제어점 벡터 중 2개의 제어점 벡터(e.g., Idx=0, 1인 제어점 벡터, 또는 Idx=0, 2인 제어점 벡터)만을 선택하고, 이를 현재 블록의 제어점 벡터로 설정할 수 있다. 또는, 상기 특정된 어파인 후보의 3개의 제어점 벡터를 현재 블록의 제어점 벡터로 설정할 수 있다. 이 경우, 현재 블록의 어파인 모델의 타입은 6-parameter로 갱신될 수 있다.
반대로, 현재 블록의 어파인 모델의 타입이 6-parameter인 경우(즉, 현재 블록이 3개의 제어점 벡터를 사용하는 것으로 결정된 경우)를 가정한다. 이때, 상기 후보 인덱스에 의해 특정된 어파인 후보가 2개의 제어점 벡터를 가진 경우, 1개의 추가적인 제어점 벡터를 생성하고, 상기 어파인 후보의 2개의 제어점 벡터 및 추가적인 제어점 벡터를 현재 블록의 제어점 벡터로 설정할 수 있다. 상기 추가적인 제어점 벡터는, 어파인 후보의 2개의 제어점 벡터, 상기 2개의 제어점 벡터에 대응하는 블록 코너의 위치, 또는 현재/주변 블록의 크기/위치 정보 중 적어도 하나에 기초하여 유도될 수 있다. 또는, 상기 특정된 어파인 후보의 2개의 제어점 벡터를 현재 블록의 제어점 벡터로 설정할 수 있다. 이 경우, 현재 블록의 어파인 모델의 타입은 4-parameter로 갱신될 수 있다.
상기 제어점 벡터를 기반으로, 현재 블록의 움직임 벡터가 유도될 수 있다(S420).
상기 움직임 벡터는, 현재 블록의 서브 블록의 단위로 유도될 수 있다. 이를 위해 현재 블록은 복수의 NxM 서브 블록으로 분할될 수 있다. 여기서, NxM 서브 블록은, 직사각형(N>M 또는 N<M) 또는 정사각형(N=M)의 형태일 수 있다. 상기 N과 M 값은, 2, 4, 8, 16, 32 또는 그 이상일 수 있다. 서브 블록의 단위로 움직임 벡터를 유도하는 방법에 대해서는 도 7을 참조하여 자세히 살펴 보도록 한다.
도 4를 참조하면, 상기 유도된 움직임 벡터를 이용하여 현재 블록에 대해 인터 예측을 수행할 수 있다(S430).
구체적으로, 현재 블록의 움직임 벡터를 이용하여 참조 블록을 특정할 수 있다. 상기 참조 블록은, 현재 블록의 서브 블록 별로 각각 특정될 수 있다. 각 서브 블록의 참조 블록은, 하나의 참조 픽쳐에 속할 수 있다. 즉, 현재 블록에 속한 서브 블록은 하나의 참조 픽쳐를 공유할 수 있다. 또는, 현재 블록의 서브 블록 별로 참조 픽쳐 인덱스가 독립적으로 설정될 수도 있다. 상기 특정된 참조 블록은 현재 블록의 예측 블록으로 설정될 수 있다.
도 4를 참조하여 살펴본 S400 내지 S430 단계는 어파인 인터 모드에서도 동일/유사하게 수행될 수 있으며, 여기서 중복된 설명은 생략하기로 한다.
다만, 전술한 후보 리스트와 후보 인덱스를 기반으로 현재 블록의 제어점 벡터 예측값이 유도되고, 비트스트림으로부터 시그날링되는 제어점 벡터 차분값과 상기 제어점 벡터 예측값을 기반으로 현재 블록의 제어점 벡터가 유도될 수 있다. 이러한 제어점 벡터의 유도 방법이 어파인 머지 모드와 구별되는 특징이라 볼 수 있다.
여기서, 어파인 인터 모드를 위한 후보 리스트의 경우, 후보 리스트에 포함 가능한 어파인 후보의 최대 개수는 2개로 제한될 수 있다. 어파인 인터 모드를 위한 후보 인덱스는 0 또는 1의 값을 가지며, 후보 리스트에 속한 2개의 어파인 후보 중 어느 하나를 특정할 수 있다.
또는, 어파인 인터 모드를 위한 어파인 후보의 개수를 증가시키는 것을 고려할 수 있다. 전술한 바와 같이, 어파인 인터 모드를 위한 후보 리스트는 복수의 어파인 후보를 포함할 수 있고, 복수의 어파인 후보는 상속된 후보, 구성된 후보, 히스토리 기반의 후보, 또는 평균 후보 중 적어도 하나를 포함할 수 있다. 또한, 상기 후보 리스트는 하나 또는 그 이상의 병진 후보(translational candidate)를 더 포함할 수도 있다. 상기 병진 후보는, 어파인 후보와 달리, 하나의 움직임 벡터를 가진 후보를 의미할 수 있다. 상기 병진 후보는, 현재 블록의 공간적 또는 시간적 주변 블록으로부터 유도될 수 있고, 제로 움직임 벡터를 가진 후보를 더 포함할 수도 있다.
이와 같이 어파인 인터 모드를 위한 후보의 개수가 증가됨에 따라, 후보 리스트에 속한 모든 또는 일부 후보를 대상으로 후보 인덱스가 시그날링될 수 있다. 이때, 상기 모든 또는 일부 후보의 개수는 2개, 3개, 4개, 5개, 또는 그 이상일 수 있다. 일부 후보를 결정하는 방법은 앞서 살펴본 바와 같다.
전술한 어파인 모드를 위한 후보 리스트의 평균 후보는, 머지 모드에 적용될 수 있다. 머지 모드의 경우, 후보 인덱스에 기초하여 후보 리스트에 속한 복수의 머지 후보 중 어느 하나를 특정하고, 상기 특정된 머지 후보를 기반으로 현재 블록의 움직임 정보를 유도하며, 유도된 움직임 정보를 기반으로 인터 예측을 수행할 수 있다.
상기 복수의 머지 후보는 공간적 후보, 시간적 후보, 넌-인접 후보(non-adjacent candidate), 히스토리 기반의 후보, 평균 후보, 또는 제로 후보 중 적어도 하나를 포함할 수 있다.
상기 평균 후보는, 후보 리스트 내의 2개의 대상 후보를 기반으로 생성될 수 있으며, 2개의 대상 후보를 선택하는 방법은 표 1을 참조하여 살펴본 바와 같다. 또한, 상기 2개의 대상 후보는 후보의 타입을 고려하여 후보 리스트로부터 선택될 수 있다. 일예로, 상기 넌-인접 후보 또는 히스토리 기반의 후보 중 적어도 하나는 상기 평균 후보를 생성하기 위한 대상 후보에 제외될 수 있다. 다만, 이에 한정되지 아니하며, 전술한 어파인 모드에서의 대상 후보 선택 방법이 머지 모드에도 동일/유사하게 적용될 수 있음은 물론이다.
상기 평균 후보의 움직임 벡터는, 제1 대상 후보(Cand0)의 움직임 벡터와 제2 대상 후보(Cand1)의 움직임 벡터 간의 평균값으로 산출될 수 있다. 평균 후보의 참조 픽쳐 인덱스, 인터 예측 방향 정보, 양방향 가중 예측을 위한 가중치 정보, 또는 지역 조도 보상의 적용 여부를 나타내는 플래그 중 적어도 하나는, 상기 2개의 대상 후보 중 어느 하나로부터 상속될 수 있다.
또한, 평균 후보는, 기-정의된 특정 조건을 고려하여 후보 리스트에 적응적으로 추가/저장될 수 있다. 여기서, 특정 조건은, 대상 후보의 이용 가능 여부(조건 1), 2개의 대상 후보가 동일한 참조 픽쳐 인덱스를 가지는지 여부(조건 2), 또는 2개의 대상 후보가 동일한 참조 픽쳐를 가지는지 여부(조건 3) 중 적어도 하나를 포함할 수 있다.
(조건 1)과 관련하여, 2개의 대상 후보가 L0 예측 방향에 대해서 이용 가능한 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 움직임 벡터와 제2 대상 후보에 대한 L0 예측 방향의 움직임 벡터 간의 평균값으로 산출될 수 있다.
2개의 대상 후보 중 어느 하나의 대상 후보만이 L0 예측 방향에 대해서 이용 가능한 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 즉, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 움직임 벡터와 제2 대상 후보에 대한 L0 예측 방향의 움직임 벡터 간의 평균값으로 산출되지 않을 수 있다. 대신에, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, L0 예측 방향에 대해서 이용 가능한 대상 후보의 L0 예측 방향의 움직임 벡터로 설정/저장될 수 있다.
동일한 방법을 통해, 평균 후보에 대한 L1 예측 방향의 움직임 벡터를 유도할 수 있으며, 자세한 설명은 생략하기로 한다.
(조건 2)와 관련하여, 2개의 대상 후보에 대한 L0 예측 방향의 참조 픽쳐 인덱스가 서로 동일한 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 움직임 벡터와 제2 대상 후보에 대한 L0 예측 방향의 움직임 벡터 간의 평균값으로 산출될 수 있다.
2개의 대상 후보에 대한 L0 예측 방향의 참조 픽쳐 인덱스가 서로 상이한 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 즉, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 움직임 벡터와 제2 대상 후보에 대한 L0 예측 방향의 움직임 벡터 간의 평균값으로 산출되지 않을 수 있다. 대신에, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 2개의 대상 후보 중 어느 하나에 대한 L0 예측 방향의 움직임 벡터로 설정/저장될 수 있다. 여기서, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 후보 리스트의 인덱스가 더 작은 후보(즉, Cand0)를 의미할 수 있다. 또는, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 참조 픽쳐 인덱스가 더 작은 후보를 의미할 수도 있다.
동일한 방법을 통해, 평균 후보에 대한 L1 예측 방향의 움직임 벡터를 유도할 수 있으며, 자세한 설명은 생략하기로 한다.
(조건 3)과 관련하여, 제1 대상 후보에 대한 L0 예측 방향의 참조 픽쳐의 출력 순서 정보(POCref0[Cand0])와 제2 대상 후보에 대한 L0 예측 방향의 참조 픽쳐의 출력 순서 정보(POCref0[Cand1])의 차이가 0인 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성될 수 있다. 이때, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는 제1 대상 후보에 대한 L0 예측 방향의 움직임 벡터와 제2 대상 후보에 대한 L0 예측 방향의 움직임 벡터 간의 평균값으로 산출될 수 있다.
상기 POCref0[Cand0]와 POCref0[Cand1]의 차이가 0이 아닌 경우, 2개의 대상 후보를 기반으로 평균 후보가 생성되지 않을 수 있다. 즉, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 제1 대상 후보에 대한 L0 예측 방향의 움직임 벡터와 제2 대상 후보에 대한 L0 예측 방향의 움직임 벡터 간의 평균값으로 산출되지 않을 수 있다. 대신에, 평균 후보에 대한 L0 예측 방향의 움직임 벡터는, 2개의 대상 후보 중 어느 하나에 대한 L0 예측 방향의 움직임 벡터로 설정/저장될 수 있다. 여기서, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 후보 리스트의 인덱스가 더 작은 후보(즉, Cand0)를 의미할 수 있다. 또는, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 참조 픽쳐 인덱스가 더 작은 후보를 의미할 수도 있다. 또는, 2개의 대상 후보 중 어느 하나는, 2개의 대상 후보 중에서 현재 블록이 속한 현재 픽쳐의 출력 순서 정보와의 차이가 더 작은 후보를 의미할 수도 있다.
