WO2017069591A1

WO2017069591A1 - Method and device for filtering image in image coding system

Info

Publication number: WO2017069591A1
Application number: PCT/KR2016/011949
Authority: WO
Inventors: 남정학; 손은용; 박내리; 허진
Original assignee: 엘지전자 주식회사
Priority date: 2015-10-23
Filing date: 2016-10-24
Publication date: 2017-04-27

Abstract

A filtering method performed by a decoding device, according to the present invention, comprises the steps of: deriving a motion vector of a block to be predicted; deriving a prediction sample for the block to be predicted, on the basis of the motion vector; deriving a block to be filtered, on the basis of neighboring blocks of the block to be predicted; deriving an area to be filtered within the block to be filtered; and filtering a restoration sample within the area to be filtered, on the basis of the prediction sample of the block to be predicted. According to the present invention, filtering can be performed through a weighted sum of a prediction sample, derived on the basis of a motion vector of a block to be predicted, and neighboring restoration samples at the boundary of the block to be predicted, thereby enabling errors between blocks to be reduced and subjective/objective image quality to be improved.

Description

Method and apparatus for image filtering in image coding system

The present invention relates to image coding technology, and more particularly, to an image filtering method and apparatus in an image coding system.

Recently, the demand for high resolution and high quality images is increasing in various applications. As an image has a high resolution and high quality, the amount of information on the image also increases.

As information volume increases, devices with various performances and networks with various environments are emerging. As devices with different capabilities and networks with different environments emerge, the same content is available in different qualities.

In detail, as the video quality of the terminal device can be supported and the network environment is diversified, in general, video of general quality may be used in one environment, but higher quality video may be used in another environment. .

For example, a consumer who purchases video content on a mobile terminal can view the same video content on a larger screen and at a higher resolution through a large display in the home.

In recent years, as broadcasts with full high definition (FHD) resolutions are being serviced, many users are already accustomed to high-definition and high-definition video, and service providers and users are paying attention to more than ultra high definition (UHD) services as well as FHD. .

Accordingly, there is a demand for an image filtering method for further improving subjective / objective picture quality.

An object of the present invention is to provide a method and apparatus for improving image coding efficiency.

Another object of the present invention is to provide a method and apparatus for improving the subjective / objective picture quality of a reconstructed picture.

Another technical problem of the present invention is to provide a filtering method and apparatus using a motion vector of a prediction target block.

Another technical problem of the present invention is to provide a filtering method and apparatus through weighted sum of a prediction sample derived based on a motion vector of a prediction target block and a reconstruction sample of a filtering target block.

According to an embodiment of the present invention, there is provided an image decoding method performed by a decoding apparatus. The method may include deriving a motion vector of a prediction target block, deriving a prediction sample for the prediction target block based on the motion vector, and deriving a filtering target block based on neighboring blocks of the prediction target block. Deriving a filtering target region in the filtering target block, and performing filtering on a reconstructed sample in the filtering target region based on a prediction sample of the prediction target block.

According to another embodiment of the present invention, a decoding apparatus for performing inter prediction is provided. The decoding apparatus derives a motion vector of a prediction target block, a prediction unit for deriving a prediction sample for the prediction target block based on the motion vector, and derives a filtering target block based on neighboring blocks of the prediction target block. And a filter unit for deriving a filtering target region in the filtering target block and performing filtering on the reconstructed sample in the filtering target region based on a prediction sample of the prediction target block.

According to another embodiment of the present invention, a video encoding method performed by an encoding apparatus is provided. The method may include deriving a motion vector of a prediction target block, deriving a prediction sample for the prediction target block based on the motion vector, and deriving a filtering target block based on neighboring blocks of the prediction target block. Determining a filtering target region in the filtering target block, performing filtering on a reconstructed sample in the filtering target region based on a prediction sample of the prediction target block, and information on filtering performed on the reconstructed sample It characterized in that it comprises the step of encoding and outputting.

According to another embodiment of the present invention, a video encoding apparatus is provided. The encoding apparatus derives a motion vector of the prediction target block, a prediction unit for deriving a prediction sample for the prediction target block based on the motion vector, and derives a filtering target block based on neighboring blocks of the prediction target block. A filter unit configured to determine a filtering target region in the filtering target block, and to perform filtering on the reconstructed sample in the filtering target region based on a prediction sample of the prediction target block, and a filtering performed on the reconstructed sample. And an entropy encoding unit for encoding and outputting the information.

According to the present invention, subjective / objective image quality may be improved by filtering reconstructed samples in the neighboring block of the prediction target block.

According to the present invention, filtering can be performed by weighted sum of neighboring reconstructed samples at the boundary of the predicted sample and the predicted block derived based on the motion vector of the predicted block, thereby reducing errors between blocks and subjective. Can improve the objective image quality.

1 is a block diagram schematically illustrating a video encoding apparatus according to an embodiment of the present invention.

2 is a block diagram schematically illustrating a video decoding apparatus according to an embodiment of the present invention.

3 illustrates positions of neighboring blocks usable in a prediction target block used in image encoding / decoding.

4 illustrates an example of filtering a neighboring block using a motion vector of a prediction target block.

5 illustrates an example of a process of filtering a neighboring block based on a motion vector of a prediction target block.

6 illustrates an example of filtering target block candidates.

7 illustrates an example of comparing motion vectors to derive the filtering target block.

8 illustrates an example of weights for performing filtering on reconstructed samples in the filtering region.

9 schematically illustrates a video encoding method by an encoding device according to the present invention.

10 schematically illustrates a video decoding method by a decoding apparatus according to the present invention.

As the present invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. However, this is not intended to limit the invention to the specific embodiments. The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the spirit of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprise" or "have" are intended to indicate that there is a feature, number, step, action, component, part, or combination thereof described on the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

On the other hand, each of the components in the drawings described in the present invention are shown independently for the convenience of description of the different characteristic functions in the video encoding apparatus / decoding apparatus, each component is a separate hardware or separate software It does not mean that it is implemented. For example, two or more of each configuration may be combined to form one configuration, or one configuration may be divided into a plurality of configurations. Embodiments in which each configuration is integrated and / or separated are also included in the present invention without departing from the spirit of the present invention.

Hereinafter, with reference to the accompanying drawings, it will be described in detail an embodiment of the present invention.

Referring to FIG. 1, the encoding apparatus 100 may include a picture divider 105, a predictor 110, a transformer 115, a quantizer 120, a reordering unit 125, an entropy encoding unit 130, An inverse quantization unit 135, an inverse transform unit 140, a filter unit 145, and a memory 150 are provided.

The picture dividing unit 105 may divide the input picture into at least one processing unit block. In this case, the block as the processing unit may be a prediction unit (PU), a transform unit (TU), or a coding unit (CU). A picture may be composed of a plurality of coding tree units (CTUs), and each CTU may be split into CUs in a quad-tree structure. A CU may be divided into quad tree structures with CUs of a lower depth. PU and TU may be obtained from a CU. For example, a PU may be partitioned from a CU into a symmetrical or asymmetrical square structure. The TU may also be divided into quad tree structures from the CU.

The predictor 110 includes an inter predictor for performing inter prediction and an intra predictor for performing intra prediction, as described below. The prediction unit 110 performs prediction on the processing unit of the picture in the picture division unit 105 to generate a prediction block including a prediction sample (or a prediction sample array). The processing unit of the picture in the prediction unit 110 may be a CU, a TU, or a PU. In addition, the prediction unit 110 may determine whether the prediction performed on the processing unit is inter prediction or intra prediction, and determine specific contents (eg, prediction mode, etc.) of each prediction method. In this case, the processing unit in which the prediction is performed and the processing unit in which the details of the prediction method and the prediction method are determined may be different. For example, the method of prediction and the prediction mode may be determined in units of PUs, and the prediction may be performed in units of TUs.

