WO2016072732A1

WO2016072732A1 - Method and apparatus for encoding/decoding video using texture synthesis based prediction mode

Info

Publication number: WO2016072732A1
Application number: PCT/KR2015/011771
Authority: WO
Inventors: 알쉬나엘레나; 알쉰알렉산더
Original assignee: 삼성전자 주식회사
Priority date: 2014-11-04
Filing date: 2015-11-04
Publication date: 2016-05-12

Abstract

Provided is a block based image prediction method comprising the steps of: determining a first prediction value of a current sample included in a current block, on the basis of a reference sample of a reference block to which the current block refers; determining the change rate of the first prediction value on the basis of peripheral samples of the reference sample; determining a texture parameter used for adjusting the change rate; and determining a second prediction value of the current sample using the first prediction value, the change rate, the texture parameter, and the temporal distance between a reference picture to which the reference block belongs, and a current picture to which the current block belongs.

Description

Method and apparatus for video encoding / decoding using texture synthesis based prediction mode

The present invention relates to a video encoding method and a decoding method. Specifically, the present invention relates to a method of encoding and decoding using texture synthesis based prediction mode.

With the development and dissemination of hardware capable of playing and storing high resolution or high definition video content, there is an increasing need for a video codec for efficiently encoding or decoding high resolution or high definition video content. According to the existing video codec, video is encoded according to a limited encoding method based on coding units having a tree structure.

Image data in the spatial domain is transformed into coefficients in the frequency domain using frequency transformation. The video codec divides an image into blocks having a predetermined size for fast operation of frequency conversion, performs DCT conversion for each block, and encodes frequency coefficients in units of blocks. Compared to the image data of the spatial domain, the coefficients of the frequency domain are easily compressed. In particular, since the image pixel value of the spatial domain is expressed as a prediction error through inter prediction or intra prediction of the video codec, when frequency conversion is performed on the prediction error, much data may be converted to zero. By substituting data repeatedly generated repeatedly with small size data, the data amount of the image can be reduced.

When a texture that changes according to a certain law is included in the image, the amount of change in the texture may be predicted to improve the prediction accuracy of the encoding / decoding apparatus.

Determining a first prediction value of a current sample included in the current block based on a reference sample of a reference block referenced by the current block, based on neighboring samples of the reference sample, change of the first prediction value Determining a degree, determining a texture parameter used to adjust the gradient, and a reference picture to which the first prediction value, the gradient, the texture parameter, and the reference block belong, and the current block A block-based image prediction method is provided that includes determining a second prediction value of the current sample by using a temporal distance of a current picture to which the current picture belongs.

The block-based image prediction method may further include obtaining texture synthesis based prediction information indicating whether texture synthesis based prediction is applied to the current block, and applying texture synthesis based prediction to the texture block based on the texture synthesis based prediction information. In this case, the change degree determining step, the texture parameter information determining step, and the second prediction value determining step may be performed.

In the block-based image prediction method, the current block may be a prediction block, and the texture synthesis based prediction information obtaining may include obtaining texture synthesis based prediction information applied only to the current block.

In the block-based image prediction method, the current block may be a prediction block, and the texture synthesis based prediction information obtaining may include obtaining texture synthesis based prediction information applied to a coding block including the current block.

The gradient may include gradient components along a predetermined number of directions, and the texture parameter information includes the predetermined number of texture parameters corresponding to the gradient components, and the predetermined number is two or more. Can be. The method of claim 5,

The degree of change includes a vertical degree of change determined based on peripheral samples located in a vertical direction of a reference sample and a horizontal degree of change determined based on peripheral samples located in a horizontal direction of a reference sample, and the texture parameter information May include a vertical texture parameter used to adjust the vertical gradient and a horizontal texture parameter used to adjust the horizontal gradient.

The determining of the degree of change may include applying an interpolation filter to the surrounding samples of the reference sample.

The determining of the second prediction value may include: adjusting the degree of change according to the texture parameter, determining a change amount of the first prediction value according to the adjusted change degree and the temporal distance, and the first prediction value. And determining the second prediction value according to the change amount of the first prediction value.

The determining of the texture parameter information may include determining the texture parameter information according to at least one of a size of the current block, a quantization parameter applied to the current block, and a slice type of the current block.

A first prediction value determiner configured to determine a first prediction value of a current sample included in the current block, based on a reference sample of a reference block referenced by the current block, based on neighboring samples of the reference sample, A gradient determining unit that determines a degree of change of a predicted value, a texture parameter determiner that determines a texture parameter used to adjust the degree of change, and the first predicted value, the gradient, the texture parameter, and the reference block Provided is a block-based image prediction apparatus including a second prediction value determiner configured to determine a second prediction value of the current sample using a temporal distance between a reference picture to which the current picture belongs and a current picture to which the current block belongs.

A video decoding method and a video encoding method are provided, wherein the image is predicted according to the provided block-based image prediction method, and the video is decoded according to the prediction result of the image.

Provided are a video decoding apparatus and a video encoding apparatus, which decode a video according to an image prediction result according to the provided block-based image prediction apparatus.

A computer-readable recording medium having recorded thereon a program for implementing the provided video decoding method and video encoding method is provided.

The encoding efficiency is improved by increasing the prediction accuracy of the image and the blocks included in the image.

1A is a block diagram of a video encoding apparatus based on coding units having a tree structure, according to an embodiment.

1B is a block diagram of a video decoding apparatus based on coding units having a tree structure, according to an embodiment.

2 illustrates a concept of coding units, according to an embodiment.

3A is a block diagram of an image encoder based on coding units, according to an embodiment.

3B is a block diagram of an image decoder based on coding units, according to an embodiment.

4 is a diagram of deeper coding units according to depths, and partitions, according to an embodiment.

5 illustrates a relationship between a coding unit and transformation units, according to an embodiment.

6 is a diagram of deeper encoding information according to depths, according to an embodiment.

7 is a diagram of deeper coding units according to depths, according to an exemplary embodiment.

8, 9, and 10 illustrate a relationship between a coding unit, a prediction unit, and a transformation unit, according to an embodiment.

11 illustrates a relationship between a coding unit, a prediction unit, and a transformation unit, according to encoding mode information.

12A illustrates a block diagram of a block-based image prediction apparatus, according to an embodiment.

12B is a flowchart of a block-based image prediction method, according to an embodiment.

13 is a diagram for describing an inter prediction mode, according to an exemplary embodiment.

14A and 14B are diagrams for describing a method of determining a degree of change of a current sample, according to an exemplary embodiment.

15 is a flowchart illustrating a method of encoding / decoding a current block according to a block-based image prediction method according to an embodiment.

Embodiments provided in the present invention may be modified and implemented within a range predictable by those skilled in the art based on the description provided by the specification and drawings.

In various embodiments described herein below, 'image' may refer to a generic image including a still image as well as a video such as a video. In addition, the term 'picture' described in the present specification means a still image to be encoded or decoded.

Hereinafter, "sample" means data to be processed as data allocated to a sampling position of an image. For example, the pixels in the spatial domain image may be samples.

Hereinafter, the 'current picture' means a picture to be currently encoded / decoded. Hereinafter, the 'current block' refers to a block that is the current encoding / decoding target among the blocks included in the current picture in the block-based image encoding / decoding method. Hereinafter, the 'current sample' refers to a sample that is currently subjected to encoding / decoding among samples included in the 'current block'.

Hereinafter, the 'inter prediction mode' refers to a prediction mode for predicting the current block from samples of pictures other than the current picture indicated by the motion information of the current block.

The coordinate (x, y) is defined below the sample located at the upper left corner of the block. Specifically, the coordinate of the sample located at the upper left corner of the block is determined as (0,0). The x value of the coordinate increases in the right direction, and the y value of the coordinate increases in the downward direction.

The present specification is largely divided into two parts. First, a block-based image prediction method will be described with reference to FIGS. 1A through 11. And the texture synthesis based prediction method is described in Figure 12a to 15.

1A is a block diagram of a video encoding apparatus 100 based on coding units having a tree structure, according to various embodiments.

According to an embodiment, the video encoding apparatus 100 that includes video prediction based on coding units having a tree structure includes an encoder 110 and an output unit 120. For convenience of description below, the video encoding apparatus 100 that includes video prediction based on coding units having a tree structure, according to an embodiment, is abbreviated as “video encoding apparatus 100”.

The encoder 110 may partition the current picture based on a maximum coding unit that is a coding unit of a maximum size for the current picture of the image. If the current picture is larger than the maximum coding unit, image data of the current picture may be split into at least one maximum coding unit. The maximum coding unit according to an embodiment may be a data unit having a size of 32x32, 64x64, 128x128, 256x256, or the like, and may be a square data unit having a square of two horizontal and vertical sizes.

The coding unit according to an embodiment may be characterized by a maximum size and depth. The depth indicates the number of times the coding unit is spatially divided from the maximum coding unit, and as the depth increases, the coding unit for each depth may be split from the maximum coding unit to the minimum coding unit. The depth of the largest coding unit is the highest depth and the minimum coding unit may be defined as the lowest coding unit. As the maximum coding unit decreases as the depth increases, the size of the coding unit for each depth decreases, and thus, the coding unit of the higher depth may include coding units of a plurality of lower depths.

As described above, the image data of the current picture may be divided into maximum coding units according to the maximum size of the coding unit, and each maximum coding unit may include coding units divided by depths. Since the maximum coding unit is divided according to depths, image data of a spatial domain included in the maximum coding unit may be hierarchically classified according to depths.

The maximum depth and the maximum size of the coding unit that limit the total number of times of hierarchically dividing the height and the width of the maximum coding unit may be preset.

The encoder 110 encodes at least one divided region obtained by dividing the region of the largest coding unit for each depth, and determines a depth at which the final encoding result is output for each of the at least one divided region. That is, the encoder 110 determines the encoding depth by selecting the depth at which the smallest encoding error occurs by encoding the image data in each coding unit according to depths in each maximum coding unit of the current picture. The determined coded depth and the image data for each maximum coding unit are output to the outputter 120.

Image data in the largest coding unit is encoded based on coding units according to depths according to at least one depth less than or equal to the maximum depth, and encoding results based on the coding units for each depth are compared. As a result of comparing the encoding error of the coding units according to depths, a depth having the smallest encoding error may be selected. At least one coding depth may be determined for each maximum coding unit.

As the depth of the maximum coding unit increases, the coding unit is divided into hierarchically and the number of coding units increases. In addition, even in the case of coding units having the same depth included in one largest coding unit, a coding error of each data is measured, and whether or not division into a lower depth is determined. Therefore, even in the data included in one largest coding unit, since the encoding error for each depth is different according to the position, the coding depth may be differently determined according to the position. Accordingly, one or more coding depths may be set for one maximum coding unit, and data of the maximum coding unit may be partitioned according to coding units of one or more coding depths.