동일한 방법을 통해, 평균 후보에 대한 L1 예측 방향의 움직임 벡터를 유도할 수 있으며, 자세한 설명은 생략하기로 한다.
L0 및 L1 예측 방향에 대해서, 2개의 대상 후보를 기반으로 평균 후보가 모두 생성된 경우에 한하여, 해당 평균 후보가 후보 리스트에 추가되도록 제한될 수 있다. 또는, L0 및 L1 예측 방향 중 어느 하나에 대해서라도 전술한 조건이 만족되어 2개의 대상 후보를 기반으로 평균 후보가 생성된 경우라면, 해당 평균 후보가 후보 리스트에 추가될 수 있다. 즉, 평균 후보에 대한 L0 및 L1 예측 방향의 움직임 벡터 각각이, 2개의 대상 후보 중 어느 하나의 움직임 벡터로 설정/저장되는 경우, 해당 평균 후보는 후보 리스트에 추가되지 않을 수 있다.
머지 모드를 위한 후보 리스트를 구성함에 있어서, 상기 평균 후보는 후보 리스트에 기-포함된 머지 후보와의 중복 체크를 기반으로 적응적으로 추가될 수 있다.
일예로, 상기 중복 체크는, 상기 평균 후보의 움직임 벡터와 기-포함된 머지 후보의 움직임 벡터가 동일한지 여부에 기초하여 수행될 수 있다.
또는, 상기 중복 체크는, 상기 평균 후보의 움직임 벡터와 기-포함된 머지 후보의 움직임 벡터 간의 차이가 소정의 문턱값보다 작은지 여부에 기초하여 수행될 수 있다. 만일, 상기 평균 후보의 움직임 벡터와 기-포함된 머지 후보의 움직임 벡터 간의 차이가 소정의 문턱값보다 작은 경우, 해당 평균 후보는 후보 리스트에 추가되지 않을 수 있고, 그렇지 않은 경우, 해당 평균 후보는 후보 리스트에 추가될 수 있다.
또한, 어파인 모드를 위한 후보 리스트 구성 방법에 대한 실시예들은 머지 모드를 위한 후보 리스트를 구성하기 위해 동일/유사하게 적용될 수 있으며, 자세한 설명은 생략하기로 한다.
도 5는 본 개시에 따른 일실시예로서, 공간적/시간적 주변 블록의 제어점 벡터로부터 어파인 후보를 유도하는 방법에 관한 것이다.
현재 블록(500)의 너비와 높이는 각각 cbW, cbH이고, 현재 블록의 위치는 (xCb, yCb)이다. 공간적 주변 블록(510-550)의 너비와 높이는 각각 nbW, nbH이고, 공간적 주변 블록의 위치는 (xNb, yNb)이다. 공간적 주변 블록은, 현재 블록의 좌측 블록(510), 좌하단 블록(540), 우상단 블록(430), 상단 블록(420) 또는 좌상단 블록(450) 중 적어도 하나를 포함할 수 있다. 또는, 상기 공간적 주변 블록은, 상기 좌상단 블록(450)의 우측으로 인접한 블록 또는 하단으로 인접한 블록 중 적어도 하나를 더 포함할 수도 있다.
공간적 후보는, N개의 제어점 벡터(cpMV)를 가질 수 있다. 여기서, N은, 1, 2, 3, 또는 그 이상의 정수일 수 있다. N 값은 서브 블록 단위로 복호화되는지 여부에 관한 정보, 어파인 모드로 부호화된 블록인지 여부에 관한 정보 또는 어파인 모델의 타입(4-parameter 또는 6-parameter)에 관한 정보 중 적어도 하나에 기초하여 결정될 수 있다.
예를 들어, 상기 정보에 따라, 해당 블록이 서브 블록 단위로 복호화되거나, 어파인 모드로 부호화된 블록인 경우, 해당 블록은 2개의 제어점 벡터를 가질 수 있다. 반면, 그렇지 않은 경우, 해당 블록은 어파인 모델 기반의 예측을 수행하지 않을 수 있다.
또는, 상기 정보에 따라, 해당 블록이 어파인 모드로 부호화된 블록이고, 어파인 모델의 타입이 6-parameter인 경우, 해당 블록은 3개의 제어점 벡터를 가질 수 있다. 반면, 그렇지 않은 경우, 해당 블록은 어파인 모델 기반의 예측을 수행하지 않을 수 있다.
전술한 정보는, 인코딩 장치에서 부호화되어 시그날링될 수도 있다. 또는, 상기 정보 전부 또는 일부는, 블록의 속성에 기초하여 디코딩 장치에서 유도될 수도 있다. 여기서, 블록은, 현재 블록을 의미할 수도 있고, 현재 블록의 공간적/시간적 주변 블록을 의미할 수도 있다. 상기 속성은, 크기, 형태, 위치, 분할 타입, 인터 모드, 잔차 계수에 관한 파라미터 등을 의미할 수 있다. 상기 인터 모드는, 디코딩 장치에 기-정의된 모드로서, 머지 모드, 스킵 모드, AMVP 모드, 어파인 모드, 인트라/인터 조합 모드, 현재 픽쳐 참조 모드 등을 의미할 수 있다. 또는, N 값은 전술한 블록의 속성에 기초하여 디코딩 장치에서 유도될 수도 있다.
본 실시예에서, N개의 제어점 벡터는, 제1 제어점 벡터(cpMV[0]), 제2 제어점 벡터(cpMV[1]), 제3 제어점 벡터(cpMV[2]), ... 제N 제어점 벡터(cpMV[N-1])로 표현될 수 있다.
예를 들어, 제1 제어점 벡터(cpMV[0]), 제2 제어점 벡터(cpMV[1]), 제3 제어점 벡터(cpMV[2]) 및 제4 제어점 벡터(cpMV[3])는, 블록의 좌상단 코너, 우상단 코너, 좌하단 코너 및 우하단 코너의 위치에 각각 대응하는 벡터일 수 있다. 여기서, 공간적 후보는, 3개의 제어점 벡터를 가지는 것을 가정하며, 3개의 제어점 벡터는 제1 개 내지 제N 제어점 벡터 중에서 선택된 임의의 제어점 벡터일 수 있다. 다만, 이에 한정되지 아니하며, 공간적 후보는 2개의 제어점 벡터를 가질 수 있고, 2개의 제어점 벡터는 제1 내지 제N 제어점 벡터 중에서 선택된 임의의 제어점 벡터일 수 있다.
이하, 공간적 후보의 제어점 벡터를 유도하는 방법에 대해서 살펴 보도록 한다.
1. 현재 블록의 경계가 CTU boundary에 접하지 않는 경우
상기 제1 제어점 벡터는, 공간적 주변 블록의 제1 제어점 벡터, 소정의 차분값, 현재 블록의 위치 정보 (xCb, yCb) 또는 공간적 주변 블록의 위치 정보 (xNb, yNb) 중 적어도 하나에 기초하여 유도될 수 있다.
상기 차분값의 개수는, 1개, 2개, 3개 또는 그 이상일 수 있다. 상기 차분값의 개수는, 전술한 블록의 속성을 고려하여 가변적으로 결정될 수도 있고, 인코딩/디코딩 장치에 동일하게 기-정의된 값일 수도 있다. 상기 차분값은, 복수의 제어점 벡터 중 어느 하나와 다른 하나 간의 차분값으로 정의될 수 있다.
예를 들어, 상기 차분값은, 제2 제어점 벡터와 제1 제어점 벡터 간의 제1 차분값, 제3 제어점 벡터와 제1 제어점 벡터 간의 제2 차분값, 제4 제어점 벡터와 제3 제어점 벡터 간의 제3 차분값, 또는 제4 제어점 벡터와 제2 제어점 벡터 간의 제4 차분값 중 적어도 하나를 포함할 수 있다.
구체적으로, 제1 제어점 벡터는 다음 수학식 1과 같이 유도될 수 있다.
[수학식 1]
cpMvLX[ 0 ][ 0 ] = ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb - yNb ) )
cpMvLX[ 0 ][ 1 ] = ( mvScaleVer + dVerX * ( xCb - xNb ) + dVerY * ( yCb - yNb ) )
수학식 1에서, 변수 mvScaleHor와 mvScaleVer는 공간적 주변 블록의 제1 제어점 벡터를 의미할 수도 있고, 제1 제어점 벡터에 k만큼 쉬프트 연산을 적용하여 유도된 값을 의미할 수도 있다. 여기서, k는 1, 2, 3, 4, 5, 6, 7, 8, 9 또는 그 이상의 정수일 수 있다. 변수 dHorX와 dVerX는, 제2 제어점 벡터와 제1 제어점 벡터 간의 제1 차분값의 x-성분과 y-성분에 각각 대응된다. 변수 dHorY와 dVerY는, 제3 제어점 벡터와 제1 제어점 벡터 간의 제2 차분값의 x-성분과 y-성분에 각각 대응된다. 전술한 변수는, 다음 수학식 2와 같이 유도될 수 있다.
[수학식 2]
mvScaleHor = CpMvLX[ xNb ][ yNb ][ 0 ][ 0 ] << 7
mvScaleVer = CpMvLX[ xNb ][ yNb ][ 0 ][ 1 ] << 7
dHorX = ( CpMvLX[ xNb + nNbW - 1 ][ yNb ][ 1 ][ 0 ] - CpMvLX[ xNb ][ yNb ][ 0 ][ 0 ] ) << ( 7 - log2NbW )
dVerX = ( CpMvLX[ xNb + nNbW - 1 ][ yNb ][ 1 ][ 1 ] - CpMvLX[ xNb ][ yNb ][ 0 ][ 1 ] ) << ( 7 - log2NbW )
dHorY = ( CpMvLX[ xNb ][ yNb + nNbH - 1 ][ 2 ][ 0 ] - CpMvLX[ xNb ][ yNb ][ 2 ][ 0 ] ) << ( 7 - log2NbH )
dVerY = ( CpMvLX[ xNb ][ yNb + nNbH - 1 ][ 2 ][ 1 ] - CpMvLX[ xNb ][ yNb ][ 2 ][ 1 ] ) << ( 7 - log2NbH )
제2 제어점 벡터는, 공간적 주변 블록의 제1 제어점 벡터, 소정의 차분값, 현재 블록의 위치 정보 (xCb, yCb), 블록 크기(너비 또는 높이) 또는 공간적 주변 블록의 위치 정보 (xNb, yNb) 중 적어도 하나에 기초하여 유도될 수 있다. 여기서, 블록 크기는, 현재 블록 및/또는 공간적 주변 블록의 크기를 의미할 수 있다. 상기 차분값은, 제1 제어점 벡터에서 살펴본 바와 같은바, 여기서 자세한 설명은 생략하기로 한다. 다만, 제2 제어점 벡터의 유도 과정에서 이용되는 차분값의 범위 및/또는 개수는 제1 제어점 벡터와 상이할 수 있다.
구체적으로, 제2 제어점 벡터는 다음 수학식 3과 같이 유도될 수 있다.
[수학식 3]
cpMvLX[ 1 ][ 0 ] = ( mvScaleHor + dHorX * ( xCb + cbWidth - xNb ) + dHorY * ( yCb - yNb ) )
cpMvLX[ 1 ][ 1 ] = ( mvScaleVer + dVerX * ( xCb + cbWidth - xNb ) + dVerY * ( yCb - yNb ) )
수학식 3에서, 변수 mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, dVerY는, 수학식 1에서 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다.