Through inter prediction, a prediction block may be generated by performing prediction based on information of at least one picture of a previous picture and / or a subsequent picture of the current picture. In addition, through intra prediction, a prediction block may be generated by performing prediction based on pixel information in a current picture.

As a method of inter prediction, a skip mode, a merge mode, an advanced motion vector prediction (AMVP), and the like can be used. In inter prediction, a reference picture may be selected for a PU and a reference block corresponding to the PU may be selected. The reference block may be selected in units of integer pixels (or samples) or fractional pixels (or samples). Subsequently, a predictive block is generated in which a residual signal with the PU is minimized and the size of the motion vector is also minimized.

The prediction block may be generated in integer pixel units, or may be generated in sub-pixel units such as 1/2 pixel unit or 1/4 pixel unit. In this case, the motion vector may also be expressed in units of integer pixels or less.

Information such as an index of a reference picture selected through inter prediction, a motion vector difference (MDV), a motion vector predictor (MVP), a residual signal, and the like may be entropy encoded and transmitted to a decoding apparatus. . When the skip mode is applied, the residual may be used as the reconstructed block, and thus the residual may not be generated, transformed, quantized, or transmitted.

When performing intra prediction, a prediction mode may be determined in units of PUs, and prediction may be performed in units of PUs. In addition, a prediction mode may be determined in units of PUs, and intra prediction may be performed in units of TUs.

In intra prediction, the prediction mode may have, for example, 33 directional prediction modes and at least two non-directional modes. The non-directional mode may include a DC prediction mode and a planner mode (Planar mode).

In intra prediction, a prediction block may be generated after applying a filter to a reference sample. In this case, whether to apply the filter to the reference sample may be determined according to the intra prediction mode and / or the size of the current block.

The residual value (the residual block or the residual signal) between the generated prediction block and the original block is input to the converter 115. In addition, the prediction mode information, the motion vector information, etc. used for the prediction are encoded by the entropy encoding unit 130 together with the residual value and transmitted to the decoding apparatus.

The transform unit 115 performs transform on the residual block in units of transform blocks and generates transform coefficients.

The transform block is a rectangular block of samples to which the same transform is applied. The transform block can be a transform unit (TU) and can have a quad tree structure.

The transformer 115 may perform the transformation according to the prediction mode applied to the residual block and the size of the block.

For example, if intra prediction is applied to a residual block and the block is a 4x4 residual array, the residual block is transformed using a discrete sine transform (DST), otherwise the residual block is transformed into a DCT (Discrete). Can be transformed using Cosine Transform.

The transform unit 115 may generate a transform block of transform coefficients by the transform.

The quantization unit 120 may generate quantized transform coefficients by quantizing the residual values transformed by the transform unit 115, that is, the transform coefficients. The value calculated by the quantization unit 120 is provided to the inverse quantization unit 135 and the reordering unit 125.

The reordering unit 125 rearranges the quantized transform coefficients provided from the quantization unit 120. By rearranging the quantized transform coefficients, the encoding efficiency of the entropy encoding unit 130 may be increased.

The reordering unit 125 may rearrange the quantized transform coefficients in the form of a 2D block into a 1D vector form through a coefficient scanning method.

The entropy encoding unit 130 entropy-codes a symbol according to a probability distribution based on the quantized transform values rearranged by the reordering unit 125 or the encoding parameter value calculated in the coding process, thereby performing a bitstream. You can output The entropy encoding method receives a symbol having various values and expresses it as a decodable column while removing statistical redundancy.

Here, the symbol means a syntax element, a coding parameter, a value of a residual signal, etc., to be encoded / decoded. An encoding parameter is a parameter necessary for encoding and decoding, and may include information that may be inferred in the encoding or decoding process as well as information encoded by an encoding device and transmitted to the decoding device, such as a syntax element. It means the information you need when you do. The encoding parameter may be, for example, a value such as an intra / inter prediction mode, a moving / motion vector, a reference image index, a coding block pattern, a residual signal presence, a transform coefficient, a quantized transform coefficient, a quantization parameter, a block size, block partitioning information, or the like. May include statistics. In addition, the residual signal may mean a difference between the original signal and the prediction signal, and a signal in which the difference between the original signal and the prediction signal is transformed or a signal in which the difference between the original signal and the prediction signal is converted and quantized It may mean. The residual signal may be referred to as a residual block in the block unit, and the residual sample in the sample unit.

When entropy encoding is applied, a small number of bits are allocated to a symbol having a high probability of occurrence and a large number of bits are allocated to a symbol having a low probability of occurrence, thereby representing the size of the bit string for the symbols to be encoded. Can be reduced. Therefore, the compression performance of video encoding can be increased through entropy encoding.

Encoding methods such as exponential golomb, context-adaptive variable length coding (CAVLC), and context-adaptive binary arithmetic coding (CABAC) may be used for entropy encoding. For example, the entropy encoding unit 130 may store a table for performing entropy encoding, such as a variable length coding (VLC) table, and the entropy encoding unit 130 may store the variable length coding. Entropy encoding can be performed using the (VLC) table. In addition, the entropy encoding unit 130 derives the binarization method of the target symbol and the probability model of the target symbol / bin, and then uses the derived binarization method or the probability model to entropy. You can also perform encoding.

In addition, if necessary, the entropy encoding unit 130 may apply a constant change to a parameter set or syntax to be transmitted.

The inverse quantizer 135 inversely quantizes the quantized values (quantized transform coefficients) in the quantizer 120, and the inverse transformer 140 inversely transforms the inverse quantized values in the inverse quantizer 135.

The residual value (or the residual sample or the residual sample array) generated by the inverse quantizer 135 and the inverse transform unit 140 and the prediction block predicted by the predictor 110 are added together to reconstruct the sample (or the reconstructed sample array). A reconstructed block including a may be generated.

In FIG. 1, it is described that a reconstructed block is generated by adding a residual block and a prediction block through an adder. In this case, the adder may be viewed as a separate unit (restore block generation unit) for generating a reconstruction block.

The filter unit 145 may apply a deblocking filter, an adaptive loop filter (ALF), and a sample adaptive offset (SAO) to the reconstructed picture.

The deblocking filter may remove distortion generated at the boundary between blocks in the reconstructed picture. The adaptive loop filter (ALF) may perform filtering based on a value obtained by comparing the reconstructed image with the original image after the block is filtered through the deblocking filter. ALF may be performed only when high efficiency is applied. The SAO restores the offset difference from the original image on a pixel-by-pixel basis for the residual block to which the deblocking filter is applied, and is applied in the form of a band offset and an edge offset.

Meanwhile, the filter unit 145 may not apply filtering to the reconstructed block used for inter prediction.

The memory 150 may store the reconstructed block or the picture calculated by the filter unit 145. The reconstructed block or picture stored in the memory 150 may be provided to the predictor 110 that performs inter prediction.

2 is a block diagram schematically illustrating a video decoding apparatus according to an embodiment of the present invention. Referring to FIG. 2, the video decoding apparatus 200 includes an entropy decoding unit 210, a reordering unit 215, an inverse quantization unit 220, an inverse transform unit 225, a prediction unit 230, and a filter unit 235. Memory 240 may be included.

When an image bitstream is input in the video encoding apparatus, the input bitstream may be decoded according to a procedure in which image information is processed in the video encoding apparatus.

The entropy decoding unit 210 may entropy decode the input bitstream according to a probability distribution to generate symbols including symbols in the form of quantized coefficients. The entropy decoding method is a method of generating each symbol by receiving a binary string. The entropy decoding method is similar to the entropy encoding method described above.