Therefore, the encoder 110 according to an embodiment may determine coding units having a tree structure included in the current maximum coding unit. The coding units having a tree structure according to an embodiment include coding units having a depth determined as a coding depth among all deeper coding units included in the maximum coding unit. The coding unit of the coding depth may be hierarchically determined according to the depth in the same region within the maximum coding unit, and may be independently determined for the other regions. Similarly, the coded depth for the current region may be determined independently of the coded depth for the other region.

The maximum depth according to an embodiment is an index related to the number of divisions from the maximum coding unit to the minimum coding unit. The maximum depth according to an embodiment may represent the total number of splits from the maximum coding unit to the minimum coding unit. For example, when the depth of the largest coding unit is 0, the depth of the coding unit obtained by dividing the largest coding unit once may be set to 1, and the depth of the coding unit divided twice may be set to 2. In this case, if the coding unit divided four times from the maximum coding unit is the minimum coding unit, since depth levels of

depths

0, 1, 2, 3, and 4 exist, the maximum depth may be set to 4.

Predictive encoding and transformation of the largest coding unit may be performed. Similarly, prediction encoding and transformation are performed based on depth-wise coding units for each maximum coding unit and for each depth less than or equal to the maximum depth.

Since the number of coding units for each depth increases each time the maximum coding unit is divided for each depth, encoding including prediction encoding and transformation should be performed on all the coding units for each depth generated as the depth deepens. For convenience of explanation, the prediction encoding and the transformation will be described based on the coding unit of the current depth among at least one maximum coding unit.

The video encoding apparatus 100 according to an embodiment may variously select a size or shape of a data unit for encoding image data. The encoding of the image data is performed through prediction encoding, transforming, entropy encoding, and the like. The same data unit may be used in every step, or the data unit may be changed in steps.

For example, the video encoding apparatus 100 may select not only a coding unit for encoding the image data, but also a data unit different from the coding unit in order to perform predictive encoding of the image data in the coding unit.

For prediction encoding of the largest coding unit, prediction encoding may be performed based on a coding unit of a coding depth, that is, a more strange undivided coding unit, according to an embodiment. Hereinafter, a more strange undivided coding unit that is the basis of prediction coding is referred to as a 'prediction unit'. The partition in which the prediction unit is divided may include a data unit in which at least one of the prediction unit and the height and the width of the prediction unit are divided. The partition may be a data unit in which the prediction unit of the coding unit is split, and the prediction unit may be a partition having the same size as the coding unit.

For example, when a coding unit having a size of 2Nx2N (where N is a positive integer) is no longer split, it becomes a prediction unit of size 2Nx2N, and the size of a partition may be 2Nx2N, 2NxN, Nx2N, NxN, or the like. According to an embodiment, the partition type includes not only symmetric partitions in which the height or width of the prediction unit is divided by a symmetrical ratio, but also partitions divided in an asymmetrical ratio, such as 1: n or n: 1, by a geometric form. It may optionally include partitioned partitions, arbitrary types of partitions, and the like.

The prediction mode of the prediction unit may be at least one of an intra mode, an inter mode, and a skip mode. For example, the intra mode and the inter mode may be performed on partitions having sizes of 2N × 2N, 2N × N, N × 2N, and N × N. In addition, the skip mode may be performed only for partitions having a size of 2N × 2N. The encoding may be performed independently for each prediction unit within the coding unit to select a prediction mode having the smallest encoding error.

Also, the video encoding apparatus 100 according to an embodiment may perform conversion of image data of a coding unit based on not only a coding unit for encoding image data, but also a data unit different from the coding unit. In order to transform the coding unit, the transformation may be performed based on a transformation unit having a size smaller than or equal to the coding unit. For example, the transformation unit may include a data unit for intra mode and a transformation unit for inter mode.

In a similar manner to the coding unit according to the tree structure according to an embodiment, the transformation unit in the coding unit is also recursively divided into smaller transformation units, so that the residual data of the coding unit is determined according to the tree structure according to the transformation depth. Can be partitioned according to the conversion unit.

For a transform unit according to an embodiment, a transform depth indicating a number of divisions between the height and the width of the coding unit divided to the transform unit may be set. For example, if the size of the transform unit of the current coding unit of size 2Nx2N is 2Nx2N, the transform depth is 0, the transform depth 1 if the size of the transform unit is NxN, and the transform depth 2 if the size of the transform unit is N / 2xN / 2. Can be. That is, the transformation unit having a tree structure may also be set for the transformation unit according to the transformation depth.

The encoded information for each coded depth requires not only the coded depth but also prediction related information and transformation related information. Accordingly, the encoder 110 may determine not only the coded depth that generated the minimum encoding error, but also a partition type obtained by dividing the prediction unit into partitions, a prediction mode for each prediction unit, and a size of a transformation unit for transformation.

A method of determining a coding unit, a prediction unit / partition, and a transformation unit according to a tree structure of a maximum coding unit according to an embodiment will be described in detail with reference to FIGS. 8 to 24.

The encoder 110 may measure a coding error of coding units according to depths using a Lagrangian Multiplier-based rate-distortion optimization technique.

When the prediction unit or the prediction block is predicted by the inter prediction mode, the encoder 110 may apply the texture synthesis based prediction method 1250 of FIG. 12B.

The outputter 120 outputs the image data of the maximum coding unit encoded and the information about the encoding modes according to depths in the form of a bitstream based on the at least one coded depth determined by the encoder 110.

The encoded image data may be a result of encoding residual data of the image.

The information about the encoding modes according to depths may include encoding depth information, partition type information of a prediction unit, prediction mode information, size information of a transformation unit, and the like.

The coded depth information may be defined using depth-specific segmentation information indicating whether to encode to a coding unit of a lower depth without encoding to the current depth. If the current depth of the current coding unit is a coding depth, since the current coding unit is encoded in a coding unit of the current depth, split information of the current depth may be defined so that it is no longer divided into lower depths. On the contrary, if the current depth of the current coding unit is not the coding depth, encoding should be attempted using the coding unit of the lower depth, and thus split information of the current depth may be defined to be divided into coding units of the lower depth.

If the current depth is not the coded depth, encoding is performed on the coding unit divided into the coding units of the lower depth. Since at least one coding unit of a lower depth exists in the coding unit of the current depth, encoding may be repeatedly performed for each coding unit of each lower depth, and recursive coding may be performed for each coding unit of the same depth.

Since coding units having a tree structure are determined in one largest coding unit and information about at least one coding mode should be determined for each coding unit of a coding depth, information about at least one coding mode may be determined for one maximum coding unit. Can be. In addition, since the data of the largest coding unit is divided hierarchically according to the depth, the coding depth may be different for each location, and thus information about the coded depth and the coding mode may be set for the data.

Therefore, the outputter 120 according to an embodiment may allocate encoding information about a corresponding coding depth and an encoding mode to at least one of a coding unit, a prediction unit, and a minimum unit included in the maximum coding unit. .

The minimum unit according to an embodiment is a square data unit having a size obtained by dividing the minimum coding unit, which is the lowest coding depth, into four divisions. The minimum unit according to an embodiment may be a square data unit having a maximum size that may be included in all coding units, prediction units, partition units, and transformation units included in the maximum coding unit.

For example, the encoding information output through the output unit 120 may be classified into encoding information according to depth coding units and encoding information according to prediction units. The encoding information for each coding unit according to depth may include prediction mode information and partition size information. The encoding information transmitted for each prediction unit includes information about an estimation direction of the inter mode, information about a reference image index of the inter mode, information about a motion vector, information about a chroma component of an intra mode, and information about an inter mode of an intra mode. And the like.

Information about the maximum size and information about the maximum depth of a coding unit defined for each picture, slice segment, or GOP may be inserted into a header, a sequence parameter set, or a picture parameter set of a bitstream.

In addition, the information on the maximum size of the transform unit and the minimum size of the transform unit allowed for the current video may also be output through a header, a sequence parameter set, a picture parameter set, or the like of the bitstream. The output unit 120 may encode and output reference information related to prediction, prediction information, slice segment type information, and the like.

According to an embodiment of the simplest form of the video encoding apparatus 100, a coding unit according to depths is a coding unit having a size in which a height and a width of a coding unit of one layer higher depth are divided by half. That is, if the size of the coding unit of the current depth is 2Nx2N, the size of the coding unit of the lower depth is NxN. In addition, the current coding unit having a size of 2N × 2N may include up to four lower depth coding units having a size of N × N.

Accordingly, the video encoding apparatus 100 determines a coding unit having an optimal shape and size for each maximum coding unit based on the size and the maximum depth of the maximum coding unit determined in consideration of the characteristics of the current picture. Coding units may be configured. In addition, since each of the maximum coding units may be encoded in various prediction modes and transformation methods, an optimal coding mode may be determined in consideration of image characteristics of coding units having various image sizes.

Therefore, if an image having a very high resolution or a very large data amount is encoded in an existing macroblock unit, the number of macroblocks per picture is excessively increased. Accordingly, since the compressed information generated for each macroblock increases, the transmission burden of the compressed information increases, and the data compression efficiency tends to decrease. Therefore, the video encoding apparatus according to an embodiment may adjust the coding unit in consideration of the image characteristics while increasing the maximum size of the coding unit in consideration of the size of the image, thereby increasing image compression efficiency.

1B is a block diagram of a video decoding apparatus 150 based on coding units having a tree structure, according to various embodiments.

According to an embodiment, the video decoding apparatus 150 including video prediction based on coding units having a tree structure may include image data and encoding information reception extracting unit 160 and a decoding unit 170. For convenience of description below, the video decoding apparatus 150 that accompanies video prediction based on coding units having a tree structure according to an embodiment is referred to as a video decoding apparatus 150.

Definition of various terms such as a coding unit, a depth, a prediction unit, a transformation unit, and information about various encoding modes for a decoding operation of the video decoding apparatus 150 according to an embodiment may be described with reference to FIG. 8 and the video encoding apparatus 100. Same as described above with reference.

The reception extractor 160 receives and parses a bitstream of an encoded video. The image data and encoding information reception extractor 160 extracts the encoded image data for each coding unit from the parsed bitstream according to the coding units having the tree structure for each maximum coding unit, and outputs the encoded image data to the decoder 170. The image data and encoding information reception extracting unit 160 may extract information about a maximum size of a coding unit of the current picture from a header, a sequence parameter set, or a picture parameter set for the current picture.

Also, the image data and encoding information reception extracting unit 160 extracts information about a coding depth and an encoding mode of coding units having a tree structure for each largest coding unit, from the parsed bitstream. The extracted information about the coded depth and the coding mode is output to the decoder 170. That is, the image data of the bit string may be divided into the largest coding units so that the decoder 170 may decode the image data for each maximum coding unit.