제3 제어점 벡터는, 공간적 주변 블록의 제1 제어점 벡터, 소정의 차분값, 현재 블록의 위치 정보 (xCb, yCb), 블록 크기(너비 또는 높이) 또는 공간적 주변 블록의 위치 정보 (xNb, yNb) 중 적어도 하나에 기초하여 유도될 수 있다. 여기서, 블록 크기는, 현재 블록 및/또는 공간적 주변 블록의 크기를 의미할 수 있다. 상기 차분값은, 제1 제어점 벡터에서 살펴본 바와 같은바, 여기서 자세한 설명은 생략하기로 한다. 다만, 제3 제어점 벡터의 유도 과정에서 이용되는 차분값의 범위 및/또는 개수는 제1 제어점 벡터 또는 제2 제어점 벡터와 상이할 수 있다.
구체적으로, 제3 제어점 벡터는 다음 수학식 4와 같이 유도될 수 있다.
[수학식 4]
cpMvLX[ 2 ][ 0 ] = ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb + cbHeight - yNb ) )
cpMvLX[ 2 ][ 1 ] = ( mvScaleVer + dVerX * ( xCb - xNb ) + dVerY * ( yCb + cbHeight - yNb ) )
수학식 4에서, 변수 mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, dVerY는, 수학식 1에서 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다. 한편, 전술한 과정을 통해, 공간적 후보의 제1 내지 제N 제어점 벡터가 유도될 수 있다.
2. 현재 블록의 경계가 CTU boundary에 접하는 경우
상기 제1 제어점 벡터는, 공간적 주변 블록의 움직임 벡터(MV), 소정의 차분값, 현재 블록의 위치 정보 (xCb, yCb) 또는 공간적 주변 블록의 위치 정보 (xNb, yNb) 중 적어도 하나에 기초하여 유도될 수 있다.
상기 움직임 벡터는, 공간적 주변 블록의 최하단에 위치한 서브 블록의 움직임 벡터일 수 있다. 상기 서브 블록은, 공간적 주변 블록의 최하단에 위치한 복수의 서브 블록 중 최좌측, 중앙 또는 최우측에 위치한 것일 수 있다. 또는, 상기 움직임 벡터는, 서브 블록의 움직임 벡터의 평균값, 최대값 또는 최소값을 의미할 수도 있다.
상기 차분값의 개수는, 1개, 2개, 3개 또는 그 이상일 수 있다. 상기 차분값의 개수는, 전술한 블록의 속성을 고려하여 가변적으로 결정될 수도 있고, 인코딩/디코딩 장치에 동일하게 기-정의된 값일 수도 있다. 상기 차분값은, 공간적 주변 블록에서 서브 블록의 단위로 저장된 복수의 움직임 벡터 중 어느 하나와 다른 하나 간의 차분값으로 정의될 수 있다. 예를 들어, 상기 차분값은, 공간적 주변 블록의 우하단 서브 블록의 움직임 벡터와 좌하단 서브 블록의 움직임 벡터 간의 차분값을 의미할 수 있다.
예를 들어, 제1 제어점 벡터는 다음 수학식 5와 같이 유도될 수 있다.
[수학식 5]
cpMvLX[ 0 ][ 0 ] = ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb - yNb ) )
cpMvLX[ 0 ][ 1 ] = ( mvScaleVer + dVerX * ( xCb - xNb ) + dVerY * ( yCb - yNb ) )
수학식 5에서, 변수 mvScaleHor와 mvScaleVer는 전술한 공간적 주변 블록의 움직임 벡터(MV) 또는 상기 움직임 벡터에 k만큼 쉬프트 연산을 적용하여 유도된 값을 의미할 수도 있다. 여기서, k는 1, 2, 3, 4, 5, 6, 7, 8, 9 또는 그 이상의 정수일 수 있다.
상기 변수 dHorX와 dVerX는, 소정의 차분값의 x-성분과 y-성분에 각각 대응된다. 여기서, 차분값은, 공간적 주변 블록 내 우하단 서브 블록의 움직임 벡터와 좌하단 서브 블록의 움직임 벡터 간의 차분값을 의미한다. 변수 dHorY와 dVerY는, 상기 변수 dHorX와 dVerX에 기초하여 유도될 수 있다. 전술한 변수는, 다음 수학식 6과 같이 유도될 수 있다.
[수학식 6]
mvScaleHor = MvLX[ xNb ][ yNb + nNbH - 1 ][ 0 ] << 7
mvScaleVer = MvLX[ xNb ][ yNb + nNbH - 1 ][ 1 ] << 7
dHorX = ( MvLX[ xNb + nNbW - 1 ][ yNb + nNbH - 1 ][ 0 ] - MvLX[ xNb ][ yNb + nNbH - 1 ][ 0 ] ) << ( 7 - log2NbW )
dVerX = ( MvLX[ xNb + nNbW - 1 ][ yNb + nNbH - 1 ][ 1 ] - MvLX[ xNb ][ yNb + nNbH - 1 ][ 1 ] ) << ( 7 - log2NbW )
dHorY = - dVerX
dVerY = dHorX
제2 제어점 벡터는, 공간적 주변 블록의 움직임 벡터(MV), 소정의 차분값, 현재 블록의 위치 정보 (xCb, yCb), 블록 크기(너비 또는 높이) 또는 공간적 주변 블록의 위치 정보 (xNb, yNb) 중 적어도 하나에 기초하여 유도될 수 있다. 여기서, 블록 크기는, 현재 블록 및/또는 공간적 주변 블록의 크기를 의미할 수 있다. 상기 움직임 벡터와 차분값에 대해서는, 제1 제어점 벡터에서 살펴본 바와 같은바, 여기서 자세한 설명은 생략하기로 한다. 다만, 제2 제어점 벡터의 유도 과정에서 이용되는 움직임 벡터의 위치, 차분값의 범위 및/또는 개수는 제1 제어점 벡터와 상이할 수 있다.
예를 들어, 제2 제어점 벡터는 다음 수학식 7과 같이 유도될 수 있다.
[수학식 7]
cpMvLX[ 1 ][ 0 ] = ( mvScaleHor + dHorX * ( xCb + cbWidth - xNb ) + dHorY * ( yCb - yNb ) )
cpMvLX[ 1 ][ 1 ] = ( mvScaleVer + dVerX * ( xCb + cbWidth - xNb ) + dVerY * ( yCb - yNb ) )
수학식 7에서, 변수 mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, dVerY는, 수학식 5에서 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다.
제3 제어점 벡터는, 공간적 주변 블록의 움직임 벡터(MV), 소정의 차분값, 현재 블록의 위치 정보 (xCb, yCb), 블록 크기(너비 또는 높이) 또는 공간적 주변 블록의 위치 정보 (xNb, yNb) 중 적어도 하나에 기초하여 유도될 수 있다. 여기서, 블록 크기는, 현재 블록 및/또는 공간적 주변 블록의 크기를 의미할 수 있다. 상기 움직임 벡터와 차분값은, 제1 제어점 벡터에서 살펴본 바와 같은바, 여기서 자세한 설명은 생략하기로 한다. 다만, 제3 제어점 벡터의 유도 과정에서 이용되는 움직임 벡터의 위치, 차분값의 범위 및/또는 개수는 제1 제어점 벡터 또는 제2 제어점 벡터와 상이할 수 있다.
예를 들어, 제3 제어점 벡터는 다음 수학식 8과 같이 유도될 수 있다.
[수학식 8]
cpMvLX[ 2 ][ 0 ] = ( mvScaleHor + dHorX * ( xCb - xNb ) + dHorY * ( yCb + cbHeight - yNb ) )
cpMvLX[ 2 ][ 1 ] = ( mvScaleVer + dVerX * ( xCb - xNb ) + dVerY * ( yCb + cbHeight - yNb ) )
수학식 8에서, 변수 mvScaleHor, mvScaleVer, dHorX, dVerX, dHorY, dVerY는, 수학식 5에서 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다. 한편, 전술한 과정을 통해, 공간적 후보의 제1 내지 제N 제어점 벡터가 유도될 수 있다.
전술한 어파인 후보의 유도 과정은, 기-정의된 공간적 주변 블록 각각에 대해서 수행될 수 있다. 기-정의된 공간적 주변 블록은, 현재 블록의 좌측 블록, 좌하단 블록, 우상단 블록, 상단 블록 또는 좌상단 블록 중 적어도 하나를 포함할 수 있다.
또는, 상기 어파인 후보의 유도 과정은, 상기 공간적 주변 블록의 그룹 별로 각각 수행될 수도 있다. 여기서, 공간적 주변 블록은, 좌측 블록 및 좌하단 블록을 포함한 제1 그룹과 우상단 블록, 상단 블록 및 좌상단 블록을 포함한 제2 그룹으로 분류될 수 있다.
예를 들어, 제1 그룹에 속한 공간적 주변 블록으로부터 하나의 어파인 후보가 유도될 수 있다. 상기 유도는, 소정의 우선 순위에 기초하여, 가용한 어파인 후보가 발견될 때까지 수행될 수 있다. 상기 우선 순위는, 좌측 블록 -> 좌하단 블록의 순서일 수도 있고, 그 역순일 수도 있다. 상기 우선 순위에 따라 제1 그룹 내의 해당 공간적 주변 블록이 어파인 모델 기반의 예측을 통해 복호화된 블록인지를 판단하고, 가장 처음 어파인 모델 기반의 예측으로 복호화된 블록이 어파인 후보로 선택될 수 있다.
마찬가지로, 제2 그룹에 속한 공간적 주변 블록으로부터 하나의 어파인 후보가 유도될 수 있다. 상기 유도는, 소정의 우선 순위에 기초하여, 가용한 어파인 후보가 발견될 때까지 수행될 수 있다. 상기 우선 순위는, 우상단 블록 -> 상단 블록 -> 좌상단 블록의 순서일 수도 있고, 그 역순일 수도 있다. 상기 우선 순위에 따라 제2 그룹 내의 해당 공간적 주변 블록이 어파인 모델 기반의 예측을 통해 복호화된 블록인지를 판단하고, 가장 처음 어파인 모델 기반의 예측으로 복호화된 블록이 어파인 후보로 선택될 수 있다.
전술한 실시예는, 시간적 주변 블록에 대해서도 동일/유사하게 적용될 수 있다. 여기서, 시간적 주변 블록은, 현재 블록과 다른 픽쳐에 속하나, 현재 블록과 동일 위치의 블록일 수 있다. 동일 위치의 블록이라 함은, 현재 블록의 좌상단 샘플의 위치, 중앙 위치 또는 현재 블록의 우하단 샘플에 인접한 샘플의 위치를 포함하는 블록일 수 있다.
또는, 시간적 주변 블록은, 상기 동일 위치의 블록에서 소정의 변이 벡터만큼 쉬프트된 위치의 블록을 의미할 수 있다. 여기서, 변이 벡터는, 전술한 현재 블록의 공간적 주변 블록 중 어느 하나의 움직임 벡터에 기초하여 결정될 수 있다.
도 6은 본 개시에 따른 일실시예로서, 공간적/시간적 주변 블록의 움직임 벡터의 조합에 기초하여 구성된 후보를 유도하는 방법을 도시한 것이다.
본 개시의 구성된 후보는, 현재 블록의 각 코너 위치에 대응하는 제어점 벡터(이하, 제어점 벡터(cpMVCorner[n])라 함) 중 적어도 2개의 조합에 기초하여 유도될 수 있다. 여기서, n은 0, 1, 2, 3일 수 있다.