For example, when variable length coding (VLC, hereinafter called 'VLC') is used to perform entropy encoding in the video encoding apparatus, the entropy decoding unit 210 also uses the VLC. Entropy decoding may be performed by implementing the same VLC table as the table. In addition, when CABAC is used to perform entropy encoding in the video encoding apparatus, the entropy decoding unit 210 may correspondingly perform entropy decoding using CABAC.

More specifically, the CABAC entropy decoding method receives a bin corresponding to each syntax element in a bitstream, and decodes syntax element information and decoding information of neighboring and decoding target blocks or information of symbols / bins decoded in a previous step. The context model may be determined using the context model, the probability of occurrence of a bin may be predicted according to the determined context model, and arithmetic decoding of the bin may be performed to generate a symbol corresponding to the value of each syntax element. have. In this case, the CABAC entropy decoding method may update the context model by using the information of the decoded symbol / bin for the context model of the next symbol / bean after determining the context model.

Information for generating the prediction block among the information decoded by the entropy decoding unit 210 is provided to the predictor 230, and a residual value where entropy decoding is performed by the entropy decoding unit 210, that is, a quantized transform coefficient It may be input to the reordering unit 215.

The reordering unit 215 may reorder the information of the bitstream entropy decoded by the entropy decoding unit 210, that is, the quantized transform coefficients, based on the reordering method in the encoding apparatus.

The reordering unit 215 may reorder the coefficients expressed in the form of a one-dimensional vector by restoring the coefficients in the form of a two-dimensional block. The reordering unit 215 scans the coefficients based on the prediction mode applied to the current block (transform block) and the size of the transform block to generate an array of coefficients (quantized transform coefficients) in the form of a two-dimensional block. Can be.

The inverse quantization unit 220 may perform inverse quantization based on the quantization parameter provided by the encoding apparatus and the coefficient values of the rearranged block.

The inverse transform unit 225 may perform inverse DCT and / or inverse DST on the DCT and the DST performed by the transform unit of the encoding apparatus with respect to the quantization result performed by the video encoding apparatus.

The inverse transformation may be performed based on a transmission unit determined by the encoding apparatus or a division unit of an image. The DCT and / or DST in the encoding unit of the encoding apparatus may be selectively performed according to a plurality of pieces of information, such as a prediction method, a size and a prediction direction of the current block, and the inverse transform unit 225 of the decoding apparatus is configured in the transformation unit of the encoding apparatus. Inverse transformation may be performed based on the performed transformation information.

The prediction unit 230 may include prediction samples (or prediction sample arrays) based on prediction block generation related information provided by the entropy decoding unit 210 and previously decoded block and / or picture information provided by the memory 240. A prediction block can be generated.

When the prediction mode for the current PU is an intra prediction mode, intra prediction for generating a prediction block based on pixel information in the current picture may be performed.

When the prediction mode for the current PU is an inter prediction mode, inter prediction on the current PU may be performed based on information included in at least one of a previous picture or a subsequent picture of the current picture. In this case, motion information required for inter prediction of the current PU provided by the video encoding apparatus, for example, a motion vector, a reference picture index, and the like, may be derived by checking a skip flag, a merge flag, and the like received from the encoding apparatus.

In inter prediction on a current picture, a prediction block may be generated such that a residual signal with a current block is minimized and a motion vector size is also minimized.

Meanwhile, the motion information derivation scheme may vary depending on the prediction mode of the current block. Prediction modes applied for inter prediction may include an advanced motion vector prediction (AMVP) mode, a merge mode, and the like.

For example, when the merge mode is applied, the encoding apparatus and the decoding apparatus may generate a merge candidate list by using the motion vector of the reconstructed spatial neighboring block and / or the motion vector corresponding to the Col block, which is a temporal neighboring block. . In the merge mode, the motion vector of the candidate block selected from the merge candidate list is used as the motion vector of the current block. The encoding apparatus may transmit, to the decoding apparatus, a merge index indicating a candidate block having an optimal motion vector selected from candidate blocks included in the merge candidate list. In this case, the decoding apparatus may derive the motion vector of the current block by using the merge index.

As another example, when the Advanced Motion Vector Prediction (AMVP) mode is applied, the encoding device and the decoding device use a motion vector corresponding to a motion vector of a reconstructed spatial neighboring block and / or a Col block, which is a temporal neighboring block, and a motion vector. A predictor candidate list may be generated. That is, the motion vector of the reconstructed spatial neighboring block and / or the Col vector, which is a temporal neighboring block, may be used as a motion vector candidate. The encoding apparatus may transmit the predicted motion vector index indicating the optimal motion vector selected from the motion vector candidates included in the list to the decoding apparatus. In this case, the decoding apparatus may select the predicted motion vector of the current block among the motion vector candidates included in the motion information candidate list using the motion vector index.

The encoding apparatus may obtain a motion vector difference MVD between the motion vector MV of the current block and the motion vector predictor MVP, and may encode the same and transmit the encoded motion vector to the decoding device. That is, MVD may be obtained by subtracting MVP from MV of the current block. In this case, the decoding apparatus may decode the received motion vector difference and derive the motion vector of the current block through the addition of the decoded motion vector difference and the motion vector predictor.

The encoding apparatus may also transmit a reference picture index or the like indicating the reference picture to the decoding apparatus.

The decoding apparatus may predict the motion vector of the current block using the motion information of the neighboring block, and may derive the motion vector for the current block using the residual received from the encoding apparatus. The decoding apparatus may generate a prediction block for the current block based on the derived motion vector and the reference picture index information received from the encoding apparatus.

As another example, when the merge mode is applied, the encoding apparatus and the decoding apparatus may generate the merge candidate list using the motion information of the reconstructed neighboring block and / or the motion information of the call block. That is, the encoding apparatus and the decoding apparatus may use this as a merge candidate for the current block when there is motion information of the reconstructed neighboring block and / or the call block.

The encoding apparatus may select a merge candidate capable of providing an optimal encoding efficiency among the merge candidates included in the merge candidate list as motion information for the current block. In this case, a merge index indicating the selected merge candidate may be included in the bitstream and transmitted to the decoding apparatus. The decoding apparatus may select one of the merge candidates included in the merge candidate list by using the transmitted merge index, and determine the selected merge candidate as motion information of the current block. Therefore, when the merge mode is applied, motion information corresponding to the reconstructed neighboring block and / or the call block may be used as the motion information of the current block. The decoding apparatus may reconstruct the current block by adding the prediction block and the residual transmitted from the encoding apparatus.

In the above-described AMVP and merge mode, the motion information of the reconstructed neighboring block and / or the motion information of the call block may be used to derive the motion information of the current block.

In the case of a skip mode, which is one of other modes used for inter prediction, information of neighboring blocks may be used in the current block as it is. Therefore, in the skip mode, the encoding apparatus does not transmit syntax information such as residual to the decoding apparatus other than information indicating which block motion information to use as the motion information of the current block.

The encoding apparatus and the decoding apparatus may generate the prediction block of the current block by performing motion compensation on the current block based on the derived motion information. Here, the prediction block may mean a motion compensated block generated as a result of performing motion compensation on the current block. Also, the plurality of motion compensated blocks may constitute one motion compensated image.

The reconstruction block may be generated using the prediction block generated by the predictor 230 and the residual block provided by the inverse transform unit 225. In FIG. 2, it is described that the reconstructed block is generated by combining the prediction block and the residual block in the adder. In this case, the adder may be viewed as a separate unit (restore block generation unit) for generating a reconstruction block. Here, the reconstruction block includes a reconstruction sample (or reconstruction sample array) as described above, the prediction block includes a prediction sample (or a prediction sample array), and the residual block is a residual sample (or a residual sample). Array). Thus, a reconstructed sample (or reconstructed sample array) may be expressed as the sum of the corresponding predictive sample (or predictive sample array) and the residual sample (residual sample array).