The information about the coded depth and the encoding mode for each largest coding unit may be set with respect to one or more coded depth information, and the information about the coding mode according to the coded depths may include partition type information, prediction mode information, and transformation unit of the corresponding coding unit. May include size information and the like. In addition, split information for each depth may be extracted as the coded depth information.

The information about the coded depth and the encoding mode according to the maximum coding units extracted by the image data and the encoding information reception extractor 160 may be different according to the depths according to the maximum coding units in the encoding stage, as in the video encoding apparatus 100 according to an exemplary embodiment. Information about a coded depth and an encoding mode determined to repeatedly perform encoding for each coding unit to generate a minimum encoding error. Accordingly, the video decoding apparatus 150 may reconstruct an image by decoding data according to an encoding method that generates a minimum encoding error.

Since the encoded information about the coded depth and the encoding mode according to an embodiment may be allocated to a predetermined data unit among the corresponding coding unit, the prediction unit, and the minimum unit, the image data and the encoding information reception extracting unit 160 may be predetermined. Information about a coded depth and an encoding mode may be extracted for each data unit. If the information about the coded depth and the coding mode of the maximum coding unit is recorded for each of the predetermined data units, the predetermined data units having the information about the same coded depth and the coding mode are inferred as data units included in the same maximum coding unit. Can be.

The decoder 170 reconstructs the current picture by decoding image data of each maximum coding unit based on the information about the coded depth and the encoding mode for each maximum coding unit. That is, the decoder 170 may decode the encoded image data based on the read partition type, the prediction mode, and the transformation unit for each coding unit among the coding units having the tree structure included in the maximum coding unit. . The decoding process may include a prediction process including intra prediction and motion compensation, and an inverse transform process.

The decoder 170 may perform intra prediction or motion compensation according to each partition and prediction mode for each coding unit, based on partition type information and prediction mode information of the prediction unit of the coding unit for each coding depth.

In addition, the decoder 170 may read transform unit information having a tree structure for each coding unit and perform inverse transform based on the transformation unit for each coding unit, for the maximum transformation for each coding unit. Through inverse transformation, the pixel value of the spatial region of the coding unit may be reconstructed.

The decoder 170 may determine the coded depth of the current maximum coding unit by using the split information for each depth. If the split information indicates that the split information is no longer split at the current depth, the current depth is the coded depth. Therefore, the decoder 170 may decode the coding unit of the current depth with respect to the image data of the current maximum coding unit by using the partition type, the prediction mode, and the transformation unit size information of the prediction unit.

That is, by observing the encoding information set for a predetermined data unit among the coding unit, the prediction unit, and the minimum unit, the data units holding the encoding information including the same split information are gathered, and the decoding unit 170 performs the same encoding. It can be regarded as one data unit to be decoded in the mode. The decoding of the current coding unit may be performed by obtaining information about an encoding mode for each coding unit determined in this way.

The reception extractor 160 may acquire the SAO type and offset from the received current layer bitstream and determine the SAO category according to the distribution of sample values for each sample of the current layer prediction image. The offset for each SAO category can be obtained. Therefore, even if the prediction error is not received for each sample, the decoder 170 may compensate the offset for each category of each sample of the current hierarchical prediction image, and determine the current hierarchical reconstruction image by referring to the compensated current hierarchical prediction image. have.

When the prediction unit or the prediction block is predicted by the inter prediction mode, the decoder 170 may apply the texture synthesis based prediction method 1250 of FIG. 12B.

As a result, the video decoding apparatus 150 may obtain information about a coding unit that generates a minimum coding error by recursively encoding each maximum coding unit in the encoding process, and use the same to decode the current picture. That is, decoding of encoded image data of coding units having a tree structure determined as an optimal coding unit for each maximum coding unit can be performed.

Therefore, even if a high resolution image or an excessively large amount of data is used, the image data can be efficiently used according to the coding unit size and the encoding mode that are adaptively determined according to the characteristics of the image by using the information about the optimum encoding mode transmitted from the encoding end. Can be decoded and restored.

2 illustrates a concept of coding units, according to various embodiments.

As an example of a coding unit, a size of a coding unit may be expressed by a width x height, and may include 32x32, 16x16, and 8x8 from a coding unit having a size of 64x64. Coding units of size 64x64 may be partitioned into partitions of size 64x64, 64x32, 32x64, and 32x32, coding units of size 32x32 are partitions of size 32x32, 32x16, 16x32, and 16x16, and coding units of size 16x16 are 16x16. Coding units of size 8x8 may be divided into partitions of size 8x8, 8x4, 4x8, and 4x4, into partitions of 16x8, 8x16, and 8x8.

As for the video data 210, the resolution is set to 1920x1080, the maximum size of the coding unit is 64, and the maximum depth is 2. As for the video data 220, the resolution is set to 1920x1080, the maximum size of the coding unit is 64, and the maximum depth is 3. For the video data 230, the resolution is set to 352x288, the maximum size of the coding unit is 16, and the maximum depth is 1. The maximum depth illustrated in FIG. 8 represents the total number of divisions from the maximum coding unit to the minimum coding unit.

When the resolution is high or the amount of data is large, it is preferable that the maximum size of the coding size is relatively large not only to improve the coding efficiency but also to accurately shape the image characteristics. Therefore, the

video data

210 and 220 having higher resolution than the video data 230 may be selected to have a maximum size of 64.

Since the maximum depth of the video data 210 is 2, the coding unit 215 of the video data 210 is divided twice from the maximum coding unit having the long axis size of 64, and the depth is deepened by two layers, so the long axis size is 32, 16. Up to coding units may be included. On the other hand, since the maximum depth of the video data 230 is 1, the coding unit 235 of the video data 230 is divided once from coding units having a long axis size of 16, and the depth of the video data 230 is deepened by one layer. Up to coding units may be included.

Since the maximum depth of the video data 220 is 3, the coding unit 225 of the video data 220 is divided three times from the maximum coding unit having the long axis size of 64, and the depth is three layers deep, so that the long axis size is 32, 16. , Up to 8 coding units may be included. As the depth increases, the expressive power of the detailed information may be improved.

3A is a block diagram of an image encoder 300 based on coding units, according to various embodiments.

The image encoder 300 according to an embodiment includes operations performed by the encoder 210 of the video encoding apparatus 900 to encode image data. That is, the intra predictor 304 performs intra prediction on the coding unit of the intra mode of the current frame 302, and the motion estimator 306 and the motion compensator 308 perform the current frame 302 of the inter mode. And the inter frame estimation and the motion compensation using the reference frame 326. The motion compensator 308 may further apply a texture synthesis based prediction method to increase the accuracy of the prediction.

The data output from the intra predictor 304, the motion estimator 306, and the motion compensator 308 are output as quantized transform coefficients through the transformer 310 and the quantizer 312. The quantized transform coefficients are restored to the data of the spatial domain through the inverse quantizer 318 and the inverse transformer 320, and the recovered data of the spatial domain is passed through the deblocking unit 322 and the offset compensator 324. Processed and output to the reference frame 326. The quantized transform coefficients may be output to the bitstream 316 via the entropy encoder 314.

In order to be applied to the video encoding apparatus 900 according to an embodiment, the intra predictor 304, the motion estimator 306, the motion compensator 308, and the transform unit, which are components of the image encoder 300, may be used. 310, quantizer 312, entropy encoder 314, inverse quantizer 318, inverse transform unit 320, deblocking unit 322, and offset compensator 324 all have the maximum per maximum coding unit. In consideration of the depth, a task based on each coding unit among the coding units having a tree structure should be performed.

In particular, the intra predictor 304, the motion estimator 306, and the motion compensator 308 are partitions of each coding unit among coding units having a tree structure in consideration of the maximum size and the maximum depth of the current maximum coding unit. And a prediction mode, and the transform unit 310 should determine the size of a transform unit in each coding unit among the coding units having a tree structure.

3B is a block diagram of an image decoder 350 based on coding units, according to various embodiments.

The bitstream 352 is parsed through the parser 354 and the encoded image data to be decoded and information about encoding necessary for decoding are parsed. The encoded image data is output as inverse quantized data through the entropy decoding unit 356 and the inverse quantization unit 358, and the image data of the spatial domain is restored through the inverse transformation unit 360.

For the image data of the spatial domain, the intra predictor 362 performs intra prediction on the coding unit of the intra mode, and the motion compensator 364 uses the reference frame 370 together to apply the coding unit of the inter mode. Perform motion compensation for the The motion compensator 364 may further increase the accuracy of prediction by applying a texture synthesis based prediction method.

Data in the spatial domain that has passed through the intra predictor 362 and the motion compensator 364 may be post-processed through the deblocking 366 and the offset compensator 368 and output to the reconstructed frame 372. In addition, the post-processed data through the deblocking unit 366 and the loop filtering unit 368 may be output as the reference frame 370.

In order to decode the image data by the decoder 170 of the video decoding apparatus 350, step-by-step operations after the parser 354 of the image decoder 350 may be performed.

In order to be applied to the video decoding apparatus 950, a parser 354, an entropy decoder 356, an inverse quantizer 358, and an inverse transform unit 360, which are components of the image decoder 350, may be used. ), The intra predictor 362, the motion compensator 364, the deblocking part 366, and the offset compensator 368 must all perform operations based on coding units having a tree structure for each largest coding unit. do.

In particular, the intra predictor 362 and the motion compensator 364 determine partitions and prediction modes for each coding unit having a tree structure, and the inverse transform unit 360 must determine the size of the transform unit for each coding unit. .

4 is a diagram illustrating deeper coding units according to depths, and partitions, according to various embodiments.

The video encoding apparatus 100 according to an embodiment and the video decoding apparatus 150 according to an embodiment use hierarchical coding units to consider image characteristics. The maximum height, width, and maximum depth of the coding unit may be adaptively determined according to the characteristics of the image, and may be variously set according to a user's request. According to the maximum size of the preset coding unit, the size of the coding unit for each depth may be determined.

The hierarchical structure 400 of a coding unit according to an embodiment illustrates a case in which a maximum height and a width of a coding unit are 64 and a maximum depth is three. In this case, the maximum depth indicates the total number of divisions from the maximum coding unit to the minimum coding unit. Since the depth deepens along the vertical axis of the hierarchical structure 400 of the coding unit according to an embodiment, the height and the width of the coding unit for each depth are divided. In addition, a prediction unit and a partition on which the prediction coding of each depth-based coding unit is shown along the horizontal axis of the hierarchical structure 400 of the coding unit are illustrated.