상기 제어점 벡터(cpMVCorner[n])는, 공간적 주변 블록 및/또는 시간적 주변 블록의 움직임 벡터에 기초하여 유도될 수 있다. 여기서, 공간적 주변 블록은, 현재 블록의 좌상단 샘플에 인접한 제1 주변 블록(A, B 또는 C), 현재 블록의 우상단 샘플에 인접한 제2 주변 블록(D 또는 E) 또는 현재 블록의 좌하단 샘플에 인접한 제3 주변 블록(F 또는 G) 중 적어도 하나를 포함할 수 있다. 시간적 주변 블록은, 현재 블록과 다른 픽쳐에 속하는 블록으로서, 현재 블록과 동일 위치의 블록(이하, 제4 주변 블록(Col)이라 함)을 의미할 수 있다. 여기서, 제4 주변 블록은, 현재 블록의 좌상단 샘플, 우상단 샘플 또는 좌하단 샘플의 위치를 포함하는 블록(H, I 또는 J)이거나, 현재 블록의 우하단 샘플의 위치에 인접한 블록을 의미할 수 있다.
제1 주변 블록은, 현재 블록의 좌상단(A), 상단(B) 또는 좌측(C)의 주변 블록을 의미할 수 있다. 소정의 우선 순위에 따라 주변 블록 A, B, C의 움직임 벡터가 가용한지 여부를 판단하고, 가용한 주변 블록의 움직임 벡터를 이용하여 제어점 벡터를 결정할 수 있다. 상기 가용성 판단은, 가용한 움직임 벡터를 가진 주변 블록이 발견될 때까지 수행될 수 있다. 여기서, 우선 순위는, A -> B -> C의 순일 수도 있다. 다만, 이에 한정되지 아니하며, A -> C -> B, C -> A -> B, B -> A -> C의 순일 수도 있다.
제2 주변 블록은, 현재 블록의 상단(D) 또는 우상단(E) 주변 블록을 의미할 수 있다. 마찬가지로, 소정의 우선 순위에 따라 주변 블록 D, E의 움직임 벡터가 가용한지 여부를 판단하고, 가용한 주변 블록의 움직임 벡터를 이용하여 제어점 벡터를 결정할 수 있다. 상기 가용성 판단은, 가용한 움직임 벡터를 가진 주변 블록이 발견될 때까지 수행될 수 있다. 여기서, 우선 순위는, D -> E의 순일 수도 있고, E -> D의 순일 수도 있다.
제3 주변 블록은, 현재 블록의 좌측(F) 또는 좌하단(G)의 주변 블록을 의미할 수 있다. 마찬가지로, 소정의 우선 순위에 따라 주변 블록의 움직임 벡터가 가용한지 여부를 판단하고, 가용한 주변 블록의 움직임 벡터를 이용하여 제어점 벡터를 결정할 수 있다. 상기 가용성 판단은, 가용한 움직임 벡터를 가진 주변 블록이 발견될 때까지 수행될 수 있다. 여기서, 우선 순위는, G -> F의 순일 수도 있고, F -> G의 순일 수도 있다.
예를 들어, 제1 제어점 벡터(cpMVCorner[0])는 제1 주변 블록의 움직임 벡터로 설정될 수 있고, 제2 제어점 벡터(cpMVCorner[1])는 제2 주변 블록의 움직임 벡터로 설정될 수 있고, 제3 제어점 벡터(cpMVCorner[2])는 제3 주변 블록의 움직임 벡터로 설정될 수 있다. 제4 제어점 벡터(cpMVCorner[3])는 제4 주변 블록의 움직임 벡터로 설정될 수 있다.
또는, 제1 제어점 벡터는, 제1 주변 블록 또는 제4 주변 블록 중 적어도 하나의 움직임 벡터를 이용하여 유도될 수 있고, 여기서, 제4 주변 블록은 좌상단 샘플의 위치를 포함하는 블록(H)일 수 있다. 제2 제어점 벡터는, 제2 주변 블록 또는 제4 주변 블록 중 적어도 하나의 움직임 벡터를 이용하여 유도될 수 있다. 여기서, 제4 주변 블록은 우상단 샘플의 위치를 포함하는 블록(I)일 수 있다. 제3 제어점 벡터는, 제3 주변 블록 또는 제4 주변 블록 중 적어도 하나의 움직임 벡터를 이용하여 유도될 수 있다. 여기서, 제4 주변 블록은 좌하단 샘플의 위치를 포함하는 블록(J)일 수 있다.
또는, 상기 제1 내지 제4 제어점 벡터 중 어느 하나는 다른 하나에 기초하여 유도될 수 있다. 예를 들어, 제2 제어점 벡터는, 제1 제어점 벡터에 소정의 오프셋 벡터를 적용하여 유도될 수 있다. 오프셋 벡터는, 제3 제어점 벡터와 제1 제어점 벡터 간의 차분 벡터이거나 상기 차분 벡터에 소정의 스케일링 팩터를 적용하여 유도될 수 있다. 스케일링 팩터는, 현재 블록 및/또는 주변 블록의 너비 또는 높이 중 적어도 하나에 기초하여 결정될 수 있다.
전술한 제1 내지 제4 제어점 벡터 중 적어도 2개의 조합을 통해, 본 개시에 따른 K개의 구성된 후보(ConstK)가 결정될 수 있다. K 값은, 1, 2, 3, 4, 5, 6, 7 또는 그 이상의 정수일 수 있다. K 값은, 인코딩 장치에서 시그날링되는 정보에 기초하여 유도될 수도 있고, 인코딩/디코딩 장치에 동일하게 기-정의된 값일 수도 있다. 상기 정보는, 후보 리스트에 포함되는 구성된 후보의 최대 개수를 지시하는 정보를 포함할 수 있다.
구체적으로, 제1 구성된 후보(Const1)는, 제1 내지 제3 제어점 벡터를 조합하여 유도될 수 있다. 예를 들어, 제1 구성된 후보(Const1)는 다음 표 9와 같은 제어점 벡터를 가질 수 있다. 한편, 제1 주변 블록의 참조 픽쳐 정보가 제2 및 제3 주변 블록의 참조 픽쳐 정보와 동일한 경우에 한하여, 표 9과 같이 제어점 벡터가 구성되도록 제한될 수도 있다. 여기서, 참조 픽쳐 정보는, 참조 픽쳐 리스트 내에서 해당 참조 픽쳐의 위치를 나타내는 참조 픽쳐 인덱스를 의미할 수도 있고, 출력 순서를 나타내는 POC(picture order count) 값을 의미할 수도 있다.
Idx 제어점 벡터
0 cpMvCorner[ 0 ]
1 cpMvCorner[ 1 ]
2 cpMvCorner[ 2 ]
제2 구성된 후보(Const2)는, 제1, 제2 및 제4 제어점 벡터를 조합하여 유도될 수 있다. 예를 들어, 제2 구성된 후보(Const2)는 다음 표 10과 같은 제어점 벡터를 가질 수 있다. 한편, 제1 주변 블록의 참조 픽쳐 정보가 제2 및 제4 주변 블록의 참조 픽쳐 정보와 동일한 경우에 한하여, 표 10과 같이 제어점 벡터가 구성되도록 제한될 수도 있다. 여기서, 참조 픽쳐 정보는 전술한 바와 같다.
Idx 제어점 벡터
0 cpMvCorner[ 0 ]
1 cpMvCorner[ 1 ]
2 cpMvCorner[ 3 ] + cpMvCorner[ 1 ] - cpMvCorner[ 0 ]
cpMvCorner[ 3 ] + cpMvCorner[ 0 ] - cpMvCorner[ 1 ]
제3 구성된 후보(Const3)는, 제1, 제3 및 제4 제어점 벡터를 조합하여 유도될 수 있다. 예를 들어, 제3 구성된 후보(Const3)는 다음 표 11과 같은 제어점 벡터를 가질 수 있다. 한편, 제1 주변 블록의 참조 픽쳐 정보가 제3 및 제4 주변 블록의 참조 픽쳐 정보와 동일한 경우에 한하여, 표 11과 같이 제어점 벡터가 구성되도록 제한될 수도 있다. 여기서, 참조 픽쳐 정보는 전술한 바와 같다.
Idx 제어점 벡터 제어점 벡터
0 cpMvCorner[0] cpMvCorner[0]
1 cpMvCorner[3] + cpMvCorner[0] - cpMvCorner[2] cpMvCorner[2]
2 cpMvCorner[2] cpMvCorner[0] + cpMvCorner[3] - cpMvCorner[2]
제4 구성된 후보(Const4)는, 제2, 제3 및 제4 제어점 벡터를 조합하여 유도될 수 있다. 예를 들어, 제4 구성된 후보(Const4)는 다음 표 12와 같은 제어점 벡터를 가질 수 있다. 한편, 제2 주변 블록의 참조 픽쳐 정보가 제3 및 제4 주변 블록의 참조 픽쳐 정보와 동일한 경우에 한하여, 표 12와 같이 구성되도록 제한될 수도 있다. 여기서, 참조 픽쳐 정보는 전술한 바와 같다.
Idx 제어점 벡터 제어점 벡터
0 cpMvCorner[1] + cpMvCorner[2] - cpMvCorner[3] cpMvCorner[2]
1 cpMvCorner[1] cpMvCorner[3]
2 cpMvCorner[2] cpMvCorner[3] + cpMvCorner[2] - cpMvCorner[1]
제5 구성된 후보(Const5)는, 제1 및 제2 제어점 벡터를 조합하여 유도될 수 있다. 예를 들어, 제5 구성된 후보(Const5)는 다음 표 13과 같은 제어점 벡터를 가질 수 있다. 한편, 제1 주변 블록의 참조 픽쳐 정보가 제2 주변 블록의 참조 픽쳐 정보와 동일한 경우에 한하여, 표 13과 같이 제어점 벡터가 구성되도록 제한될 수도 있다. 여기서, 참조 픽쳐 정보는 전술한 바와 같다.
Idx 제어점 벡터
1 cpMvCorner[ 0 ]
2 cpMvCorner[ 1 ]
제6 구성된 후보(Const6)는, 제1 및 제3 제어점 벡터를 조합하여 유도될 수 있다. 예를 들어, 제6 구성된 후보(Const6)는 다음 표 14와 같은 제어점 벡터를 가질 수 있다. 한편, 제1 주변 블록의 참조 픽쳐 정보가 제3 주변 블록의 참조 픽쳐 정보와 동일한 경우에 한하여, 표 14와 같이 제어점 벡터가 구성되도록 제한될 수도 있다. 여기서, 참조 픽쳐 정보는 전술한 바와 같다.
Idx 제어점 벡터 제어점 벡터
1 cpMvCorner[ 0 ] cpMvCorner[ 0 ]
2 cpMvCorner[ 2 ] cpMvCorner[ 1 ]
표 14에서, cpMvCorner[ 1 ]는, 상기 제1 및 제3 제어점 벡터를 기반으로 유도된 제2 제어점 벡터일 수 있다. 제2 제어점 벡터는, 제1 제어점 벡터, 소정의 차분값 또는 현재/주변 블록의 크기 중 적어도 하나에 기초하여 유도될 수 있다. 예를 들어, 제2 제어점 벡터는, 다음 수학식 9와 같이 유도될 수 있다.
[수학식 9]
cpMvCorner[ 1 ][ 0 ] = ( cpMvCorner[ 0 ][ 0 ] << 7 ) + ( ( cpMvCorner[ 2 ][ 1 ] - cpMvCorner[ 0 ][ 1 ] ) << ( 7 + Log2( cbHeight / cbWidth ) ) )
cpMvCorner[ 1 ][ 1 ] = ( cpMvCorner[ 0 ][ 1 ] << 7 ) + ( ( cpMvCorner[ 2 ][ 0 ] - cpMvCorner[ 0 ][ 0 ] ) << ( 7 + Log2( cbHeight / cbWidth ) ) )
전술한 K개의 구성된 후보(ConstK)에서, K 값은 후보 리스트에 배열되는 구성된 후보의 위치 또는 우선순위를 한정하는 것은 아니다.
또한, 상기 제1 내지 제6 구성된 후보 모두 상기 후보 리스트에 포함될 수도 있고, 일부만이 후보 리스트에 포함될 수도 있다.