The residual is not transmitted for the block to which the skip mode is applied, and the prediction block may be a reconstruction block.

The reconstructed block and / or picture may be provided to the filter unit 235. The filter unit 235 may apply deblocking filtering, sample adaptive offset (SAO), and / or ALF to the reconstructed block and / or picture.

The memory 240 may store the reconstructed picture or block to use as a reference picture or reference block and provide the reconstructed picture to the output unit.

Entropy decoding unit 210, reordering unit 215, inverse quantization unit 220, inverse transform unit 225, predictor 230, filter unit 235, and memory 240 included in the decoding device 200. ) Components directly related to the decoding of an image, for example, an entropy decoding unit 210, a reordering unit 215, an inverse quantization unit 220, an inverse transform unit 225, a prediction unit 230, and a filter unit ( 235) and the like may be distinguished from other components by a decoder or a decoder.

In addition, the decoding apparatus 200 may further include a parsing unit (not shown) for parsing information related to the encoded image included in the bitstream. The parsing unit may include the entropy decoding unit 210 or may be included in the entropy decoding unit 210. Such a parser may also be implemented as one component of the decoder.

As described above, since the compression coding process of the image is performed in units of blocks, distortion may occur at boundaries between blocks in the reconstructed picture. In order to reduce an error that increases in the boundary region between blocks generated during encoding and decoding, after the generation of the reconstructed sample of the predicted block is completed, the neighboring block of the predicted block that has already been decoded is filtered to the neighboring block. Can improve the quality.

In general, image compression may divide an image into specific block units (eg, coding tree units, CTUs), and encode and decode blocks by one block in a sequential order from the upper left of the image. The order may be a raster scan order. In addition, in image compression, in order to remove spatial and temporal redundancy, the prediction target block may be encoded and decoded by referring to information of neighboring blocks that have been encoded and decoded among neighboring blocks of the prediction target block.

3 illustrates positions of neighboring blocks usable in a prediction target block used in image encoding / decoding. Referring to FIG. 3, a usable peripheral block may be a left peripheral block, an upper peripheral block, a left-above peripheral block, or a right-above peripheral block.

However, the information of the prediction target block may also be used to improve visual quality of neighboring blocks that are already decoded. The present invention provides a method for improving the quality of the neighboring block based on the motion information of the prediction target block.

When the image is divided and encoded and decoded on the basis of a quad-tree or an equal division scheme, samples (or pixels) located at the boundary of the block may have a relatively large quantization error. Blocking degradation may occur. By applying filtering to a sample located at the boundary of the neighboring block based on the motion vector of the prediction target block, an error of the sample located at the boundary of the neighboring block can be reduced. According to the present invention, a prediction sample of a sample located at the boundary of the neighboring block may be derived based on the motion vector of the prediction target block, and the value of the sample located at the boundary of the prediction sample and the neighboring block is determined. Filtering may be performed through a weight sum of.

4 illustrates an example of filtering a neighboring block using a motion vector of a prediction target block. Referring to FIG. 4, a filtering target region may be determined based on a motion vector of the prediction target block, and filtering may be performed on the filtering target region. The filtering subject region may include samples located at a boundary of at least one of the above-described neighboring blocks.

Referring to FIG. 5, the encoding apparatus and the decoding apparatus according to the present invention are provided through the filtering target block determining unit 500, the filtering target region determining unit 510, the filter parameter deriving unit 520, and the filtering execution unit 530. Filtering can be performed on neighboring blocks.

The filtering target block determiner 500 may derive a filtering target block to perform filtering among neighboring blocks adjacent to the prediction target block. The size of the filtering target block may be the same as or different from the size of the prediction target block. The filtering object block may be derived using a motion vector of the prediction object block. The filtering target region determiner 510 may derive a boundary region adjacent to the prediction target block to be filtered in the filtering target block. The filtering target region may be variably determined according to the size of the filtering target block. The filtering parameter derivation unit 520 may derive whether to perform filtering on the filtering target region and a weight value used for filtering. The filtering performing unit 530 may perform the filtering derived by the filter parameter deriving unit 520 on a region determined by the filtering target region determining unit 510 with respect to the block determined by the filtering target block determining unit 500. Filtering can be applied using the presence and absence of filter parameters. The filtering target block determining unit 500, the filtering target region determining unit 510, the filter parameter deriving unit 520, and the filtering performing unit 530 may be included in a filter unit, and may be implemented by a processor. Can be.

The filtering target block determiner 500 may derive a neighboring block to perform filtering among neighboring blocks adjacent to the prediction target block. Conditions derived to the filtering target block may be as follows.

The filtering target block may be one of a left neighboring block, an upper neighboring block, and the upper left neighboring block among adjacent neighboring blocks of the prediction object block. In addition, the motion information of the filtering target block may be different from the motion information of the prediction target block. The motion information may include a motion vector and a reference picture index associated with the motion vector.

Although FIG. 5 illustrates the filtering target block determining unit 500, the filtering target region determining unit 510, the filter parameter deriving unit 520, and the filtering performing unit 530 separately according to functions. For example, the filtering target block determining unit 500, the filtering target region determining unit 510, the filter parameter deriving unit 520, and the filtering performing unit 530 may be included in a filter unit. . That is, all of the above functions may be performed by one filter unit.

6 illustrates an example of the filtering target block candidates. Referring to FIG. 6, a left neighboring block, an upper neighboring block, and a left upper neighboring block may be included in the filtering target block candidates. The filtering target block candidate may include not only the neighboring block to which the inter prediction mode is applied but also the neighboring block to which the intra prediction mode is applied.

When the motion vector of the filtering target block candidate is different from the motion vector of the prediction target block, the filtering target block candidate may be derived as the filtering target block. The motion vector of the prediction target block may be included in the motion information of the prediction target block. The motion information of the prediction target block may be bi-predicted motion information or may be L0 or L1 predicted motion information. The L0 or L1 predicted motion information may also be referred to as uni-predicted motion information. The bi-predicted motion information may include an L0 motion vector and an L1 motion vector, and the unipredicted motion information may include only one of the L0 motion vector and the L1 motion vector. The L0 motion vector represents a motion vector relating to L0 prediction, and the L1 motion vector represents a motion vector relating to L1 prediction. L0 represents a reference picture list L0 (list 0), and L1 represents a reference picture list L1 (list 1). In detail, when the motion information of the prediction target block is bi-predictive motion information, the L0 reference picture index, the L0 motion vector, the L1 reference picture index, and the L1 motion vector may be included. When the motion information of the prediction block is unipredicted motion information, the L0 reference picture index and the L0 motion vector may be included, or the L1 reference picture index and the L1 motion vector may be included.

When the motion information of the prediction target block is mono-predictive motion information, if the filtering target block candidate has the same reference picture index and the motion vector as the reference picture index and the motion vector of the prediction target block, the filtering target block candidate is the It may not be derived to the filtering target block. For example, when the unipredicted motion information of the prediction target block includes an L0 reference picture index and an L0 motion vector, and the same as the L0 reference picture index and the L0 motion vector of the filtering target block candidate, the filtering target block candidate may be used. It may not be derived to the filtering target block.

As another example, when the unipredicted motion information of the prediction target block includes an L1 reference picture index and an L1 motion vector, and the same as the L1 reference picture index and the L1 motion vector of the filtering target block candidate, the filtering target block candidate is the It may not be derived to the filtering target block.