That is, the coding unit 410 has a depth of 0 as the largest coding unit of the hierarchical structure 400 of the coding unit, and the size, that is, the height and the width of the coding unit, is 64x64. A depth deeper along the vertical axis includes a coding unit 420 of depth 1 having a size of 32x32, a coding unit 430 of depth 2 having a size of 16x16, and a coding unit 440 of depth 3 having a size of 8x8. A coding unit 440 of depth 3 having a size of 8 × 8 is a minimum coding unit.

Prediction units and partitions of the coding unit are arranged along the horizontal axis for each depth. That is, if a coding unit 410 having a size of 0 depth 64x64 is a prediction unit, the prediction unit includes a partition 410 having a size 64x64, a partition 412 having a size 64x32, and a size included in a coding unit 410 having a size 64x64. 32x64 partitions 414, 32x32 partitions 416.

Similarly, the prediction unit of the coding unit 420 having a size of 32x32 having a depth of 1 includes a partition 420 having a size of 32x32, partitions 422 having a size of 32x16, and a partition having a size of 16x32 included in the coding unit 420 having a size of 32x32. 424, partitions 426 of size 16x16.

Similarly, the prediction unit of the coding unit 430 of size 16x16 having a depth of 2 includes a partition 430 of size 16x16, partitions 432 of size 16x8, and a partition of size 8x16 included in the coding unit 430 of size 16x16. 434, partitions 436 of size 8x8.

Similarly, the prediction unit of the coding unit 440 of size 8x8 having a depth of 3 includes a partition 440 of size 8x8, partitions 442 of size 8x4 and a partition of size 4x8 included in the coding unit 440 of size 8x8. 444, partitions 446 of size 4x4.

The encoder 110 of the video encoding apparatus 100 according to an embodiment encodes each coding unit of each depth included in the maximum coding unit 410 to determine the coded depth of the maximum coding unit 410. Should be performed.

The number of deeper coding units according to depths for including data having the same range and size increases as the depth increases. For example, four coding units of depth 2 are required for data included in one coding unit of depth 1. Therefore, in order to compare the encoding results of the same data for each depth, each of the coding units having one depth 1 and four coding units having four depths 2 should be encoded.

For each depth coding, encoding is performed for each prediction unit of each coding unit along a horizontal axis of the hierarchical structure 400 of the coding unit, and a representative coding error, which is the smallest coding error at a corresponding depth, may be selected. . In addition, a depth deeper along the vertical axis of the hierarchical structure 400 of the coding unit, encoding may be performed for each depth, and the minimum coding error may be searched by comparing the representative coding error for each depth. The depth and the partition in which the minimum coding error occurs in the maximum coding unit 410 may be selected as the coding depth and the partition type of the maximum coding unit 410.

5 illustrates a relationship between a coding unit and transformation units, according to various embodiments.

The video encoding apparatus 100 according to an embodiment or the video decoding apparatus 150 according to an embodiment encodes or decodes an image in coding units having a size smaller than or equal to the maximum coding unit for each maximum coding unit. The size of a transformation unit for transformation in the encoding process may be selected based on a data unit that is not larger than each coding unit.

For example, in the video encoding apparatus 100 or the video decoding apparatus 150 according to the embodiment, when the current coding unit 510 is 64x64 size, the 32x32 size conversion unit 520 is selected. The conversion can be performed.

In addition, the data of the 64x64 coding unit 510 is transformed into 32x32, 16x16, 8x8, and 4x4 size transformation units of 64x64 size or less, and then encoded, and the transformation unit having the least error with the original is selected. Can be.

6 is a diagram of deeper encoding information according to depths, according to various embodiments.

The output unit 120 of the video encoding apparatus 100 according to an embodiment is information about an encoding mode, and information about a partition type 600 and information 610 about a prediction mode for each coding unit of each coded depth. In operation 620, information 620 about the size of the transform unit may be encoded and transmitted.

The information 600 about the partition type is a data unit for predictive encoding of the current coding unit and indicates information about a partition type in which the prediction unit of the current coding unit is divided. For example, the current coding unit CU_0 of size 2Nx2N is any one of a partition 602 of size 2Nx2N, a partition 604 of size 2NxN, a partition 606 of size Nx2N, and a partition 608 of size NxN. It can be divided and used. In this case, the information 600 about the partition type of the current coding unit represents one of a partition 602 of size 2Nx2N, a partition 604 of size 2NxN, a partition 606 of size Nx2N, and a partition 608 of size NxN. It is set to.

Information 610 relating to the prediction mode indicates the prediction mode of each partition. For example, through the information 610 on the prediction mode, whether the partition indicated by the information 600 on the partition type is performed in one of the intra mode 612, the inter mode 614, and the skip mode 616. Whether or not can be set.

In addition, the information about the transform unit size 620 indicates which transform unit to transform the current coding unit into. For example, the transform unit may be one of a first intra transform unit size 622, a second intra transform unit size 624, a first inter transform unit size 626, and a second inter transform unit size 628. have.

The reception extractor 160 of the video decoding apparatus 150 according to an embodiment may include information about partition type 600, information 610 about prediction mode, and transform unit size for each depth-decoding unit. The information 620 can be extracted and used for decoding.

7 is a diagram illustrating deeper coding units according to depths, according to various embodiments.

Segmentation information may be used to indicate a change in depth. The split information indicates whether a coding unit of a current depth is split into coding units of a lower depth.

The prediction unit 710 for predictive encoding of the coding unit 700 having depth 0 and 2N_0x2N_0 size includes a partition type 712 having a size of 2N_0x2N_0, a partition type 714 having a size of 2N_0xN_0, a partition type 716 having a size of N_0x2N_0, and a N_0xN_0 It may include a partition type 718 of size. Although only

partitions

712, 714, 716, and 718 in which the prediction unit is divided by a symmetrical ratio are illustrated, as described above, the partition type is not limited thereto, and asymmetric partitions, arbitrary partitions, geometric partitions, and the like. It may include.

For each partition type, prediction coding must be performed repeatedly for one 2N_0x2N_0 partition, two 2N_0xN_0 partitions, two N_0x2N_0 partitions, and four N_0xN_0 partitions. For partitions having a size 2N_0x2N_0, a size N_0x2N_0, a size 2N_0xN_0, and a size N_0xN_0, prediction encoding may be performed in an intra mode and an inter mode. The skip mode may be performed only for prediction encoding on partitions having a size of 2N_0x2N_0.

If the encoding error by one of the

partition types

712, 714, 716 of sizes 2N_0x2N_0, 2N_0xN_0, and N_0x2N_0 is the smallest, it is no longer necessary to divide it into lower depths.

If the encoding error due to the partition type 718 of size N_0xN_0 is the smallest, the depth 0 is changed to 1 and split (720), and the encoding is repeatedly performed on the depth 2 and the coding units 730 of the partition type of the size N_0xN_0. We can search for the minimum coding error.

The prediction unit 740 for prediction encoding of the coding unit 730 having a depth of 1 and a size of 2N_1x2N_1 (= N_0xN_0) includes a partition type 742 of size 2N_1x2N_1, a partition type 744 of size 2N_1xN_1, and a partition type of size N_1x2N_1. 746, a partition type 748 of size N_1 × N_1.

In addition, if the encoding error due to the partition type 748 having the size N_1xN_1 is the smallest, the depth 1 is changed to the depth 2 and divided (750), and repeatedly for the depth 2 and the coding units 760 of the size N_2xN_2. The encoding may be performed to search for a minimum encoding error.

When the maximum depth is d, depth-based coding units may be set until depth d-1, and split information may be set up to depth d-2. That is, when encoding is performed from the depth d-2 to the depth d-1 and the encoding is performed to the depth d-1, the prediction encoding of the coding unit 780 of the depth d-1 and the size 2N_ (d-1) x2N_ (d-1) The prediction unit 790 for is a partition type 792 of size 2N_ (d-1) x2N_ (d-1), partition type 794 of size 2N_ (d-1) xN_ (d-1), size A partition type 796 of N_ (d-1) x2N_ (d-1) and a partition type 798 of size N_ (d-1) xN_ (d-1) may be included.

Among the partition types, one partition 2N_ (d-1) x2N_ (d-1), two partitions 2N_ (d-1) xN_ (d-1), two sizes N_ (d-1) x2N_ Prediction encoding is repeatedly performed for each partition of (d-1) and four partitions of size N_ (d-1) xN_ (d-1), so that a partition type having a minimum encoding error may be searched. .

Even if the encoding error of the partition type 798 of size N_ (d-1) xN_ (d-1) is the smallest, the maximum depth is d, so the coding unit CU_ (d-1) of the depth d-1 is no longer The encoding depth of the current maximum coding unit 700 may be determined as the depth d-1, and the partition type may be determined as N_ (d-1) xN_ (d-1) without going through a division process into lower depths. In addition, since the maximum depth is d, split information is not set for the coding unit 752 having the depth d-1.

The data unit 799 may be referred to as a 'minimum unit' for the current maximum coding unit. According to an embodiment, the minimum unit may be a square data unit having a size obtained by dividing the minimum coding unit, which is the lowest coding depth, into four divisions. Through this iterative encoding process, the video encoding apparatus 100 compares the encoding errors for each depth of the coding unit 700, selects the depth at which the smallest encoding error occurs, and determines the encoding depth. The partition type and the prediction mode may be set to the encoding mode of the coded depth.

In this way, the depth with the smallest error can be determined by comparing the minimum coding errors for all depths of

depths

0, 1, ..., d-1, d, and can be determined as the coding depth. The coded depth, the partition type of the prediction unit, and the prediction mode may be encoded and transmitted as information about an encoding mode. In addition, since the coding unit must be split from the depth 0 to the coded depth, only the split information of the coded depth is set to '0', and the split information for each depth except the coded depth should be set to '1'.

The image data and encoding information reception extractor 160 of the video decoding apparatus 150 according to an embodiment extracts information about a coding depth and a prediction unit for the coding unit 700 to decode the coding unit 712. It is available. The video decoding apparatus 150 according to an embodiment may identify a depth having split information of '0' as an encoding depth by using split information according to depths, and use the decoding information by using information about an encoding mode for a corresponding depth. have.

8, 9, and 10 illustrate a relationship between coding units, prediction units, and transformation units, according to various embodiments.

The coding units 810 are coding units according to coding depths, which are determined by the video encoding apparatus 100 according to an embodiment with respect to the maximum coding unit. The prediction unit 860 is partitions of prediction units of each coding depth of each coding depth among the coding units 810, and the transformation unit 870 is transformation units of each coding depth for each coding depth.

If the depth-based coding units 810 have a depth of 0, the coding units 812 have a depth of 1, and the

coding units

814, 816, 818, 828, 850, and 852 have a depth of 2. The

coding units

820, 822, 824, 826, 830, 832, and 848 have a depth of three, and the

coding units

840, 842, 844, and 846 have a depth of four.