예를 들어, 현재 블록이 3개의 제어점 벡터를 이용하는 것으로 결정된 경우, 제1 내지 제4 제어점 벡터 중 3개의 조합을 통해 생성되는 구성된 후보만이 이용될 수 있다. 현재 블록이 2개의 제어점 벡터를 이용하는 것으로 결정된 경우, 제1 내지 제4 제어점 벡터 중 적어도 2개의 조합을 통해 생성되는 구성된 후보가 이용되거나, 제1 내지 제 4 제어점 벡터 중 2개의 조합만을 통해 생성되는 구성된 후보가 이용될 수도 있다.
또는, 후보 리스트에 포함되는 어파인 후보의 최대 개수를 고려하여 일부 구성된 후보만이 후보 리스트에 포함될 수도 있다. 여기서, 최대 개수는, 인코딩 장치에서 시그날링되는 최대 개수 정보에 기초하여 결정되거나, 전술한 현재 블록의 속성을 고려하여 가변적으로 결정될 수도 있다. 이때, 구성된 후보(ConstK)의 K 값은 후보 리스트에 삽입되는 우선순위를 의미할 수도 있다.
한편, 상기 구성된 후보에 속한 2개의 제어점 벡터 간의 차이값이 소정의 문턱값보다 작은 경우, 해당 구성된 후보는 후보 리스트에 삽입되지 않을 수 있다. 상기 2개의 제어점 벡터 간의 차이값은, 수평 방향의 차이값과 수직 방향의 차이값으로 구분될 수 있다. 여기서, 수평 방향의 차이값은 제1 제어점 벡터(610)과 제2 제어점 벡터(620) 간의 차이값을 의미하고, 수직 방향의 차이값은 제1 제어점 벡터(610)와 제3 제어점 벡터(630) 간의 차이값을 의미할 수 있다. 상기 문턱값은, 0 또는 인코딩/디코딩 장치에 동일하게 기-정의된 크기의 벡터를 의미할 수 있다.
도 7은 본 개시의 일실시예로서, 현재 블록의 서브 블록 단위로 움직임 벡터를 유도하는 방법을 도시한 것이다.
전술한 바와 같이, 현재 블록의 제어점 벡터를 이용하여, 현재 블록의 움직임 벡터가 유도될 수 있다. 이때, 현재 블록은 복수의 서브 블록으로 분할되고, 서브 블록의 단위로 움직임 벡터가 유도될 수 있다.
상기 서브 블록은, 인코딩/디코딩 장치에 동일하게 기-정의된 크기/형태를 가질 수 있다. 예를 들어, 서브 블록의 크기/형태는, 4x4, 8x8, 16x16 등과 같은 정방형이거나, 2x4, 2x8, 4x8, 4x16 등과 같은 비정방형일 수 있다. 또는, 상기 서브 블록은 너비와 높이의 합이 8, 12, 16 또는 그 이상의 정수인 블록으로 정의될 수도 있다. 또는, 상기 서브 블록은 너비와 높이의 곱이 16, 32, 64 또는 그 이상의 정수인 블록으로 정의될 수도 있다.
또는, 서브 블록의 크기/형태는, 전술한 블록의 속성에 기초하여 가변적으로 유도될 수도 있다. 예를 들어, 현재 블록의 크기가 소정의 문턱크기보다 크거나 같은 경우, 현재 블록은 제1 서브 블록의 단위(e.g., 8x8, 16x16)로 분할되고, 그렇지 않은 경우, 현재 블록은 제2 서브 블록의 단위(e.g., 4x4)로 분할될 수 있다.
또는, 서브 블록의 크기/형태에 관한 정보가 인코딩 장치에서 부호화되어 디코딩 장치로 시그날링될 수도 있다. 상기 정보는, 서브 블록의 크기 또는 형태 중 적어도 하나를 나타내며, 이는 시퀀스, 픽쳐, 슬라이스, 타일, 코딩 트리 유닛(CTU), 또는 코딩 유닛 중 적어도 하나의 레벨에서 시그날링될 수 있다.
도 7에 도시된 바와 같이, 현재 블록(700)을 대표하는 제어점은, 좌상단 코너 위치(711)와 우상단 코너 위치(721)를 포함할 수 있다. 다만, 이에 한정되지 아니하며, 상기 제어점은, 좌상단 코너 위치(711), 우상단 코너 위치(721) 및 좌하단 코너 위치(731)의 세 지점을 포함하거나, 추가적인 다수의 지점을 더 포함할 수 있다.
2개의 제어점(711, 721)이 이용되는 경우, 제1 제어점(711)에 대응하는 제1 제어점 벡터(710), 제2 제어점(721)에 대응하는 제2 제어점 벡터(720), 서브 블록의 위치(x, y) 또는 현재 블록의 크기(w 또는 h) 중 적어도 하나를 이용하여, 현재 블록의 각 서브 블록에 대한 움직임 벡터가 유도될 수 있다. 예를 들어, 각 서브 블록의 움직임 벡터는 다음 수학식 10과 같이 유도될 수 있다.
[수학식 10]
Figure PCTKR2022021532-appb-img-000001
한편, 현재 블록은 보정된 제1 제어점 벡터 및 제3 제어점 벡터를 가지는 경우, 현재 블록 내 각 서브 블록에 대한 움직임 벡터는, 보정된 제1 제어점 벡터, 제3 제어점 벡터, 해당 서브 블록의 위치, 또는 현재 블록의 높이 중 적어도 하나를 이용하여 유도될 수 있다.
3개의 제어점이 이용되는 경우, 제1 제어점(711)에 대응하는 제1 제어점 벡터(710), 제2 제어점(721)에 대응하는 제2 제어점 벡터(720), 제3 제어점(731)에 대응하는 제3 제어점 벡터(730), 서브 블록의 위치(x, y) 또는 현재 블록의 크기(w 또는 h) 중 적어도 하나를 이용하여, 현재 블록의 각 서브 블록에 대한 움직임 벡터가 유도될 수 있다. 이때, 제2 제어점 벡터와 제1 제어점 벡터 간의 차분벡터가 이용될 수 있고, 제3 제어점 벡터와 제1 제어점 벡터 간의 차분벡터가 이용될 수 있다. 상기 차분벡터는, 수평 방향(x축 방향)과 수직 방향(y축 방향) 각각에 대해서 산출될 수 있다.
도 8은 본 개시에 따른 일실시예로서, 어파인 모델 기반의 인터 예측을 수행하는 인터 예측부(332)의 개략적인 구성을 도시한 것이다.
도 4를 참조하여 디코딩 장치에서 수행되는 어파인 모델 기반의 인터 예측 방법에 대해서 살펴보았으며, 이는 디코딩 장치의 인터 예측부(332)에서 동일하게 수행될 수 있다. 이에, 여기서 중복된 설명은 생략하기로 한다.
도 8을 참조하면, 인터 예측부(332)는, 제어점 벡터 유도부(800), 움직임 벡터 유도부(810) 및 예측 샘플 생성부(820)를 포함할 수 있다.
상기 제어점 벡터 유도부(800)는, 현재 블록의 움직임 정보를 예측하기 위한 후보 리스트를 생성할 수 있다. 상기 후보 리스트는, 어파인 후보 또는 제로 후보 중 적어도 하나를 포함하며, 상기 어파인 후보는 상속된 후보, 구성된 후보, 히스토리 기반의 후보, 평균 후보, 또는 조합된 후보 중 적어도 하나를 포함할 수 있다. 상기 후보 리스트를 구성하고, 이를 재배열하는 방법에 대해서는 도 4를 참조하여 자세히 살펴본 바와 같으며, 자세한 설명은 생략하기로 한다. 상기 제어점 벡터 유도부(800)는, 상기 후보 리스트와 후보 인덱스에 기초하여, 현재 블록의 제어점 벡터를 유도할 수 있다.
상기 움직임 벡터 유도부(810)는, 유도된 제어점 벡터를 기반으로, 현재 블록의 움직임 벡터를 유도할 수 있다. 상기 움직임 벡터는 현재 블록의 서브 블록의 단위로 유도될 수 있으며, 이를 위해 현재 블록은 복수의 서브 블록으로 분할될 수 있다. 현재 블록의 서브 블록 단위로 움직임 벡터를 유도하는 방법에 대해서는 도 7을 참조하여 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다.
상기 예측 샘플 생성부(820)는, 상기 유도된 움직임 벡터를 이용하여 현재 블록에 대한 인터 예측을 수행할 수 있다.
또는, 머지 모드의 경우, 인터 예측부(332)는, 움직임 정보 유도부와 예측 샘플 생성부를 포함할 수 있다.
상기 움직임 정보 유도부는, 현재 블록의 머지 모드를 위한 후보 리스트를 생성하고, 필요한 경우 이를 재배열할 수 있다. 이에 대해서는 도 4를 참조하여 자세히 살펴본 바와 같으며, 자세한 설명은 생략하기로 한다. 상기 움직임 정보 유도부는, 상기 후보 리스트와 후보 인덱스에 기초하여, 현재 블록의 움직임 정보를 유도할 수 있다.
상기 예측 샘플 생성부는, 상기 유도된 움직임 정보를 이용하여 현재 블록에 대한 인터 예측을 수행할 수 있다.
도 9는 본 개시에 따른 일실시예로서, 인코딩 장치에 의해 수행되는 어파인 모델 기반의 인터 예측 방법을 도시한 것이다.
이하, 도 4를 참조하여 디코딩 장치에서 수행되는 어파인 모델 기반의 인터 예측 방법을 살펴보았으며, 이는 인코딩 장치에서 의해 수행되는 어파인 모델 기반의 인터 예측 방법에도 동일/유사하게 적용될 수 있다. 이에, 여기서 중복된 설명은 생략하기로 한다.
도 9를 참조하면, 현재 블록의 움직임 정보를 예측하기 위한 후보 리스트를 생성할 수 있다(S900).
상기 후보 리스트는, 어파인 후보 또는 제로 후보 중 적어도 하나를 포함하며, 상기 어파인 후보는 상속된 후보, 구성된 후보, 히스토리 기반의 후보, 평균 후보, 또는 조합된 후보 중 적어도 하나를 포함할 수 있다. 상기 후보 리스트를 구성하고, 이를 재배열하는 방법에 대해서는 도 4를 참조하여 자세히 살펴본 바와 같으며, 자세한 설명은 생략하기로 한다.
도 9를 참조하면, 상기 후보 리스트에 기초하여 상기 현재 블록의 제어점 벡터를 결정할 수 있다(S910).
상기 후보 리스트에 속한 복수의 후보 중에서 최적의 후보를 결정할 수 있고, 상기 결정된 최적의 후보를 기반으로 현재 블록의 제어점 벡터를 결정할 수 있다. 이때, 상기 현재 블록의 제어점 벡터를 결정하기 위해 이용된 최적의 후보를 특정하는 후보 인덱스가 부호화될 수 있다.
도 9를 참조하면, 상기 유도된 제어점 벡터를 기반으로, 현재 블록의 움직임 벡터가 유도될 수 있다(S920).
상기 움직임 벡터는 현재 블록의 서브 블록의 단위로 유도될 수 있으며, 이를 위해 현재 블록은 복수의 서브 블록으로 분할될 수 있다. 현재 블록의 서브 블록 단위로 움직임 벡터를 유도하는 방법에 대해서는 도 7을 참조하여 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다.
도 9를 참조하면, 상기 유도된 움직임 벡터를 이용하여 현재 블록에 대해 인터 예측을 수행할 수 있다(S930).
상기 움직임 벡터에 기초하여, 현재 블록의 서브 블록 각각에 대응하는 참조 서브 블록이 특정될 수 있고, 상기 특정된 참조 서브 블록을 기반으로 현재 블록 내 해당 서브 블록의 예측 블록을 획득할 수 있다. 상기 현재 블록의 예측 블록을 기반으로 상기 현재 블록을 부호화하여, 비트스트림을 생성할 수 있다.
또는, 머지 모드의 경우, 후보 리스트에 속한 복수의 머지 후보 중 어느 하나를 기반으로 현재 블록의 움직임 정보를 결정하고, 결정된 움직임 정보를 기반으로 인터 예측을 수행할 수 있다. 상기 복수의 머지 후보 중 어느 하나를 특정하는 후보 인덱스가 부호화될 수 있다.
도 10은 본 개시에 따른 일실시예로서, 어파인 모델 기반의 인터 예측을 수행하는 인터 예측부(221)의 개략적인 구성을 도시한 것이다.