On the other hand, when the filtering target block candidate has the same reference picture index as the reference picture index of the prediction target block, but the motion vector is different, it may be derived as the filtering target block. For example, the short-predicted motion information of the prediction target block includes an L0 reference picture index and an L0 motion vector, the L0 reference picture index of the filtering target block candidate and the L0 reference picture index of the prediction target block are the same, and When the L0 motion vector of the filtering target block candidate and the L0 motion vector of the prediction target block are not the same, the filtering target block candidate may be derived as the filtering target block.

As another example, the unidirectional motion information of the prediction target block includes an L1 reference picture index and an L1 motion vector, the L1 reference picture index of the filtering target block candidate and the L1 reference picture index of the prediction target block are the same, When the L1 motion vector of the filtering target block candidate and the L1 motion vector of the prediction target block are not the same, the filtering target block candidate may be derived as the filtering target block.

On the other hand, when the motion information of the prediction target block is bi-predictive motion information, the motion information of the filtering target block candidate matches both the L0 prediction motion information of the prediction target block and the L1 prediction motion information of the prediction target block. The filtering target block candidate may not be derived as the filtering target block.

If the motion information of the filtering target block candidate does not match one of the L0 prediction motion information of the prediction target block and the L1 prediction motion information of the prediction target block, the filtering target block candidate may be derived as the filtering target block. have. For example, when the L0 motion vector of the prediction target block and the L0 motion vector of the filtering target block candidate do not match, the filtering target block candidate may be derived as the filtering target block. In this case, the case where the L0 motion vector of the prediction target block and the L0 motion vector of the filtering target block candidate do not coincide with each other and may include a case where the L0 motion vector of the filtering target block candidate does not exist. In this case, the filtering may be performed based on the L0 motion vector of the prediction target block.

As another example, when the L1 motion vector of the prediction target block and the L1 motion vector of the filtering target block candidate do not match, the filtering target block candidate may be derived as the filtering target block. In this case, the case where the L1 motion vector of the prediction target block and the L1 motion vector of the filtering target block candidate do not coincide with each other and may include a case where the L1 motion vector of the filtering target block candidate does not exist. In this case, the filtering may be performed based on the L1 motion vector of the prediction target block.

If the motion information of the filtering target block candidate does not match both the L0 prediction motion information of the prediction target block and the L1 prediction motion information of the prediction target block, the L0 motion vector and the prediction target block of the prediction target block. The filtering may be performed based on the L1 motion vector.

When filtering is performed based on one of the L0 motion vector and the L1 motion vector of the prediction target block, and when filtering is performed based on the L0 motion vector and the L1 motion vector of the prediction block, the weight value is different. Can be.

7 illustrates an example of comparing motion vectors to derive the filtering target block. FIG. 7A illustrates an example of comparing motion vectors when the size of the filtering target block candidate is the same as the size of the prediction target block. FIG. 7B illustrates an example of comparing motion vectors when the size of the filtering target block candidates is smaller than the size of the prediction target block.

As shown in FIG. 7A, when MV _L which is a motion vector of a filtering target block candidate and a motion vector MV _C of a prediction target block are different, filtering may be performed on the prediction target block side of the filtering target block. have. In other words, the filtering target block candidate may be derived as a filtering target block, and a boundary where the boundary of the prediction target block and the boundary of the filtering target block overlap with each other may be referred to as a reference boundary. A filtering target region in the reference boundary direction of the filtering target block may be derived, and filtering may be performed on the filtering target region.

As shown in (b) of FIG. 7, when the filtering target block candidates are divided into sub-blocks smaller than the prediction target block (that is, the size of the filtering target block candidates is smaller than the size of the prediction target block). ), A motion vector of each filtering target block candidate and a motion vector of the prediction target block may be compared.

Specifically, for example, since the motion vector MV _C of the prediction target block is different from MV _L2 , the motion vector of the filtering target block candidate located on the upper right side of the filtering target block candidates, the filtering target block candidate is a filtering target block. Can be derived. A filtering target region in the reference boundary direction of the filtering target block may be derived, and filtering may be performed on the filtering target region.

On the other hand, since the motion vector MV _C of the prediction target block is the same as the motion vector MV _L4 of the filtering target block candidate located on the lower right side among the filtering target block candidates, the filtering target block candidate may not be derived as the filtering target block. Can be.

Meanwhile, when the motion vector of the filtering target block candidate does not exist, for example, when an intra prediction mode is applied to the filtering target block candidate, the motion vector of the filtering target block candidate is different from the motion vector of the prediction target block. In this case, the filtering target block candidate may be derived as the filtering target block, and filtering may be performed.

On the other hand, when the prediction target block is divided into PU units, whether to perform filtering, that is, whether to derive the filtering target block by comparing the motion vector of each sub-block with the motion vector of the filtering target block located around the prediction target block Can be determined.

The filtering target region determiner may derive a filtering target region to which samples are filtered when filtering the boundary of the prediction target block. Conditions for derivation into the filtering target region may be as follows.

The filtering target region may be derived based on the size of the prediction target block. In addition, the filtering target region may be derived based on a size of a filtering target block including the filtering target region. The filtering target region may be derived based on a difference between the size of the prediction target block and the size of the filtering target block including the filtering target region. The filtering target region may be derived based on whether an intra prediction mode is applied to a filtering target block including the filtering target region. The filtering target region may be derived based on the presence or absence of a quantization coefficient of the filtering target block including the filtering target region.

For example, since characteristics of a block may vary according to the size of the prediction target block or the size of the filtering target block, a filtering target region may be derived based on the size of the prediction target block and the size of the filtering target block. Can be. The filtering target region may be as shown in Table 1 below.

Here,> wxh indicates a case where the size of the block is larger than wxh, and <= wxh indicates a case where the size of the block is wxh or less. The decoding apparatus may determine the filtering target region based on a mapping table of a table set in advance, for example, Table 1, above, without performing a separate calculation process.

Meanwhile, when the value of the difference between the size of the prediction target block and the size of the filtering target block is smaller than a predefined threshold value TH1, filtering may not be performed. If the value of the difference from the size of the block to be filtered is greater than the predefined threshold value TH2, filtering may not be performed.

As another example, the filtering target region may be derived based on whether an intra prediction mode is applied to the filtering target block. When the intra prediction mode is applied to the filtering target block, the filtering target region may be fixed to a preset range regardless of the size of the filtering target block. When the intra prediction mode is applied, the filtering target block may include a relatively larger error than when the inter prediction mode is applied on the boundary side of the block. Therefore, filtering may be performed to compensate for the error. However, when the quantization coefficient of the filtering target block does not exist, it may be determined that an error of the filtering target block is small and excluded from the filtering target block, that is, the filtering may not be performed.

Referring to FIG. 8, the weight may be derived based on a distance from the reference boundary to the reconstructed sample. The filtering parameter derivation unit may derive a weight used for the filtering. According to the above description, the filtering subject region determiner may determine how many reconstructed samples are included in the filtered subject region based on the reference boundary. In this case, a sample adjacent to the reference boundary may be referred to as a first reconstructed sample, and may be sequentially ordered in a direction away from the reference boundary. In this case, different weights may be used for each position of the reconstructed sample.

8 exemplarily shows a filtering weight for each reconstruction sample for each position. FIG. 8A illustrates an example of weights derived from the reference boundary from the reference boundary to the second reconstruction sample, and FIG. 8B illustrates the filtering region from the reference boundary to the fourth reconstruction sample. An example of weights in case of derivation is shown. Here, P _C may represent prediction samples of a position corresponding to the reconstruction samples of the filtering target region derived based on the motion vector of the prediction target block, and P _N may represent the reconstruction samples of the filtering target region. Can be represented. P _N {A, B} represents the weight for the reconstructed sample located at A or B, and P _C {A, B} represents the weight for the predictive sample located at A or B. For example, in FIG. 8A, P _N {A, B} = {1/2, 3/4}, P _C {A, B} = {1/2, 1/4} days have. In detail, the weight of the reconstructed sample of the first position based on the A position, that is, the reference boundary may be 1/2, and the weight of the predicted sample of the first position based on the reference boundary may be 1/2. . In addition, the weight of the reconstructed sample of the second position based on the reference position B, that is, the reference boundary may be 3/4, and the weight of the predicted sample of the second position based on the reference boundary may be 1/4.