Some

partitions

814, 816, 822, 832, 848, 850, 852, and 854 of the prediction units 860 are obtained by splitting coding units. That is,

partitions

814, 822, 850, and 854 are partition types of 2NxN,

partitions

816, 848, and 852 are partition types of Nx2N, and partition 832 is partition types of NxN. The prediction units and partitions of the coding units 810 according to depths are smaller than or equal to each coding unit.

The image data of some 852 of the transformation units 870 is transformed or inversely transformed into a data unit having a smaller size than that of the coding unit. In addition, the

transformation units

814, 816, 822, 832, 848, 850, 852, and 854 are data units having different sizes or shapes when compared to corresponding prediction units and partitions among the prediction units 860. That is, even if the video encoding apparatus 100 and the video decoding apparatus 150 according to the embodiment are intra prediction / motion estimation / motion compensation operations and transform / inverse transform operations for the same coding unit, Each can be performed on a separate data unit.

Accordingly, coding is performed recursively for each coding unit having a hierarchical structure for each largest coding unit to determine an optimal coding unit. Thus, coding units having a recursive tree structure may be configured. The encoding information may include split information about a coding unit, partition type information, prediction mode information, and transformation unit size information.

The output unit 120 of the video encoding apparatus 100 according to an embodiment outputs encoding information about coding units having a tree structure, and the encoding information reception extraction unit of the video decoding apparatus 150 according to an embodiment. The 160 may extract encoding information about coding units having a tree structure from the received bitstream.

The split information indicates whether the current coding unit is split into coding units of a lower depth. If the split information of the current depth d is 0, partition type information, prediction mode, and transform unit size information are defined for the coded depth because the depth in which the current coding unit is no longer divided into the lower coding units is a coded depth. Can be. If it is to be further split by the split information, encoding should be performed independently for each coding unit of the divided four lower depths.

The prediction mode may be represented by one of an intra mode, an inter mode, and a skip mode. Intra mode and inter mode can be defined in all partition types, and skip mode can be defined only in partition type 2Nx2N.

The partition type information indicates the symmetric partition types 2Nx2N, 2NxN, Nx2N and NxN, in which the height or width of the prediction unit is divided by the symmetrical ratio, and the asymmetric partition types 2NxnU, 2NxnD, nLx2N, nRx2N, which are divided by the asymmetrical ratio. Can be. The asymmetric partition types 2NxnU and 2NxnD are divided into heights 1: 3 and 3: 1, respectively, and the asymmetric partition types nLx2N and nRx2N are divided into 1: 3 and 3: 1 widths, respectively.

The conversion unit size may be set to two kinds of sizes in the intra mode and two kinds of sizes in the inter mode. That is, if the transformation unit split information is 0, the size of the transformation unit is set to the size 2Nx2N of the current coding unit. If the transform unit split information is 1, a transform unit having a size obtained by dividing the current coding unit may be set. In addition, if the partition type for the current coding unit having a size of 2Nx2N is a symmetric partition type, the size of the transform unit may be set to NxN, and if the asymmetric partition type is N / 2xN / 2.

Encoding information of coding units having a tree structure according to an embodiment may be allocated to at least one of a coding unit, a prediction unit, and a minimum unit unit of a coding depth. The coding unit of the coding depth may include at least one prediction unit and at least one minimum unit having the same encoding information.

Therefore, if the encoding information held by each adjacent data unit is checked, it may be determined whether the adjacent data units are included in the coding unit having the same coding depth. In addition, since the coding unit of the corresponding coding depth may be identified by using the encoding information held by the data unit, the distribution of the coded depths within the maximum coding unit may be inferred.

Therefore, in this case, when the current coding unit is predicted with reference to the neighboring data unit, the encoding information of the data unit in the depth-specific coding unit adjacent to the current coding unit may be directly referenced and used.

In another embodiment, when the prediction coding is performed by referring to the neighboring coding unit, the data adjacent to the current coding unit in the coding unit according to depths is encoded by using the encoding information of the adjacent coding units according to depths. The neighboring coding unit may be referred to by searching.

FIG. 11 illustrates a relationship between coding units, prediction units, and transformation units, according to encoding mode information of Table 1. FIG.

The maximum coding unit 1100 includes

coding units

1102, 1104, 1106, 1112, 1114, 1116, and 1118 of a coded depth. Since one coding unit 1118 is a coding unit of a coded depth, split information may be set to zero. The partition type information of the coding unit 1118 of size 2Nx2N is partition type 2Nx2N 1122, 2NxN 1124, Nx2N (1126), NxN (1128), 2NxnU (1132), 2NxnD (1134), nLx2N (1136). And nRx2N 1138.

The transform unit split information (TU size flag) is a type of transform index, and a size of a transform unit corresponding to the transform index may be changed according to a prediction unit type or a partition type of a coding unit.

For example, if the partition type information is set to one of the symmetric partition types 2Nx2N 1122, 2NxN 1124, Nx2N 1126, and NxN (1128), and if the conversion unit partition information is 0, a conversion unit of size 2Nx2N ( 1142 is set, and if the transform unit split information is 1, a transform unit 1144 of size NxN may be set.

When the partition type information is set to one of the asymmetric partition types 2NxnU (1132), 2NxnD (1134), nLx2N (1136), and nRx2N (1138), if the conversion unit partition information (TU size flag) is 0, a conversion unit of size 2Nx2N ( 1152 is set, and if the transform unit split information is 1, a transform unit 1154 having a size N / 2 × N / 2 may be set.

Although the conversion unit splitting information (TU size flag) described above with reference to FIG. 5 is a flag having a value of 0 or 1, the conversion unit splitting information according to an embodiment is not limited to a 1-bit flag and is set according to a setting. , 1, 2, 3., etc., and may be divided hierarchically. The transformation unit partition information may be used as an embodiment of the transformation index.

In this case, when the transformation unit split information according to an embodiment is used together with the maximum size of the transformation unit and the minimum size of the transformation unit, the size of the transformation unit actually used may be expressed. The video encoding apparatus 100 according to an embodiment may encode maximum transform unit size information, minimum transform unit size information, and maximum transform unit split information. The encoded maximum transform unit size information, minimum transform unit size information, and maximum transform unit split information may be inserted into the SPS. The video decoding apparatus 150 according to an embodiment may use the maximum transform unit size information, the minimum transform unit size information, and the maximum transform unit split information to use for video decoding.

For example, (a) if the current coding unit is 64x64 in size and the maximum transform unit size is 32x32, (a-1) when the transform unit split information is 0, the size of the transform unit is 32x32, (a-2) When the split information is 1, the size of the transform unit may be set to 16 × 16, and (a-3) when the split unit information is 2, the size of the transform unit may be set to 8 × 8.

As another example, (b) if the current coding unit is size 32x32 and the minimum transform unit size is 32x32, (b-1) when the transform unit split information is 0, the size of the transform unit may be set to 32x32. Since the size cannot be smaller than 32x32, no further conversion unit split information can be set.

As another example, (c) if the current coding unit is 64x64 and the maximum transform unit split information is 1, the transform unit split information may be 0 or 1, and no other transform unit split information may be set.

Therefore, when the maximum transform unit split information is defined as 'MaxTransformSizeIndex', the minimum transform unit size is 'MinTransformSize', and the transform unit split information is 0, the minimum transform unit possible in the current coding unit is defined as 'RootTuSize'. The size 'CurrMinTuSize' can be defined as in relation (1) below.

CurrMinTuSize

= max (MinTransformSize, RootTuSize / (2 ^ MaxTransformSizeIndex)) ... (1)

Compared to the minimum transform unit size 'CurrMinTuSize' possible in the current coding unit, 'RootTuSize', which is a transform unit size when the transform unit split information is 0, may indicate a maximum transform unit size that can be adopted in the system. That is, according to relation (1), 'RootTuSize / (2 ^ MaxTransformSizeIndex)' is a transformation obtained by dividing 'RootTuSize', which is the size of the transformation unit when the transformation unit division information is 0, by the number of times corresponding to the maximum transformation unit division information. Since the unit size is 'MinTransformSize' is the minimum transform unit size, a smaller value among them may be the minimum transform unit size 'CurrMinTuSize' possible in the current coding unit.

According to an embodiment, the maximum transform unit size RootTuSize may vary depending on a prediction mode.

For example, if the current prediction mode is the inter mode, RootTuSize may be determined according to the following relation (2). In relation (2), 'MaxTransformSize' represents the maximum transform unit size and 'PUSize' represents the current prediction unit size.

RootTuSize = min (MaxTransformSize, PUSize) ......... (2)

That is, when the current prediction mode is the inter mode, 'RootTuSize', which is a transform unit size when the transform unit split information is 0, may be set to a smaller value among the maximum transform unit size and the current prediction unit size.

If the prediction mode of the current partition unit is a mode when the prediction mode is an intra mode, 'RootTuSize' may be determined according to Equation (3) below. 'PartitionSize' represents the size of the current partition unit.

RootTuSize = min (MaxTransformSize, PartitionSize) ........... (3)

That is, if the current prediction mode is the intra mode, the conversion unit size 'RootTuSize' when the conversion unit split information is 0 may be set to a smaller value among the maximum conversion unit size and the current partition unit size.

However, it should be noted that the current maximum conversion unit size 'RootTuSize' according to an embodiment that changes according to the prediction mode of the partition unit is only an embodiment, and a factor determining the current maximum conversion unit size is not limited thereto.

According to the video encoding method based on the coding units of the tree structure described above with reference to FIGS. 8 to 11, the image data of the spatial domain is encoded for each coding unit of the tree structure, and the video decoding method based on the coding units of the tree structure. As a result, decoding is performed for each largest coding unit, and image data of a spatial region may be reconstructed to reconstruct a picture and a video that is a picture sequence. The reconstructed video can be played back by a playback device, stored in a storage medium, or transmitted over a network.

12A to 15 illustrate an inter prediction mode according to a texture synthesis based prediction method.

The inter prediction mode is a prediction mode that predicts a current block from reconstructed samples included in a reconstructed picture before the current picture. According to the inter prediction mode, the current block is predicted according to reference samples indicated by a motion vector and a reference picture index corresponding to the current block. The motion vector represents a spatial difference between the current block and the reference samples referenced by the current block, and the reference picture index represents a unique number of the reference picture. When the value of the motion vector is an integer, reference samples included in the reference block are determined as prediction values of the current block. If the value of the motion vector is a fraction, the prediction block of the current block is determined by interpolating the reference samples included in the reference block.

In the inter prediction mode, the reference samples indicated by the motion vector and the reference picture index are generally copied or interpolated and used as the prediction value of the current block. However, if the current block includes a texture that changes according to a certain physical law, the prediction accuracy determined from the reference sample may be adjusted to increase the accuracy of the prediction. The process of adjusting the prediction value is called texture synthesis based prediction.