도 9를 참조하여 인코딩 장치에서 수행되는 어파인 모델 기반의 인터 예측 방법에 대해서 살펴보았으며, 이는 인코딩 장치의 인터 예측부(221)에서 동일하게 수행될 수 있다. 이에, 여기서 중복된 설명은 생략하기로 한다.
도 10을 참조하면, 인터 예측부(221)는, 제어점 벡터 결정부(1000), 움직임 벡터 유도부(1010) 및 예측 샘플 생성부(1020)를 포함할 수 있다.
상기 제어점 벡터 결정부(1000)는, 현재 블록의 움직임 정보를 예측하기 위한 후보 리스트를 생성할 수 있다. 상기 후보 리스트는, 어파인 후보 또는 제로 후보 중 적어도 하나를 포함하며, 상기 어파인 후보는 상속된 후보, 구성된 후보, 히스토리 기반의 후보, 평균 후보, 또는 조합된 후보 중 적어도 하나를 포함할 수 있다. 상기 후보 리스트를 구성하고, 이를 재배열하는 방법에 대해서는 도 4를 참조하여 자세히 살펴본 바와 같으며, 자세한 설명은 생략하기로 한다.
상기 제어점 벡터 결정부(1000)는, 상기 후보 리스트에 속한 복수의 후보 중에서 최적의 후보를 결정할 수 있고, 상기 결정된 최적의 후보를 기반으로 현재 블록의 제어점 벡터를 결정할 수 있다. 이때, 엔트로피 인코딩부(240)는 상기 현재 블록의 제어점 벡터를 결정하기 위해 이용된 최적의 후보를 특정하는 후보 인덱스를 부호화하여, 이를 비트스트림에 삽입할 수 있다.
상기 움직임 벡터 유도부(1010)는, 유도된 제어점 벡터를 기반으로, 현재 블록의 움직임 벡터를 유도할 수 있다. 상기 움직임 벡터는 현재 블록의 서브 블록의 단위로 유도될 수 있으며, 이를 위해 현재 블록은 복수의 서브 블록으로 분할될 수 있다. 현재 블록의 서브 블록 단위로 움직임 벡터를 유도하는 방법에 대해서는 도 7을 참조하여 살펴본 바와 같으며, 여기서 자세한 설명은 생략하기로 한다.
상기 예측 샘플 생성부(1020)는, 상기 유도된 움직임 벡터를 이용하여 현재 블록에 대한 인터 예측을 수행할 수 있다.
또는, 머지 모드의 경우, 인터 예측부(221)는, 움직임 정보 결정부와 예측 샘플 생성부를 포함할 수 있다.
상기 움직임 정보 결정부는, 후보 리스트에 속한 복수의 머지 후보 중 어느 하나를 기반으로 현재 블록의 움직임 정보를 결정하고, 결정된 움직임 정보를 기반으로 인터 예측을 수행할 수 있다. 이때, 엔트로피 인코딩부(240)는 상기 현재 블록의 움직임 벡터를 결정하기 위해 이용된 최적의 머지 후보를 특정하는 후보 인덱스를 부호화하여, 이를 비트스트림에 삽입할 수 있다.
상기 예측 샘플 생성부는, 상기 결정된 움직임 정보를 이용하여 현재 블록에 대한 인터 예측을 수행할 수 있다.
상술한 실시예에서, 방법들은 일련의 단계 또는 블록으로써 순서도를 기초로 설명되고 있지만, 해당 실시예는 단계들의 순서에 한정되는 것은 아니며, 어떤 단계는 상술한 바와 다른 단계와 다른 순서로 또는 동시에 발생할 수 있다. 또한, 당업자라면 순서도에 나타내어진 단계들이 배타적이지 않고, 다른 단계가 포함되거나 순서도의 하나 또는 그 이상의 단계가 본 문서의 실시예들의 범위에 영향을 미치지 않고 삭제될 수 있음을 이해할 수 있을 것이다.
상술한 본 문서의 실시예들에 따른 방법은 소프트웨어 형태로 구현될 수 있으며, 본 문서에 따른 인코딩 장치 및/또는 디코딩 장치는 예를 들어 TV, 컴퓨터, 스마트폰, 셋톱박스, 디스플레이 장치 등의 영상 처리를 수행하는 장치에 포함될 수 있다.
본 문서에서 실시예들이 소프트웨어로 구현될 때, 상술한 방법은 상술한 기능을 수행하는 모듈(과정, 기능 등)로 구현될 수 있다. 모듈은 메모리에 저장되고, 프로세서에 의해 실행될 수 있다. 메모리는 프로세서 내부 또는 외부에 있을 수 있고, 잘 알려진 다양한 수단으로 프로세서와 연결될 수 있다. 프로세서는 ASIC(application-specific integrated circuit), 다른 칩셋, 논리 회로 및/또는 데이터 처리 장치를 포함할 수 있다. 메모리는 ROM(read-only memory), RAM(random access memory), 플래쉬 메모리, 메모리 카드, 저장 매체 및/또는 다른 저장 장치를 포함할 수 있다. 즉, 본 문서에서 설명한 실시예들은 프로세서, 마이크로 프로세서, 컨트롤러 또는 칩 상에서 구현되어 수행될 수 있다. 예를 들어, 각 도면에서 도시한 기능 유닛들은 컴퓨터, 프로세서, 마이크로 프로세서, 컨트롤러 또는 칩 상에서 구현되어 수행될 수 있다. 이 경우 구현을 위한 정보(ex. information on instructions) 또는 알고리즘이 디지털 저장 매체에 저장될 수 있다.
또한, 본 명세서의 실시예(들)이 적용되는 디코딩 장치 및 인코딩 장치는 멀티미디어 방송 송수신 장치, 모바일 통신 단말, 홈 시네마 비디오 장치, 디지털 시네마 비디오 장치, 감시용 카메라, 비디오 대화 장치, 비디오 통신과 같은 실시간 통신 장치, 모바일 스트리밍 장치, 저장 매체, 캠코더, 주문형 비디오(VoD) 서비스 제공 장치, OTT 비디오(Over the top video) 장치, 인터넷 스트리밍 서비스 제공 장치, 3차원(3D) 비디오 장치, VR(virtual reality) 장치, AR(argumente reality) 장치, 화상 전화 비디오 장치, 운송 수단 단말 (ex. 차량(자율주행차량 포함) 단말, 비행기 단말, 선박 단말 등) 및 의료용 비디오 장치 등에 포함될 수 있으며, 비디오 신호 또는 데이터 신호를 처리하기 위해 사용될 수 있다. 예를 들어, OTT 비디오(Over the top video) 장치로는 게임 콘솔, 블루레이 플레이어, 인터넷 접속 TV, 홈시어터 시스템, 스마트폰, 태블릿 PC, DVR(Digital Video Recorder) 등을 포함할 수 있다.
또한, 본 명세서의 실시예(들)이 적용되는 처리 방법은 컴퓨터로 실행되는 프로그램의 형태로 생산될 수 있으며, 컴퓨터가 판독할 수 있는 기록 매체에 저장될 수 있다. 본 명세서의 실시예(들)에 따른 데이터 구조를 가지는 멀티미디어 데이터도 또한 컴퓨터가 판독할 수 있는 기록 매체에 저장될 수 있다. 상기 컴퓨터가 판독할 수 있는 기록 매체는 컴퓨터로 읽을 수 있는 데이터가 저장되는 모든 종류의 저장 장치 및 분산 저장 장치를 포함한다. 상기 컴퓨터가 판독할 수 있는 기록 매체는, 예를 들어, 블루레이 디스크(BD), 범용 직렬 버스(USB), ROM, PROM, EPROM, EEPROM, RAM, CD-ROM, 자기 테이프, 플로피 디스크 및 광학적 데이터 저장 장치를 포함할 수 있다. 또한, 상기 컴퓨터가 판독할 수 있는 기록 매체는 반송파(예를 들어, 인터넷을 통한 전송)의 형태로 구현된 미디어를 포함한다. 또한, 인코딩 방법으로 생성된 비트스트림이 컴퓨터가 판독할 수 있는 기록 매체에 저장되거나 유무선 통신 네트워크를 통해 전송될 수 있다.
또한, 본 명세서의 실시예(들)는 프로그램 코드에 의한 컴퓨터 프로그램 제품으로 구현될 수 있고, 상기 프로그램 코드는 본 명세서의 실시예(들)에 의해 컴퓨터에서 수행될 수 있다. 상기 프로그램 코드는 컴퓨터에 의해 판독가능한 캐리어 상에 저장될 수 있다.
도 11은 본 개시의 실시예들이 적용될 수 있는 컨텐츠 스트리밍 시스템의 예를 나타낸다.
도 11을 참조하면, 본 명세서의 실시예(들)이 적용되는 컨텐츠 스트리밍 시스템은 크게 인코딩 서버, 스트리밍 서버, 웹 서버, 미디어 저장소, 사용자 장치 및 멀티미디어 입력 장치를 포함할 수 있다.
상기 인코딩 서버는 스마트폰, 카메라, 캠코더 등과 같은 멀티미디어 입력 장치들로부터 입력된 컨텐츠를 디지털 데이터로 압축하여 비트스트림을 생성하고 이를 상기 스트리밍 서버로 전송하는 역할을 한다. 다른 예로, 스마트폰, 카메라, 캠코더 등과 같은 멀티미디어 입력 장치들이 비트스트림을 직접 생성하는 경우, 상기 인코딩 서버는 생략될 수 있다.
상기 비트스트림은 본 명세서의 실시예(들)이 적용되는 인코딩 방법 또는 비트스트림 생성 방법에 의해 생성될 수 있고, 상기 스트리밍 서버는 상기 비트스트림을 전송 또는 수신하는 과정에서 일시적으로 상기 비트스트림을 저장할 수 있다.
상기 스트리밍 서버는 웹 서버를 통한 사용자 요청에 기초하여 멀티미디어 데이터를 사용자 장치에 전송하고, 상기 웹 서버는 사용자에게 어떠한 서비스가 있는지를 알려주는 매개체 역할을 한다. 사용자가 상기 웹 서버에 원하는 서비스를 요청하면, 상기 웹 서버는 이를 스트리밍 서버에 전달하고, 상기 스트리밍 서버는 사용자에게 멀티미디어 데이터를 전송한다. 이때, 상기 컨텐츠 스트리밍 시스템은 별도의 제어 서버를 포함할 수 있고, 이 경우 상기 제어 서버는 상기 컨텐츠 스트리밍 시스템 내 각 장치 간 명령/응답을 제어하는 역할을 한다.
상기 스트리밍 서버는 미디어 저장소 및/또는 인코딩 서버로부터 컨텐츠를 수신할 수 있다. 예를 들어, 상기 인코딩 서버로부터 컨텐츠를 수신하게 되는 경우, 상기 컨텐츠를 실시간으로 수신할 수 있다. 이 경우, 원활한 스트리밍 서비스를 제공하기 위하여 상기 스트리밍 서버는 상기 비트스트림을 일정 시간동안 저장할 수 있다.
상기 사용자 장치의 예로는, 휴대폰, 스마트 폰(smart phone), 노트북 컴퓨터(laptop computer), 디지털방송용 단말기, PDA(personal digital assistants), PMP(portable multimedia player), 네비게이션, 슬레이트 PC(slate PC), 태블릿 PC(tablet PC), 울트라북(ultrabook), 웨어러블 디바이스(wearable device, 예를 들어, 워치형 단말기 (smartwatch), 글래스형 단말기 (smart glass), HMD(head mounted display)), 디지털 TV, 데스크탑 컴퓨터, 디지털 사이니지 등이 있을 수 있다.
상기 컨텐츠 스트리밍 시스템 내 각 서버들은 분산 서버로 운영될 수 있으며, 이 경우 각 서버에서 수신하는 데이터는 분산 처리될 수 있다.
본 명세서에 기재된 청구항들은 다양한 방식으로 조합될 수 있다. 예를 들어, 본 명세서의 방법 청구항의 기술적 특징이 조합되어 장치로 구현될 수 있고, 본 명세서의 장치 청구항의 기술적 특징이 조합되어 방법으로 구현될 수 있다. 또한, 본 명세서의 방법 청구항의 기술적 특징과 장치 청구항의 기술적 특징이 조합되어 장치로 구현될 수 있고, 본 명세서의 방법 청구항의 기술적 특징과 장치 청구항의 기술적 특징이 조합되어 방법으로 구현될 수 있다.