As another example, in the case of FIG. 8B, P _N {A, B, C, D} = {1/4, 1/2, 3/4, 7/8}, P _C {A, B, C , D} = {3/4, 1/2, 1/4, 1/8}. In detail, the weight of the reconstructed sample of the first position based on the A position, that is, the reference boundary may be 1/4, and the weight of the predicted sample of the first position based on the reference boundary may be 3/4. . In addition, the weight of the reconstructed sample of the second position based on the reference position B, that is, the reference boundary may be 1/2, and the weight of the predicted sample of the second position based on the reference boundary may be 1/2. In addition, the weight of the reconstructed sample of the third position based on the C position, that is, the reference boundary may be 3/4, and the weight of the predicted sample of the third position based on the reference boundary may be 1/4. In addition, the weight of the reconstructed sample of the fourth position based on the reference position D, that is, the reference boundary may be 7/8, and the weight of the predicted sample of the fourth position based on the reference boundary may be 1/7.

Based on the reference boundary, the closer the distance to the reference boundary is, the greater the weight for the prediction sample derived based on the motion vector of the prediction target block is, and as the inside of the filtering target block increases, ie As the distance from the reference boundary increases, the weight for the reconstructed sample may increase.

Meanwhile, the weight for performing the filtering may be set for each position of the reconstructed sample included in the filtering target region in the encoding / decoding process. The encoding apparatus and the decoding apparatus may have a weight table including weight candidates of the reconstructed samples. In this case, the decoding apparatus may receive the weight index through a bitstream, and the weight for the reconstructed sample for each position and the weight for the prediction sample based on the candidate included in the weight table indicated by the weight index. Can be derived. The weight index may be transmitted in a slice unit or a CU unit.

In addition, the encoding apparatus may generate a flag indicating whether filtering is applied to neighboring blocks based on the motion vector of the prediction target block, and may encode and transmit the flag. The flag may be referred to as a block-boundary filtering (bbf) flag. The weight index and bbf flag may be transmitted through a syntax as shown in Table 2 below.

Referring to Table 5, the bbf_flag syntax element may correspond to the bbf flag, and the weight_idx syntax element may correspond to the weight index. For example, when the value of the bbf flag is 1, filtering may be performed on neighboring blocks based on the motion vector of the prediction target block, and when the value of the bbf flag is 0, filtering may not be performed. have. In addition, the weight index may be transmitted when the value of the bbf flag is 1. The weight index may indicate a candidate included in the weight table. The weight table may be as shown in Table 3 below.

Table 3 shows an example of a weight table applied when the filtering target region includes up to fourth reconstructed samples based on the reference boundary. In Table 3, various weighting candidates may be added to the weighting table. The weights for the first sample, the second sample, the third sample, and the fourth sample may be indicated sequentially from the front of the weighting candidate. The weights shown in Table 3 may be weights for the prediction samples derived based on the motion vector of the prediction target block. In this case, weights for the reconstructed samples corresponding to the prediction samples may be obtained in the form of (1-α) based on the weights. Here, α may represent a weight for the prediction sample.

9 schematically illustrates a video encoding method by an encoding device according to the present invention. The method disclosed in FIG. 9 may be performed by the encoding apparatus disclosed in FIG. 1. Specifically, for example, S900 to S910 of FIG. 9 may be performed by the prediction unit of the encoding apparatus, S920 to S940 may be performed by the filter unit of the encoding apparatus, and S950 may be entropy of the encoding apparatus. It may be performed by the encoding unit.

The encoding apparatus derives a motion vector of the prediction target block (S900). The motion vector of the prediction target block may be included in the motion information of the prediction target block. The motion information of the prediction target block may be bi-predicted motion information or may be L0 or L1 predicted motion information. The L0 or L1 predicted motion information may also be referred to as uni-predicted motion information. The bi-predicted motion information may include an L0 motion vector and an L1 motion vector, and the unipredicted motion information may include only one of the L0 motion vector and the L1 motion vector. The L0 motion vector represents a motion vector relating to L0 prediction, and the L1 motion vector represents a motion vector relating to L1 prediction. L0 represents a reference picture list L0 (list 0), and L1 represents a reference picture list L1 (list 1). In detail, when the motion information of the prediction target block is bi-predictive motion information, the L0 reference picture index, the L0 motion vector, the L1 reference picture index, and the L1 motion vector may be included. When the motion information of the prediction block is unipredicted motion information, the L0 reference picture index and the L0 motion vector may be included, or the L1 reference picture index and the L1 motion vector may be included.

The encoding apparatus derives a prediction sample for the prediction block based on the motion vector (S910). The encoding apparatus may obtain a predictive sample value on the reference picture indicated by the motion vector, and generate the predictive sample.

Meanwhile, when the prediction target block has bi-predicted motion information and the motion vector of the prediction target block includes an L0 motion vector and an L1 motion vector, the encoding apparatus filters the neighboring block that has already been decoded. The motion vector for deriving the prediction sample may be different by comparing the motion vector of the filtering target block candidate with the L0 motion vector and the L1 motion vector of the prediction target block.

For example, when the motion vector of the filtering target block candidate is not the same as both the L0 motion vector and the L1 motion vector, a prediction sample of the prediction target block is based on the L0 motion vector and the L1 motion vector. Can be derived.

As another example, when the motion vector of the filtering target block candidate is the same as the L0 motion vector and the motion vector of the filtering target block candidate is not the same as the L1 motion vector, the prediction object based on the L1 motion vector. The prediction sample of the block can be derived.

As another example, when the motion vector of the filtering target block candidate is not the same as the L0 motion vector and the motion vector of the filtering target block candidate is the same as the L1 motion vector, the prediction target block based on the L0 motion vector. The prediction sample of can be derived.

The encoding apparatus derives a filtering target block based on neighboring blocks of the prediction target block (S920). The filtering target block may be at least one of a left neighboring block, a left upper neighboring block, and an upper neighboring block of the prediction target block. In addition, the neighboring block may have the same size as or different from the prediction target block. For example, the neighboring block may be smaller in size than the prediction target block.

In addition, the encoding apparatus may derive a neighboring block which is already decoded among the neighboring blocks of the prediction target block as a filtering target block candidate, and compare the motion vector of the filtering target block candidate and the motion vector of the prediction target block. have. When the motion vector of the filtering target block candidate and the motion vector of the prediction target block are not the same, the filtering target block candidate may be derived as the filtering target block.

For example, the motion information of the prediction target block may be uni-prediction motion information, and the uni-prediction motion information may include a L0 motion vector. In this case, when the motion vector of the filtering target block candidate is not the same as the L0 motion vector, the filtering target block candidate may be derived as the filtering target block. In addition, when the filtering target block candidate has the same reference picture index as the L0 reference picture index of the prediction target block, but the motion vector of the filtering target block candidate is not the same as the L0 motion vector, the filtering target block candidate is derived. Can be.

As another example, the motion information of the prediction target block may be unipredicted motion information, and the unipredicted motion information may include an L1 motion vector. In this case, when the motion vector of the filtering target block candidate is not the same as the L1 motion vector, the filtering target block candidate may be derived as the filtering target block. In addition, when the filtering target block candidate has the same reference picture index as the L1 reference picture index of the prediction target block, but the motion vector of the filtering target block candidate is not the same as the L1 motion vector, the filtering target block candidate is derived. Can be.