For example, if the current picture contains a reed forest swaying in the wind, the reeds are shaken in a regular pattern by the wind. This reed's movement can be physically represented by a partial differential equation. When the prediction values of the current block are adjusted according to the result predicted by the partial differential equation, the accuracy of the prediction values corresponding to the current block including the reed forest may be increased. Similarly, even when there are waves of the lake and leaves shaking caused by the wind, the prediction value adjustment due to the partial differential equation may be applied.

The partial differential equations used for the prediction of the current sample are represented in equation (1).

[Equation 1]

Is the degree of change of the predicted value for the current sample. Y _x means the degree of change in the horizontal direction of the current sample. Y _y means the degree of change in the vertical direction of the current sample. Y means the sample value of the reference sample referenced by the current sample. v _x , v _y , and w denote texture parameters applied to Y _x , Y _y , and Y, respectively.

The degree of change means the amount of change in the predicted value per hour. The degree of change may be determined according to a vertical degree of change indicating a degree of change in the vertical direction of the current sample and a horizontal degree of change indicating a degree of change in the horizontal direction. The vertical gradient is calculated based on the surrounding samples located in the vertical direction of the reference sample. The horizontal gradient is calculated based on the peripheral samples located in the horizontal direction of the reference sample. According to an embodiment, the sample value of the reference sample may be used to calculate the horizontal gradient and the vertical gradient.

The gradient may be determined by the sample value of the reference sample as well as the vertical gradient and the horizontal gradient. For example, when the predicted value of all the samples included in the current block increases or decreases at a constant rate, the sample value of the reference sample may be used for the gradient calculation. Therefore, the sample value of the reference sample is used to calculate the vertical gradient and horizontal gradient, which indirectly influences the determination of the gradient, while determining the gradient as an equivalent variable used to determine the gradient along with the vertical gradient and the horizontal gradient. Can have a direct impact on

Texture parameter indicates the weight (weight) given to Y _x, Y _y, Y in order to improve the accuracy of prediction. The texture parameter may have various values depending on the nature of the texture. According to an embodiment, the encoding apparatus determines the most efficient texture parameter and transmits the texture parameter to the decoding apparatus through a bitstream. According to another embodiment, the encoding apparatus and the decoding apparatus may share a predetermined rule for determining the texture parameter, and determine the texture parameter according to the predetermined rule. Therefore, the texture parameter may not be included in the bitstream. The predetermined rule may be determined according to the size of the current block, the quantization parameter applied to the current block, and the slice type of the current block.

The amount of change in the predicted value of the current sample determined according to Equation 1 is added to the predicted value of the current sample. As a result, the prediction value of the current sample can be accurate. In the following description, a prediction value of a current sample before texture synthesis based prediction is referred to as a first prediction value and a prediction value of a current sample after texture synthesis based prediction as a second prediction value.

Equation 1 is a partial differential equation applied in a case where the time domain and the space domain are continuous, and thus cannot be used in texture synthesis based prediction. Therefore, in the texture synthesis based prediction in which the time domain and the spatial domain are discrete, the texture synthesis based prediction is performed according to Equation 2 in which Equation 1 is modified.

[Equation 2]

In Equation 2, Pred _{i, j} represents a second prediction value, and MC [Ref _{i, j} ] represents a first prediction value. GRAD _x [Ref _{i, j} ], GRAD _y [Ref _{i, j} ] represent the horizontal gradient of the current sample and the vertical gradient of the current sample, respectively. Ref _{i, j} represents the sample value of the reference sample referenced by the current sample. v _x and v _y represent texture parameters applied to GRAD _x [Ref _{i, j} ] and GRAD _y [Ref _{i, j} ], respectively. w indicates the reference sample adjustment parameter applied to Ref _{i, j} . τ represents a temporal difference between the current picture and the reference picture. The temporal difference may be a value obtained by adjusting a frame difference between the current picture and the reference picture.

In equation (2), the degree of change is calculated in the same manner as in equation (1). To obtain the amount of change of the current sample according to the calculated degree of change, the degree of change is multiplied by τ, which represents a time difference between the current picture and the reference picture. However, when the texture parameter reflects the temporal difference between the current picture and the reference picture, τ is not multiplied. The second prediction value Pred _{i, j} may be obtained by adding the first prediction value MC [Pred _{i, j} ] to the change amount.

[Equation 3]

Equation 3 is a modification of Equation 2 and provides an embodiment of a method of calculating GRAD _x [Ref _{i, j} ] and GRAD _y [Ref _{i, j} ]. GRAD _x [Ref _{i, j} ], GRAD _y [Ref _{i, j} ] can be determined by samples located around the interpolation filter and the reference sample. In Equation 3, GRAD _x [Ref _{i, j} ] and GRAD _y [Ref _{i, j} ] are calculated using a 3-tap interpolation filter using [-1,0,1] as the filter coefficient.

The interpolation filter is applied to Y _{i + 1, j} and Y _{i-1, j} _, which are peripheral samples of the reference sample, to obtain a horizontal gradient. Therefore, GRAD _x [Ref _{i, j} ] representing the horizontal degree of change may be represented as (Yi + 1, j − Yi-1, j) / 2.

In a similar manner, the interpolation filter is applied to Y _{i, j + 1} and Y _{i, j-1,} which are peripheral samples of the reference sample. Accordingly, GRAD _y [Ref _{i, j} ] representing the vertical degree of change may be represented by (Y _{i, j + 1} -Y _{i, j-1} ) / 2.

[Equation 4]

Equation 4 shows the calculation process according to Equation 2 in detail. iVx, iVy, iW is v _x, v _y, represents an Integer value w after the scaling. In addition, iVx, iVy, and iW reflect the time difference between the current picture and the reference picture. And g _m represents the filter coefficient of the interpolation filter used to calculate the vertical gradient and the horizontal gradient. Ref _{i + m, j} means samples located in the horizontal direction of the reference sample, and Ref _{i, j + m} means samples located in the vertical direction of the reference sample. Δ _bts means the difference between the bit depth of MC [Ref _{i, j} ] and the bit depth for which calculation for texture synthesis based prediction is performed. Δ _b means the difference between the bit depth where the calculation for texture synthesis based prediction is performed and the bit depth of Pred _{i, j} .

_M (g _m Ref _{i + m, j} ) represents the horizontal gradient determined by the samples located in the horizontal direction of the interpolation filter and the reference sample. For example, when the filter used to calculate the horizontal gradient uses only three filter coefficients, m may have a value from -1 to 1. At this time, ∑ _m (g _m Ref _{i + m, j} ) has the same value as g ₋₁ Ref _{i-1, j} + g ₀ Ref _{i, j} + g ₁ Ref _{i + 1, j} .

_M (g _m Ref _{i, j + m} ) represents the vertical gradient determined by the samples located in the vertical direction of the interpolation filter and the reference sample. For example, when the filter used to calculate the vertical gradient uses only three filter coefficients, m may have a value from -1 to 1. At this time, ∑ _m (g _m Ref _{i + m, j} ) has the same value as g ₋₁ Ref _{i, j-1} + g ₀ Ref _{i, j} + g ₁ Ref _{i, j + 1} .

In Equation 4 iVx, iVy, because the iW reflects the temporal difference between the current picture and the reference picture, iVx, iVy, as determined in accordance with _{iW (- iVx Σ m (g} m Ref i + m, j) - iVy Σ m (g _m Ref _{i, j + m} )-iWRef _{i, j + m} ) represents the amount of change in the current sample.

A change amount is added to the first prediction value to obtain a second prediction value according to the change amount. A shift operation is performed to equalize the bit depths of the change amount, the first prediction value, and the second prediction value. First, a left shift operation is applied to MC [Ref _{i, j} ] representing the first predicted value by Δ _bts to match the bit depth of the first predicted value and the bit depth of the change amount. And after the predicted value obtained by adding the change amount to the first, second to apply the right-shift operation by the sum △ _b 1 of the predicted value and the change amount obtained second prediction value.

The texture synthesis based prediction method provided in Equations 1 to 4 may be modified and implemented by those skilled in the art without departing from the essential characteristics of the embodiments.

12A to 15, the above-described texture synthesis based prediction method is described in more detail.

12A is a block diagram of the block-based image prediction apparatus 1200 according to an embodiment. In detail, the block diagram of FIG. 12A illustrates an apparatus for performing an embodiment of a video decoding apparatus using an intra prediction mode.

The block-based image prediction apparatus 1200 may include a first prediction value determiner 1210, a gradient determiner 1220, a texture parameter determiner 1230, and a second prediction value determiner 1240. In FIG. 12A, the first prediction value determiner 1210, the change degree determiner 1220, the texture parameter determiner 1230, and the second prediction value determiner 1240 are represented by separate structural units. Accordingly, the intra first predictive value determiner 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second predictive value determiner 1240 may be combined to be implemented in the same component unit. Alternatively, the functions of the first prediction value determiner 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second prediction value determiner 1240 may be implemented in two or more component units. It may be.

In FIG. 12A, the first prediction value determiner 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second prediction value determiner 1240 are represented by a unit located in one device. The apparatuses in charge of the functions of the first predictive value determiner 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second predictive value determiner 1240 are not necessarily physically adjacent to each other. Accordingly, according to an embodiment, the first predictor 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second predictor determiner 1240 may be distributed.

The first predictor 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second predictor determiner 1240 of FIG. 12A may be implemented by one processor, according to an exemplary embodiment. have. In some embodiments, the present invention may also be implemented by a plurality of processors.

The block-based image prediction apparatus 1200 stores data generated by the first prediction value determiner 1210, the gradient determiner 1220, the texture parameter determiner 1230, and the second prediction value determiner 1240. And storage (not shown). In addition, the first prediction value determiner 1210, the change degree determiner 1220, the texture parameter determiner 1230, and the second prediction value determiner 1240 may extract and use data stored in a storage (not shown). .

The block-based image prediction apparatus 1200 of FIG. 12A is not limited to a physical apparatus. For example, some of the functions of the block-based image prediction apparatus 1200 may be implemented in software instead of hardware.

The first prediction value determiner 1210 determines a first prediction value of the current sample included in the current block based on the reference samples of the reference block referenced by the current block.

The first prediction value determiner 1210 may obtain a motion vector and a reference picture index of the current block from the bitstream. The first prediction value determiner 1210 determines reference samples referenced by samples of the current block based on the motion vector and the reference picture index. The first prediction value of the current sample is determined according to the determined reference samples.

13 illustrates a method of determining a first prediction value of a current sample in inter prediction mode. The picture on the right is the current picture 1300 and the picture on the left is the reference picture 1310. The current picture 1300 includes a current block 1302, and the reference picture 1310 includes a reference block 1312 to which the current block 1302 refers. The position of the reference block 1312 is determined by the upper left sample of the current block 1302 and the motion vector 1320. The value obtained by adding the motion vector 1320 to the position of the upper left sample of the current block 1302 indicates the position of the upper left sample of the reference block 1312.