Claims (13)

  1. 현재 블록의 후보 리스트를 생성하는 단계; 여기서, 상기 후보 리스트는 복수의 어파인 후보를 포함하고, 상기 복수의 어파인 후보는 상속된 후보, 구성된 후보, 또는 히스토리 기반의 후보 중 적어도 하나를 포함함,
    상기 후보 리스트에 평균 후보를 추가하는 단계; 여기서, 상기 평균 후보는, 상기 후보 리스트로부터 소정의 우선순서에 따라 선택된 2개의 어파인 후보에 기초하여 생성됨, 및
    상기 평균 후보를 포함한 상기 후보 리스트에 기초하여 상기 현재 블록의 인터 예측을 수행하는 단계를 포함하는, 영상 디코딩 방법.
  2. 제1항에 있어서,
    상기 선택된 2개의 어파인 후보가 서로 다른 어파인 모델의 타입을 가지는 경우, 상기 평균 후보에 대한 어파인 모델의 타입은, 상기 선택된 2개의 어파인 후보 중 어느 하나에 대한 어파인 모델의 타입으로 결정되는, 영상 디코딩 방법.
  3. 제1항에 있어서, 상기 후보 리스트를 생성하는 단계는,
    상기 후보 리스트에 속한 복수의 어파인 후보 중 적어도 하나에 대한 어파인 모델의 타입을 고려하여, 상기 후보 리스트에 속한 복수의 어파인 후보를 재배열하는 단계를 포함하는, 영상 디코딩 방법.
  4. 제1항에 있어서,
    상기 평균 후보는, 상기 2개의 어파인 후보의 이용 가능성, 상기 2개의 어파인 후보 간 참조 픽쳐 인덱스의 동일 여부, 또는 상기 2개의 어파인 후 보 간 참조 픽쳐의 동일 여부 중 적어도 하나에 기초하여, 상기 후보 리스트에 적응적으로 추가되는, 영상 디코딩 방법.
  5. 제1항에 있어서,
    상기 상속된 후보는, 상기 현재 블록의 공간적 주변 블록이 어파인 모드로 부호화된 블록인지 여부, 상기 후보 리스트에 기-포함된 상속된 후보의 개수, 또는 상기 현재 블록의 형태 중 적어도 하나를 고려하여, 상기 후보 리스트에 구성되는, 영상 디코딩 방법.
  6. 제1항에 있어서,
    상기 평균 후보는, 상기 후보 리스트에 속한 복수의 어파인 후보 중 적어도 하나와의 중복 체크에 기초하여 상기 후보 리스트에 추가되고,
    상기 중복 체크는, 상기 평균 후보에 대한 어파인 모델의 타입에 기초하여 수행되는, 영상 디코딩 방법.
  7. 제1항에 있어서,
    템플릿 기반의 코스트에 기초하여, 상기 후보 리스트에 포함된 상기 상속된 후보, 상기 구성된 후보, 또는 상기 평균 후보 중 적어도 하나에 대해서 재배열이 수행되는, 영상 디코딩 방법.
  8. 제1항에 있어서,
    상기 평균 후보의 움직임 정보는, 상기 선택된 2개의 어파인 후보 중 적어도 하나의 움직임 정보에 기초하여 결정되고,
    상기 움직임 정보는, 지역 조도 보상이 적용되는지 여부를 나타내는 플래그 또는 양방향 가중 예측을 위한 가중치 정보 중 적어도 하나를 포함하는, 영상 디코딩 방법.
  9. 제1항에 있어서,
    비트스트림으로부터 상기 현재 블록의 후보 인덱스를 획득하는 단계를 더 포함하고,
    상기 후보 인덱스는, 상기 평균 후보를 포함한 상기 후보 리스트 내에서 일부 어파인 후보 중 어느 하나를 특정하는, 영상 디코딩 방법.
  10. 제1항에 있어서,
    상기 후보 리스트는, 조합된 후보를 더 포함하고,
    상기 조합된 후보는, 상기 복수의 어파인 후보 중 어느 하나에 대한 L0 예측 방향의 움직임 정보와 상기 복수의 어파인 후보 중 다른 하나에 대한 L1 예측 방향의 움직임 정보를 조합하여 생성되는, 영상 디코딩 방법.
  11. 현재 블록의 후보 리스트를 생성하는 단계; 여기서, 상기 후보 리스트는 복수의 어파인 후보를 포함하고, 상기 복수의 어파인 후보는 상속된 후보, 구성된 후보, 또는 히스토리 기반의 후보 중 적어도 하나를 포함함,
    상기 후보 리스트에 평균 후보를 추가하는 단계; 여기서, 상기 평균 후보는, 상기 후보 리스트로부터 소정의 우선순서에 따라 선택된 2개의 어파인 후보에 기초하여 생성됨, 및
    상기 평균 후보를 포함한 상기 후보 리스트에 기초하여 상기 현재 블록의 인터 예측을 수행하는 단계를 포함하는, 영상 인코딩 방법.
  12. 제11항에 따른 영상 인코딩 방법에 의해 생성된 비트스트림을 저장하는 컴퓨터 판독 가능한 저장 매체.
  13. 현재 블록의 후보 리스트를 생성하는 단계; 여기서, 상기 후보 리스트는 복수의 어파인 후보를 포함하고, 상기 복수의 어파인 후보는 상속된 후보, 구성된 후보, 또는 히스토리 기반의 후보 중 적어도 하나를 포함함,
    상기 후보 리스트에 평균 후보를 추가하는 단계; 여기서, 상기 평균 후보는, 상기 후보 리스트로부터 소정의 우선순서에 따라 선택된 2개의 어파인 후보에 기초하여 생성됨,
    상기 평균 후보를 포함한 상기 후보 리스트에 기초하여 상기 현재 블록의 인터 예측을 수행하는 단계;
    상기 인터 예측을 통해 생성된 상기 현재 블록의 예측 블록을 기반으로 상기 현재 블록을 부호화하여 비트스트림을 생성하는 단계; 및
    상기 비트스트림을 포함한 데이터를 전송하는 단계를 포함하는, 영상 정보에 대한 데이터 전송 방법.
PCT/KR2022/021532 2022-01-04 2022-12-28 영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체 WO2023132556A1 (ko)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US202263296419P 2022-01-04 2022-01-04
US63/296,419 2022-01-04