As another example, the motion information of the prediction target block may be bi-predictive motion information, and the bi-predictive motion information may include an L0 motion vector and an L1 motion vector. In this case, when the motion vector of the filtering target block candidate is not the same as the L0 motion vector and the L1 motion vector, the filtering target block candidate may be derived as the filtering target block. Specifically, when the motion vector of the filtering target block candidate is not the same as the L0 motion vector, the filtering target block candidate may be derived as the filtering target block. In addition, when the motion vector of the filtering target block candidate is not the same as the L1 motion vector, the filtering target block candidate may be derived as the filtering target block. In addition, when the motion vector of the filtering target block candidate is not the same as both the L0 motion vector and the L1 motion vector, the filtering target block candidate may be derived as the filtering target block.

As another example, when an intra prediction mode is applied to the filtering target block candidate, the filtering target block candidate may be derived as the filtering target block. In this case, it may be assumed that the motion vector of the filtering target block candidate and the motion vector of the prediction target block are not the same.

Meanwhile, when the quantization coefficient of the filtering target block candidate does not exist among the filtering target block candidates, the filtering target block candidate may not be derived as the filtering target block.

In addition, when the value of the difference between the size of the prediction target block and the size of the filtering target block candidate is smaller than a predefined threshold value TH1, the prediction target block may not be derived. When the value of the difference between the size of and the size of the filtering target block candidate is larger than a predefined threshold value TH2, it may not be derived to the filtering target block.

The encoding apparatus determines a filtering target region in the filtering target block (S930). The filtering target region may include reconstructed samples around a reference boundary. The reference boundary may represent a boundary where a boundary of the prediction target block and a boundary of the filtering target block overlap. In detail, as an example, the filtering target region may be determined based on a difference value between the value of the motion vector of the filtering target block and the value of the motion vector of the prediction target block.

As another example, the filtering target region may be determined based on the size of the prediction target block, may be determined based on the size of the filtering target block, and based on the size of the prediction target block and the size of the filtering target block. Can be determined. The filtering target region may be as shown in Table 1 above.

As another example, the filtering target region may be determined based on whether an intra prediction mode is applied to the filtering target block. When the intra prediction mode is applied to the filtering target block, the filtering target region may be fixed to a preset range regardless of the size of the filtering target block.

As another example, the filtering target region may be determined based on the presence or absence of a quantization coefficient of the filtering target block.

The encoding apparatus performs filtering on the reconstructed samples in the filtering target region based on the prediction samples of the prediction target block (S940). The encoding apparatus may perform filtering based on a weighted sum of the value of the prediction sample and the value of the reconstructed sample. In this case, the weight of the prediction sample and the weight of the reconstruction sample may be derived based on a distance from a reference boundary to the reconstruction sample. The reference boundary may represent a boundary where a boundary of the prediction target block and a boundary of the filtering target block overlap.

For example, the weight for the prediction sample may increase as the distance from the reference boundary is closer to the reference boundary, and the weight for the reconstructed sample may increase as the distance from the reference boundary increases. Can be. The weight of the prediction sample and the weight of the reconstruction sample may be the same as the content shown in FIG. 8.

As another example, the encoding apparatus may configure a weight table based on the weight candidates for the weighted sum. In this case, the weight table may be as shown in Table 3 above. The specific candidate, which is one of the candidates of the weight table, may be determined as the weight of the prediction sample and the weight of the reconstruction sample, and an index indicating the specific candidate may be generated. The index may be called a weighted index.

Meanwhile, when the prediction target block has bipredicted motion information, the prediction sample is derived based on one of the L0 motion vector and the L1 motion vector of the prediction target block, and the L0 motion vector and the L1. When the prediction sample is derived based on the motion vector, the weight value may be different.

The encoding apparatus encodes and outputs information on the filtering performed on the reconstructed sample (S950). The encoding apparatus may entropy-encode the information about the filtering indicating the filtering performed on the reconstructed sample in the filtering target region of the filtering target block based on the motion vector of the prediction target block and output the result in a bitstream form. In addition, the encoding apparatus may generate a flag indicating whether to perform the filtering on the reconstructed sample, encode it, and output the flag in the form of the bitstream. For example, the flag may be called a block-boundary filtering (bbf) flag. For example, when the filtering on the reconstructed sample is performed, the value of the bbf flag may be 1, and when the filtering on the reconstructed sample is not performed, the value of the bbf flag may be 0. FIG.

In addition, the encoding apparatus may generate an index indicating a candidate in the weight table, encode it, and output the index in the form of the bitstream. For example, the index may be called a weighted index.

The bbf flag and the weight index may be transmitted to the decoding apparatus in the form of the bitstream. The bitstream may be transmitted to a decoding apparatus via a network or a storage medium.

10 schematically illustrates a video decoding method by a decoding apparatus according to the present invention. The method disclosed in FIG. 10 may be performed by the decoding apparatus disclosed in FIG. 2. Specifically, for example, S1000 to S1010 of FIG. 10 may be performed by the prediction unit of the decoding apparatus, and S1020 to S1040 may be performed by the filter unit of the decoding apparatus.

The decoding apparatus derives a motion vector of the prediction target block (S1000). The motion vector of the prediction target block may be included in the motion information of the prediction target block. The motion information of the prediction target block may be bi-predicted motion information or may be L0 or L1 predicted motion information. The L0 or L1 predicted motion information may also be referred to as uni-predicted motion information. The bi-predicted motion information may include an L0 motion vector and an L1 motion vector, and the unipredicted motion information may include only one of the L0 motion vector and the L1 motion vector. The L0 motion vector represents a motion vector relating to L0 prediction, and the L1 motion vector represents a motion vector relating to L1 prediction. L0 represents a reference picture list L0 (list 0), and L1 represents a reference picture list L1 (list 1). In detail, when the motion information of the prediction target block is bi-predictive motion information, the L0 reference picture index, the L0 motion vector, the L1 reference picture index, and the L1 motion vector may be included. When the motion information of the prediction block is unipredicted motion information, the L0 reference picture index and the L0 motion vector may be included, or the L1 reference picture index and the L1 motion vector may be included.

The decoding apparatus derives a prediction sample for the prediction target block based on the motion vector (S1010). The decoding apparatus may obtain a predictive sample value on the reference picture indicated by the motion vector, and generate the predictive sample.

On the other hand, when the prediction target block has bi-predicted motion information and the motion vector of the prediction target block includes an L0 motion vector and an L1 motion vector, the decoding apparatus filters the neighboring block that has already been decoded. The motion vector for deriving the prediction sample may be different by comparing the motion vector of the filtering target block candidate with the L0 motion vector and the L1 motion vector of the prediction target block.

The decoding apparatus derives a filtering target block based on neighboring blocks of the prediction target block (S1020). The filtering target block may be at least one of a left neighboring block, a left upper neighboring block, and an upper neighboring block of the prediction target block. In addition, the neighboring block may have the same size as or different from the prediction target block. For example, the neighboring block may be smaller in size than the prediction target block.

In addition, the decoding apparatus may derive the decoded neighboring block among the neighboring blocks of the prediction target block as a filtering target block candidate, and compare the motion vector of the filtering target block candidate and the motion vector of the prediction target block. have. When the motion vector of the filtering target block candidate and the motion vector of the prediction target block are not the same, the filtering target block candidate may be derived as the filtering target block.