If both the x component and the y component of the motion vector 1320 are integers, the values of the reference samples are copied to determine the first prediction value of the current samples. If at least one of the x component and the y component of the motion vector 1320 is a fraction, the interpolated value is determined as the first prediction value of the current samples using the reference samples and the interpolation filter.

The reference picture 1310 and the current picture 1300 have a Picture Order Counter (POC) meaning a frame number arranged in a playback order. The temporal difference between the reference picture 1310 and the current picture 1300 may be represented by a POC difference between the POC of the reference picture 1310 and the POC of the current picture 1300.

Τ 1330 of FIG. 13 may be obtained by multiplying the POC difference by a time difference between adjacent frames. The value τ 1330 can be used in the texture synthesis process. When the change in the gradient due to the value τ 1330 is reflected in the texture parameter, the value τ 1330 may not be used separately in the texture synthesis process.

The gradient determiner 1220 determines the gradient of the first prediction value based on the surrounding samples of the reference sample. The gradient is a value that indicates how much the current sample changes per hour. As the degree of change increases, the amount of change in the first prediction value due to the texture synthesis based prediction increases.

The gradient is determined according to the reference sample corresponding to the current sample and the surrounding samples located in the periphery of the reference sample. The peripheral sample of the reference sample may include peripheral samples located in the vertical direction or the horizontal direction from the reference sample. In addition, the peripheral sample of the reference sample may include samples located in a direction other than the vertical direction or the horizontal direction. For example, it may include samples located diagonally.

The gradient may include a vertical gradient and a horizontal gradient. Vertical gradient refers to gradients obtained from peripheral samples located in a vertical direction from a reference sample. If the sample value of the peripheral sample located at the bottom of the reference sample is larger than the sample value of the peripheral sample located at the top, the vertical gradient is positive. In the opposite case, the vertical gradient has a negative value. As the difference between the sample value of the peripheral sample located at the bottom of the reference sample and the sample value of the peripheral sample located at the top, the absolute value of the vertical gradient increases.

The horizontal gradient refers to the gradient obtained from peripheral samples located in the horizontal direction from the reference sample. When the sample value of the peripheral sample located to the right of the reference sample is larger than the sample value of the peripheral sample located to the left, the horizontal gradient has a positive value. In the opposite case, the horizontal gradient has a negative value. The absolute value of the horizontal gradient increases as the difference between the sample value of the peripheral sample positioned on the right side of the reference sample and the sample value of the peripheral sample positioned on the left side increases.

The gradient determining unit 1220 may determine the degree of change in other directions in addition to the vertical degree of change and the horizontal degree of change. For example, the degree of change in the diagonal direction can be determined.

The gradient determiner 1220 may apply an interpolation filter to the reference sample and the surrounding samples of the reference sample to calculate the horizontal gradient and the vertical gradient. 14A and 14B, the calculation method of the horizontal gradient and the vertical gradient according to the interpolation filter will be described.

FIG. 14A discloses a method for calculating horizontal gradient and vertical gradient using a filter tap of [-1, 0, 1]. In the case of using the 2-tap filter of [-1, 0, 1], according to the embodiment of Equation 3, the horizontal gradient is (Y _{i + 1, j} -Y _{i-1, j} ) / 2, which is a vertical change. The degree is determined as (Y _{i, j + 1} -Y _{i, j-1} ) / 2.

Reference block 1400 of FIG. 14A includes samples 1410-1428. For example, when the current sample points to the sample 1410 according to the motion vector of the current block, the samples 1412 and 1414 are used to obtain a horizontal gradient of the current sample.

Samples

1416 and 1418 are then used to find the vertical gradient of the current sample. The sample 1410 corresponding to the current sample in the two tap filter is not used.

Since the horizontal gradient is determined by (Y _{i + 1, j} -Y _{i-1, j} ) / 2, the horizontal gradient of the current sample is obtained by dividing the sample 1414 minus the sample 1412 by 2. Is determined. Thus 10 is determined as the horizontal gradient of the current sample. ((110-90) / 2 = 10)

Since the vertical gradient is determined by (Y _{i, j + 1} -Y _{i, j-1} ) / 2, the vertical gradient of the current sample by dividing the sample 1418 minus the sample 1416 by 2 Is determined. Thus, -15 is determined as the horizontal gradient of the current sample. ((95-125) / 2 = -15)

As another example, when the current sample points to the sample 1420 according to the motion vector of the current block, the samples 1422 and 1424 are used to obtain a horizontal gradient of the current sample.

Samples

1426 and 1428 are then used to find the vertical gradient of the current sample. Since the center filter coefficient of the two tap filter is 0, the sample 1420 corresponding to the current sample is not used.

Since the horizontal gradient is determined by (Yi + 1, j-Yi-1, j) / 2, the horizontal gradient of the current sample is determined by dividing the value of the sample 1424 minus the sample 1422 by 2. . Thus 10 is determined as the horizontal gradient of the current sample. ((160-140) / 2 = 10)

Since the vertical gradient is determined by (Yi, j + 1-Yi, j-1) / 2, the vertical gradient of the current sample is determined by dividing the sample 1428 minus the sample 1426 by two. . Thus, 14 is determined as the horizontal gradient of the current sample. ((160-132) / 2 = 14)

14B discloses a method for calculating horizontal gradient and vertical gradient using a six-tap filter with filter coefficients [6, -25, -2, 30, -12, 3].

When the current sample points to the sample 1460 according to the motion vector of the current block, the

samples

1460, 1462, 1464, 1466, 1468, and 1470 are used to obtain a horizontal gradient of the current sample. The

samples

1460, 1472, 1474, 1476, 1478, and 1480 are used to obtain a vertical gradient of the current sample.

The interpolation filter can be applied not only when the current sample is a luma sample but also when it is a chroma sample. According to an embodiment, a six tap filter may be used when the current sample is a luma sample, and a four tap filter may be used when the chroma sample is a chroma sample. FarcMV means a fractional part of the motion vector of the current block. For example, if farcMV is zero, there is no fractional part of the motion vector. If farcMV is 1/4, the motion vector has a fractional part of 1/4.

The following six tap filters can be used for the luma sample.

farcMV = 0: {6, -25, -2, 30, -12, 3}

farcMV = 1/4: {3, -10, -25, 40, -11, 3}

farcMV = 1/2: {-1, 4, -38, 38, -4, 1}

farcMV = 3/4: {-3, 11, -40, 25, 10, -3}

The following four tap filters can be used for chroma samples.

farcMV = 0: {-11, -3, 19, -5}

farcMV = 1/8: {-8, -10, 23, -5}

farcMV = 2/8: {-4, -17, 26, -5}

farcMV = 3/8: {-1, -21, 26, -4}

farcMV = 4/8: {2, -25, 25, -2}

farcMV = 5/8: {4, -26, 21, 1}

farcMV = 6/8: {5, -26, 17, 4}

farcMV = 7/8: {5, -23, 10, 8}

The number and specific values of the filter coefficients included in the interpolation filter may be changed according to embodiments.

The texture parameter determiner 1230 determines the texture parameter used to adjust the magnitude of the gradient. The texture parameter refers to a weight applied to the degree of change determined by the degree of change determiner 1220.

The texture parameter may include a horizontal texture parameter corresponding to the horizontal gradient and a vertical texture parameter corresponding to the vertical gradient. If the horizontal texture parameter is larger than the vertical texture parameter, the horizontal gradient has more influence on the determination of the gradient than the vertical gradient. If the vertical texture parameter is larger than the horizontal texture parameter, the vertical gradient has more influence on the determination of the gradient than the horizontal gradient.

The texture parameter determiner 1230 may include not only texture parameters but also reference sample adjustment parameters used to adjust the reference samples. Also, the texture parameter determiner 1230 may further determine other types of texture parameters in addition to the horizontal texture parameter and the vertical texture parameter when the gradient includes the horizontal gradient and the vertical gradient. have. For example, if there is a diagonal gradient, a diagonal texture parameter can be obtained.

The texture parameter determiner 1230 may obtain a texture parameter from the bitstream. According to an embodiment, the texture parameter may be newly acquired for each coding block or prediction block. According to another embodiment, the obtained texture parameter may be obtained at a higher data unit level of a coding block and applied to all prediction units of the higher data unit.

Instead of obtaining the texture parameter from the bitstream, the texture parameter determiner 1230 may determine the texture parameter according to at least one of the size of the current block, the quantization parameter applied to the current block, and the slice type of the current block.

For example, when the texture parameter is determined by the size of the current block, the horizontal texture parameter may be determined as Log2 (width of the current block) -1 and the vertical texture parameter may be determined as Log2 (height of the current block) -1.

For example, when the texture parameter is determined by the quantization parameter or the slice type of the current block, the texture parameter may be determined according to a statistical value determined according to the degree to which the quantization parameter and the slice type affect the image quality of the reconstructed image.

The texture parameter determiner 1230 may determine the texture parameter according to a temporal distance between the reference picture to which the reference block belongs and the current picture to which the current block belongs. By determining the texture parameter according to τ 1330 of FIG. 13, the amount of computation required in the texture synthesis based prediction process may be reduced.

The second prediction value determiner 1240 determines a second prediction value of the current sample by using a first prediction value, a degree of change, a texture parameter, and a temporal distance between a reference picture to which the reference block belongs and a current picture to which the current block belongs. do.

First, the second prediction value determiner 1240 adjusts the degree of change according to the texture parameter. The second prediction value determiner 1240 adjusts the degree of change by multiplying the degree of change determined by the change degree determiner 1220 with the texture parameter determined by the texture parameter determiner 1230.

According to an exemplary embodiment, the second prediction value determiner 1240 may adjust the vertical gradient by multiplying the vertical gradient by the vertical texture parameter. Also, the second prediction value determiner 1240 may adjust the horizontal gradient by multiplying the horizontal gradient and the horizontal texture parameter.

When the first prediction value is necessary for determining the degree of change, the second prediction value determiner 1240 may obtain the adjusted first prediction value by multiplying the first prediction value by the first prediction value adjustment parameter.

The second prediction value determiner 1240 may obtain the degree of change by adding the adjusted degree of vertical change and the adjusted degree of horizontal change. If the first prediction value is necessary for the degree of change calculation, the adjusted degree of change may be determined by adding up the adjusted vertical degree of change, the adjusted horizontal change, and the adjusted first prediction value.

After determining the adjusted degree of change, the second prediction value determiner 1240 determines the amount of change of the first predicted value according to the adjusted degree of change and the temporal distance. In detail, the second prediction value determiner 1240 may determine a change amount of the first prediction value by multiplying the adjusted change degree and the temporal distance. If the texture parameter reflects the temporal distance between the current picture and the reference picture, the adjusted change is determined as the change amount of the first prediction value.