Publications (1)

Publication Number Publication Date
WO2023132556A1 true WO2023132556A1 (ko) 2023-07-13

Family

ID=87073843

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2022/021532 WO2023132556A1 (ko) 2022-01-04 2022-12-28 영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체

Country Status (1)

Country Link
WO (1) WO2023132556A1 (ko)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116996695A (zh) * 2023-09-27 2023-11-03 深圳大学 一种全景图像压缩方法、装置、设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200055124A (ko) * 2017-10-05 2020-05-20 인터디지털 브이씨 홀딩스 인코포레이티드 비디오 인코딩 및 비디오 디코딩을 위한 방법들 및 장치들
WO2020112451A1 (en) * 2018-11-27 2020-06-04 Interdigital Vc Holdings, Inc. Combining affine candidates
KR20200066737A (ko) * 2017-11-30 2020-06-10 엘지전자 주식회사 영상 코딩 시스템에서 인터 예측에 따른 영상 디코딩 방법 및 장치
US20200296383A1 (en) * 2019-03-15 2020-09-17 Tencent America LLC Methods of accessing affine history-based motion vector predictor buffer
WO2021043166A1 (en) * 2019-09-02 2021-03-11 Beijing Bytedance Network Technology Co., Ltd. Improvement of merge candidates

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200055124A (ko) * 2017-10-05 2020-05-20 인터디지털 브이씨 홀딩스 인코포레이티드 비디오 인코딩 및 비디오 디코딩을 위한 방법들 및 장치들
KR20200066737A (ko) * 2017-11-30 2020-06-10 엘지전자 주식회사 영상 코딩 시스템에서 인터 예측에 따른 영상 디코딩 방법 및 장치
WO2020112451A1 (en) * 2018-11-27 2020-06-04 Interdigital Vc Holdings, Inc. Combining affine candidates
US20200296383A1 (en) * 2019-03-15 2020-09-17 Tencent America LLC Methods of accessing affine history-based motion vector predictor buffer
WO2021043166A1 (en) * 2019-09-02 2021-03-11 Beijing Bytedance Network Technology Co., Ltd. Improvement of merge candidates

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116996695A (zh) * 2023-09-27 2023-11-03 深圳大学 一种全景图像压缩方法、装置、设备及介质
CN116996695B (zh) * 2023-09-27 2024-04-05 深圳大学 一种全景图像压缩方法、装置、设备及介质

Similar Documents

Publication Publication Date Title
WO2020171444A1 (ko) Dmvr 기반의 인터 예측 방법 및 장치
WO2020184848A1 (ko) Dmvr 기반의 인터 예측 방법 및 장치
WO2020189893A1 (ko) Bdof 기반의 인터 예측 방법 및 장치
WO2020055161A1 (ko) 영상 코딩 시스템에서 서브 블록 단위의 움직임 예측에 기반한 영상 디코딩 방법 및 장치
WO2020055107A1 (ko) 영상 코딩 시스템에서 어파인 mvp 후보 리스트를 사용하는 어파인 움직임 예측에 기반한 영상 디코딩 방법 및 장치
WO2021096290A1 (ko) 변환에 기반한 영상 코딩 방법 및 그 장치
WO2020171647A1 (ko) 영상 코딩 시스템에서 인트라 예측을 사용하는 영상 디코딩 방법 및 그 장치
WO2020197152A1 (ko) 영상 코딩 시스템에서 인트라 예측 모드 후보 리스트를 사용하는 영상 디코딩 방법 및 그 장치
WO2021172881A1 (ko) 인터 예측에 기반한 영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체
WO2021172914A1 (ko) 레지듀얼 코딩에 대한 영상 디코딩 방법 및 그 장치
WO2020256400A1 (ko) 루마 맵핑 및 크로마 스케일링 기반 비디오 또는 영상 코딩
WO2020197223A1 (ko) 영상 코딩 시스템에서의 인트라 예측 기반 영상 코딩
WO2020204418A1 (ko) 비디오 또는 영상 코딩 방법 및 그 장치
WO2023132556A1 (ko) 영상 인코딩/디코딩 방법 및 장치, 그리고 비트스트림을 저장한 기록 매체
WO2021060834A1 (ko) 서브픽처 기반 영상 부호화/복호화 방법, 장치 및 비트스트림을 전송하는 방법
WO2021086061A1 (ko) 변환에 기반한 영상 코딩 방법 및 그 장치
WO2021006698A1 (ko) 영상 코딩 시스템에서 영상 코딩 방법 및 장치
WO2021006697A1 (ko) 레지듀얼 코딩에 대한 영상 디코딩 방법 및 그 장치
WO2020251268A1 (ko) 크로마 성분에 대한 영상 디코딩 방법 및 그 장치
WO2020197155A1 (ko) 영상 코딩 시스템에서 영상 디코딩 방법 및 그 장치, 영상 인코딩 방법 및 그 장치
WO2021172916A1 (ko) 영상 디코딩 방법 및 그 장치
WO2021172907A1 (ko) 영상 디코딩 방법 및 그 장치
WO2021137556A1 (ko) 변환에 기반한 영상 코딩 방법 및 그 장치
WO2020197224A1 (ko) 인트라 예측 모드 후보 리스트를 구성하는 영상 코딩
WO2021086056A1 (ko) 변환에 기반한 영상 코딩 방법 및 그 장치

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22919093

Country of ref document: EP

Kind code of ref document: A1