The decoding apparatus derives a filtering target region in the filtering target block (S1030). The filtering target region may include reconstructed samples around a reference boundary. The reference boundary may represent a boundary where a boundary of the prediction target block and a boundary of the filtering target block overlap. In detail, as an example, the filtering target region may be derived based on a difference value between the value of the motion vector of the filtering target block and the value of the motion vector of the prediction target block.

As another example, the filtering target region may be derived based on the size of the prediction target block, may be derived based on the size of the filtering target block, and may be derived from the size of the prediction target block and the size of the filtering target block. Can be derived based on this. The filtering target region may be as shown in Table 1 above.

As another example, the filtering target region may be derived based on whether an intra prediction mode is applied to the filtering target block. When the intra prediction mode is applied to the filtering target block, the filtering target region may be fixed to a preset range regardless of the size of the filtering target block.

As another example, the filtering target region may be derived based on the presence or absence of a quantization coefficient of the filtering target block.

The decoding apparatus performs filtering on the reconstructed samples in the filtering target region based on the prediction samples of the prediction target block (S1040). The decoding apparatus may perform filtering based on a weighted sum of the value of the prediction sample and the value of the reconstructed sample. In this case, the weight of the prediction sample and the weight of the reconstruction sample may be derived based on a distance from a reference boundary to the reconstruction sample. The reference boundary may represent a boundary where a boundary of the prediction target block and a boundary of the filtering target block overlap.

As another example, the decoding apparatus may configure a weight table based on the weight candidates for the weighted sum. The weight table may be as shown in Table 3 above. In this case, the decoding apparatus may obtain index information indicating one of the candidates included in the weight table through the bitstream, and the candidate indicating the weight for the prediction sample and the weight for the reconstruction sample is indicated by the index. Can be derived based on The index may be called a weighted index.

Meanwhile, the decoding apparatus may obtain flag information indicating whether to perform filtering on the reconstructed sample based on the motion vector of the prediction target block through a bitstream. The flag may be referred to as a block-boundary filtering (bbf) flag. For example, when the value of the bbf flag is 1, the decoding apparatus may perform filtering on the reconstructed sample. When the value of the bbf flag is 0, the decoding device may not perform filtering on the reconstructed sample.

According to the present invention described above, subjective / objective picture quality may be improved by filtering reconstructed samples in neighboring blocks of the prediction target block.

In addition, according to the present invention, filtering can be performed by weighted sum of neighboring reconstructed samples at the boundary of the predicted sample and the predicted block derived based on the motion vector of the predicted block, thereby reducing errors between blocks. As a result, subjective / objective image quality can be improved.

In the above-described embodiment, the methods are described based on a flowchart as a series of steps or blocks, but the present invention is not limited to the order of steps, and any steps may occur in a different order or simultaneously from other steps as described above. have. In addition, those skilled in the art will appreciate that the steps shown in the flowcharts are not exclusive and that other steps may be included or one or more steps in the flowcharts may be deleted without affecting the scope of the present invention.

The above-described method according to the present invention may be implemented in software, and the encoding device and / or the decoding device according to the present invention may perform image processing of, for example, a TV, a computer, a smartphone, a set-top box, a display device, and the like. It can be included in the device.

When embodiments of the present invention are implemented in software, the above-described method may be implemented as a module (process, function, etc.) for performing the above-described function. The module may be stored in memory and executed by a processor. The memory may be internal or external to the processor and may be coupled to the processor by various well known means. The processor may include application-specific integrated circuits (ASICs), other chipsets, logic circuits, and / or data processing devices. The memory may include read-only memory (ROM), random access memory (RAM), flash memory, memory card, storage medium and / or other storage device.

Claims

In the filtering method performed by the decoding apparatus,

Deriving a motion vector of the predicted block;

Deriving a prediction sample for the prediction target block based on the motion vector;

Deriving a filtering target block based on neighboring blocks of the prediction target block;

Deriving a filtering target region in the filtering target block; And

And filtering the reconstructed sample in the filtering target region based on the predicted sample of the predicted target block.
The method of claim 1,

The filtering of the reconstructed samples in the filtering target region based on the predicted samples in the prediction target block may include:

And performing filtering based on a weighted sum of the value of the prediction sample and the value of the reconstructed sample.
The method of claim 2,

In applying the weighted sum, a weight for the prediction sample and a weight for the reconstruction sample are derived based on a distance from a reference boundary to the reconstruction sample,

The reference boundary is a filtering method, characterized in that the boundary of the boundary of the prediction target block and the boundary of the filtering target block overlap.
The method of claim 2,

Constructing weight candidates for the weighted sum; And

Obtaining an index indicating one of the weighting candidates through a bitstream,

And the weight for the prediction sample and the weight for the reconstruction sample are derived based on the candidate indicated by the index.
The method of claim 1,

The filtering object block may be at least one of a left neighboring block, a left upper neighboring block, and an upper neighboring block of the prediction target block.
The method of claim 1,

And if there is no quantization coefficient of a neighboring block among the neighboring blocks, the corresponding neighboring block is not derived as the filtering target block.
The method of claim 1,

Deriving the filtering target block based on the neighboring blocks of the prediction target block,

Deriving an already decoded neighboring block among the neighboring blocks of the prediction target block as a filtering target block candidate;

Comparing the motion vector of the filtering target block candidate and the motion vector of the prediction target block; And

And when the motion vector of the filtering target block candidate and the motion vector of the prediction target block are not the same, deriving the filtering target block candidate as the filtering target block.
The method of claim 1,

Receiving a block-boundary filtering (BBF) flag via the bitstream; And

The method may further include determining whether to perform filtering on the restored sample.

And when the value of the BBF flag is 1, it is determined that filtering is performed on the reconstructed sample.
The method of claim 1,

When the prediction target block has bi-predicted motion information, the motion vector of the prediction target block includes an L0 motion vector and an L1 motion vector,

Deriving the filtering target block based on the neighboring blocks of the prediction target block,

Deriving an already decoded neighboring block among the neighboring blocks of the prediction target block as a filtering target block candidate;

Comparing the motion vector of the filtering target block candidate with the L0 motion vector and the L1 motion vector of the prediction target block; And

When the motion vector of the filtering target block candidate is not the same as the L0 motion vector and the L1 motion vector of the prediction target block, further comprising deriving the filtering target block candidate as the filtering target block. Filtering method.
The method of claim 9,

If the motion vector of the filtering target block candidate is the same as the L0 motion vector and the motion vector of the filtering target block candidate is not the same as the L1 motion vector, the prediction of the prediction block based on the L1 motion vector. Derive a predictive sample,

Deriving a prediction sample of the prediction target block based on the L0 motion vector and the L1 motion vector when the motion vector of the filtering target block candidate is not the same as both the L0 motion vector and the L1 motion vector. Characterized by a filtering method.
The method of claim 7, wherein

And when an intra prediction mode is applied to the filtering target block candidate, the filtering target block candidate is derived as a filtering target block.
The method of claim 7, wherein

And the filtering target region is derived based on a difference value between the value of the motion vector of the filtering target block candidate and the value of the motion vector of the prediction target block.
The method of claim 1,

The filtering method region is determined based on the size of the prediction target block and the size of the filtering target block.
The method of claim 1,

The filtering target region is derived based on whether an intra prediction mode is applied to the filtering target block.

And when the intra prediction mode is applied to the filtering target block, the filtering target region is derived to a preset range.
In the filtering method performed by the encoding device,

Deriving a motion vector of the predicted block;

Deriving a prediction sample for the prediction target block based on the motion vector;

Deriving a filtering target block based on neighboring blocks of the prediction target block;

Determining a filtering target region in the filtering target block;

Performing filtering on reconstructed samples in the filtering target region based on predicted samples of the predicted block; And

And encoding and outputting information on the filtering performed on the reconstructed sample.