Finally, the second prediction value determiner 1240 determines the second prediction value according to the change amount of the first prediction value and the first prediction value. In detail, the second prediction value determiner 1240 determines the second prediction value by adding the first prediction value and the amount of change.

According to an embodiment, the functions of the gradient determiner 1220, the texture parameter determiner 1230, and the second predictor determiner 1240 may indicate that texture synthesis based prediction information is applied to the texture block based on texture synthesis based prediction. Case may be performed.

The texture synthesis based prediction information indicates whether texture synthesis based prediction is applied to the current block. The texture synthesis based prediction information may be defined as a flag having a value of zero or one. For example, when texture synthesis based prediction information indicates 0, texture synthesis based prediction may not be applied to the current block. Conversely, when texture synthesis based prediction information indicates 1, texture synthesis based prediction may be applied to the current block.

According to an embodiment, texture synthesis based prediction information may exist for each prediction block. Therefore, texture synthesis based prediction information for each prediction block may be obtained from the bitstream.

According to another embodiment, texture synthesis based prediction information may exist for a coding block. At this time, texture synthesis based prediction information indicating whether texture synthesis based prediction is applied to the prediction blocks included in the coding block is obtained from the bitstream. Therefore, when texture synthesis based prediction information about a coding block indicates 1, texture synthesis based prediction may be applied to all prediction blocks included in the coding block.

According to an embodiment, texture synthesis based prediction information may be obtained for a data unit that is higher than a coding block. When the texture synthesis based prediction information indicates that texture synthesis based prediction is collectively applied to the data unit, texture synthesis based prediction is applied to all prediction blocks included in the data unit.

12B is a flowchart of a block-based image prediction method 1250 according to an embodiment. In detail, the flowchart of FIG. 12B illustrates an embodiment of a block-based image prediction method using a texture synthesis based prediction method.

In operation 12, the first prediction value of the current sample included in the current block is determined based on the reference sample of the reference block referenced by the current block.

In step 14, based on the surrounding samples of the reference sample, the degree of change of the first prediction value is determined. The gradient may include a vertical gradient determined based on the peripheral samples positioned in the vertical direction of the reference sample and a horizontal gradient determined based on the peripheral samples positioned in the horizontal direction of the reference sample.

The gradient in step 14 may be determined by applying an interpolation filter on the surrounding samples of the reference sample. The horizontal gradient may be determined by applying an interpolation filter to peripheral samples located in the horizontal direction. Similarly, the vertical gradient can be determined by applying an interpolation filter to the surrounding samples located in the vertical direction.

In step 16 the texture parameter used to adjust the gradient is determined. The texture parameter may include a vertical texture parameter used to adjust the vertical gradient and a horizontal texture parameter used to adjust the horizontal gradient.

According to one embodiment, in step 16 the texture parameter may be obtained from the bitstream. According to another embodiment, in step 16, the texture parameter information may be determined according to at least one of the size of the current block, the quantization parameter applied to the current block, and the slice type of the current block.

In operation 18, the second prediction value of the current sample is determined using the first prediction value, the gradient, the texture parameter, and the temporal distance between the reference picture to which the reference block belongs and the current picture to which the current block belongs. Specifically, the degree of change is adjusted according to the texture parameter, and the amount of change is determined according to the adjusted degree of change and the temporal distance. When the temporal distance is reflected in the texture parameter, the step of determining the amount of change according to the adjusted degree of change and the temporal distance is omitted. Finally, the second prediction value is obtained by adding the change amount and the first prediction value.

Obtaining texture synthesis based prediction information indicating whether texture synthesis based prediction is applied to the current block before step 14 may be included. When the texture synthesis based prediction information indicates that texture synthesis based prediction is applied to the current block, steps 14 to 18 may be performed on the current sample.

According to an embodiment, the obtained texture synthesis based prediction information may be obtained from a bitstream with respect to a prediction block. As another example, texture synthesis based prediction information may be obtained for a coding block and applied to all prediction blocks included in the coding block. As another example, texture synthesis based prediction information may be obtained with respect to a higher data unit of a coding block and applied to all prediction blocks included in the higher data unit.

The block-based image prediction method 1250 according to the above-described embodiment may be performed by the block-based image prediction apparatus 1200.

According to FIG. 15, an embodiment of a video encoding / decoding method using a texture synthesis based prediction method is provided.

In operation 1510, when the current block is predicted according to the inter prediction mode, a reference block to which the current block refers is determined. The reference block is determined by the motion vector and the reference picture index corresponding to the current block.

In operation 1520, a first prediction value is obtained for each sample of the current block from the reference block. A first prediction value is determined from the reference samples included in the reference block. If the fractional parts of the x and y components of the motion vector are both zero, the sample value of the reference sample is used as the first prediction value. However, when at least one of the x component and the y component of the motion vector has a fractional portion other than 0, the first prediction value is determined by interpolating the reference samples included in the reference block.

In operation 1530, it is determined whether the texture synthesis based prediction mode is applied to the current block. The current block may be a prediction block that is currently encoded / decoded. Whether the texture synthesis based prediction mode is applied may be determined by texture synthesis based prediction information. The texture synthesis based prediction information takes the form of a flag.

Texture synthesis based prediction information may be obtained from the bitstream for the current block. Alternatively, the texture synthesis based prediction information may be obtained with respect to a higher data unit of the current block.

If the texture synthesis based prediction information indicates that the texture synthesis based prediction mode is applied to the current block, step 1540 is performed. On the contrary, when the texture synthesis based prediction information indicates that the texture synthesis based prediction mode is not applied to the current block, no further prediction process is performed in step 1580, and encoding of the current block according to the first prediction value obtained in step 1520. / Decoding is performed.

In step 1540, a gradient for each sample of the current block is obtained from the reference block. Acquiring the gradient of the samples included in the current block may be performed by the gradient determiner 1220 of FIG. 12A.

In step 1550 a texture parameter used to adjust the gradient is obtained. The obtaining of the texture parameter of the samples included in the current block may be performed by the texture parameter determiner 1230 of FIG. 12A.

In operation 1560, the first prediction value, the degree of change, the texture parameter, and the temporal distance between the reference picture to which the reference block belongs and the current picture to which the current block belongs, become a second prediction value for each sample of the current block. The determining of the second prediction value of the samples included in the current block may be performed by the second prediction value determiner 1240 of FIG. 12A.

In operation 1570, the current block is encoded / decoded according to the second prediction value. In detail, in the encoding process, residual values may be obtained by obtaining difference values between an original value and a second prediction value of samples included in a current block, and transforming, quantizing, and entropy encoding the difference values. The encoding information used in the residual data and the texture synthesis based prediction mode may be transmitted to the decoding apparatus. In the decoding process, residual data may be obtained from a bitstream, and entropy decoding, inverse quantization, and inverse transformation of the residual data may obtain difference values between original values and second prediction values of samples included in a current block. The current block may be reconstructed based on the second prediction values and the difference values obtained in steps 1510 to 1560.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium. The computer-readable recording medium may include a storage medium such as a magnetic storage medium (eg, a ROM, a floppy disk, a hard disk, etc.) and an optical reading medium (eg, a CD-ROM, a DVD, etc.).

Those skilled in the art to which the various embodiments disclosed so far will understand that it can be implemented in a modified form without departing from the essential characteristics of the embodiments disclosed herein. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The disclosure scope of the present specification is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the scope of the disclosure.

Claims

Determining a first prediction value of a current sample included in the current block based on a reference sample of a reference block referenced by the current block;

Determining a degree of change of the first prediction value based on peripheral samples of the reference sample;

Determining a texture parameter used to adjust the gradient; And

Determining a second prediction value of the current sample by using the first prediction value, the gradient, the texture parameter, and a temporal distance between a reference picture to which the reference block belongs and a current picture to which the current block belongs. Block based image prediction method.
The method of claim 1,

Acquiring texture synthesis based prediction information indicating whether texture synthesis based prediction is applied to the current block;

When the texture synthesis based prediction information indicates that texture synthesis based prediction is applied to the current block, the block degree determining step, the texture parameter information determining step, and the second prediction value determining step are performed. Image Prediction Method.
The method of claim 2,

The current block is a prediction block,

The acquiring of texture synthesis based prediction information may include obtaining texture synthesis based prediction information applied only to the current block.
The method of claim 2,

The current block is a prediction block,

The acquiring of texture synthesis based prediction information may include obtaining texture synthesis based prediction information applied to a coding block including the current block.
The method of claim 1,

The gradient includes gradient components along a predetermined number of directions,

The texture parameter information includes the predetermined number of texture parameters corresponding to the gradient components,

The predetermined number is block-based image prediction method, characterized in that more than two.
The method of claim 5,

The degree of change includes a vertical degree of change determined based on peripheral samples located in a vertical direction of a reference sample and a horizontal degree of change determined based on peripheral samples located in a horizontal direction of a reference sample,

The texture parameter information includes a vertical texture parameter used to adjust the vertical gradient and a horizontal texture parameter used to adjust the horizontal gradient.
The method of claim 1,

Determining the degree of change,

And applying an interpolation filter on the neighboring samples of the reference sample.
The method of claim 1,

The second prediction value determining step,

Adjusting the degree of change in accordance with the texture parameter;

Determining a change amount of the first predicted value according to the adjusted change degree and the temporal distance; And

And determining the second prediction value according to the first prediction value and the amount of change of the first prediction value.
The method of claim 1,

Determining the texture parameter information,

And determining the texture parameter information according to at least one of a size of the current block, a quantization parameter applied to the current block, and a slice type of the current block.
A first prediction value determiner that determines a first prediction value of a current sample included in the current block based on a reference sample of a reference block referenced by the current block;

A change degree determiner which determines a change degree of the first prediction value based on surrounding samples of the reference sample;

A texture parameter determiner which determines a texture parameter used to adjust the degree of change; And

A second prediction value for determining a second prediction value of the current sample by using the first prediction value, the degree of change, the texture parameter, and a temporal distance between a reference picture to which the reference block belongs and a current picture to which the current block belongs Block-based image prediction device including a determination unit.
A video decoding method comprising predicting an image according to the block-based image prediction method of claim 1 and decoding the video according to the prediction result of the image.
The video encoding method of claim 1, wherein the video is predicted according to the block-based image prediction method, and the video is encoded according to the prediction result of the image.
The video decoding apparatus of claim 10, wherein the video is decoded according to an image prediction result of the block-based image prediction apparatus.
The video encoding apparatus of claim 10, wherein the video is encoded according to an image prediction result of the block-based image prediction apparatus.
A computer-readable recording medium having recorded thereon a program for implementing the block-based image prediction method of claim 1.