Method and apparatus for processing video signal using graph-based transform

Patent metadata

Field	Value
Publication number	US-10742988-B2
Application number	US-201615550721-A
Country	US
Kind code	B2
Filing date	Feb 4, 2016
Priority date	Feb 12, 2015
Publication date	Aug 11, 2020
Grant date	Aug 11, 2020

Abstract

The present invention provides a method of decoding a video signal using graph-based transform comprising the steps: extracting prediction unit partition information of a current coding unit from the video signal; obtaining a graph-based transform kernel from predetermined table information based on the prediction unit partition information; and performing an inverse-transform of a transform unit using the graph-based transform kernel, wherein the graph-based transform kernel corresponds to at least one of the prediction unit partition information and an edge weight, and the edge weight is a predetermined value representing a correlation between pixels.

First claim

The invention claimed is:

1. A method of decoding a video signal using graph-based transform, comprising:
extracting prediction unit partition information of a current coding unit from the video signal;
obtaining a graph-based transform kernel from predetermined table information based on the prediction unit partition information; and
performing an inverse-transform of a transform unit using the graph-based transform kernel,
wherein the graph-based transform kernel corresponds to at least one of the prediction unit partition information and an edge weight, and the edge weight is a predetermined value representing a correlation between pixels,
wherein the edge weight applied to a graph edge overlapped with a prediction unit partition boundary is smaller than the edge weight applied to the graph edge not overlapped with the prediction unit partition boundary,
wherein, based on the current coding unit being configured with a plurality of transform units by a quadtree decomposition, different transform types are applied to a first transform unit overlapped with the prediction unit partition boundary and a second transform unit not overlapped with the prediction unit partition boundary,
wherein the graph-based transform kernel is applied to the first transform unit, and another transform kernel instead of the graph-based transform kernel is applied to the second transform unit, and
wherein the another transform kernel comprises discrete cosine transform (DCT) or discrete sine transform (DST).

2. A method of encoding a video signal using graph-based transform, the method comprising:
calculating an edge weight within a graph corresponding to a coding unit, wherein the edge weight is a value representing a correlation between pixels;
generating separately graph-based transform kernels for rows and columns of a transform unit based on coding information of a prediction unit; and
performing a transform for the transform unit using the graph-based transform kernel,
wherein the graph-based transform kernels are generated based on the edge weight,
wherein the edge weight applied to a graph edge overlapped with a prediction unit partition boundary is smaller than the edge weight applied to the graph edge not overlapped with the prediction unit partition boundary,
wherein, based on a current coding unit being configured with a plurality of transform units by a quadtree decomposition, different transform types are applied to a first transform unit overlapped with the prediction unit partition boundary and a second transform unit not overlapped with the prediction unit partition boundary,
wherein the graph-based transform kernel is applied to the first transform unit, and another transform kernel instead of the graph-based transform kernel is applied to the second transform unit, and
wherein the another transform kernel comprises discrete cosine transform (DCT) or discrete sine transform (DST).

3. The method of claim 2 , wherein the coding information of the prediction unit comprises at least one of partition information and motion information.

4. The method of claim 3 , wherein the graph-based transform kernel is applied, if a difference value between motion information of a plurality of prediction units is larger than a predetermined threshold value, when the coding unit is configured with a plurality of prediction units.

5. An apparatus for decoding a video signal using graph-based transform, comprising:
a processor configured to
extract prediction unit partition information of a current coding unit from the video signal; and
obtain a graph-based transform kernel from predetermined table information based on the prediction unit partition information and that performs an inverse-transform of a transform unit using the graph-based transform kernel,
wherein the graph-based transform kernel corresponds to at least one of the prediction unit partition information and an edge weight, and the edge weight is a predetermined value representing a correlation between pixels,
wherein the edge weight applied to a graph edge overlapped with a prediction unit partition boundary is smaller than the edge weight applied to the graph edge not overlapped with the prediction unit partition boundary,
wherein, based on the current coding unit being configured with a plurality of transform units by a quadtree decomposition, different transform types are applied to a first transform unit overlapped with the prediction unit partition boundary and a second transform unit not overlapped with the prediction unit partition boundary,
wherein the graph-based transform kernel is applied to the first transform unit, and another transform kernel instead of the graph-based transform kernel is applied to the second transform unit, and
wherein the another transform kernel comprises discrete cosine transform (DCT) and discrete sine transform (DST).

6. An apparatus for encoding a video signal using graph-based transform, comprising:
a processor configured to calculate an edge weight within a graph corresponding to a coding unit,
wherein the edge weight is a value representing a correlation between pixels and that generates separately graph-based transform kernels for rows and columns of a transform unit based on coding information of a prediction unit and that performs transform of the transform unit using the graph-based transform kernel,
wherein the graph-based transform kernels are generated based on the edge weight,
wherein the edge weight applied to a graph edge overlapped with a prediction unit partition boundary is smaller than the edge weight applied to the graph edge not overlapped with the prediction unit partition boundary,
wherein, based on a current coding unit being configured with a plurality of transform units by a quadtree decomposition, different transform types are applied to a first transform unit overlapped with the prediction unit partition boundary and a second transform unit not overlapped with the prediction unit partition boundary,
wherein the graph-based transform kernel is applied to the first transform unit, and another transform kernel instead of the graph-based transform kernel is applied to the second transform unit, and
wherein the another transform kernel comprises discrete cosine transform (DCT) or discrete sine transform (DST).

7. The apparatus of claim 6 , wherein coding information of the prediction unit comprises at least one of partition information and motion information, and
the graph-based transform kernel is applied, if a difference value between motion information of a plurality of prediction units is larger than a predetermined threshold value, when the coding unit is configured with a plurality of prediction units.

Assignees

• Lg Electronics Inc

Inventors

• Koo Moonmo
• Heo Jin
• Son Eunyong
• Yea Sehoon
• Park Seungwook
• Kim Kyuwoon
• Lee Bumshik

Classifications

• H04N19/176

  the region being a block, e.g. a macroblock · CPC title

• H04N19/96 Primary

  Tree coding, e.g. quad-tree coding · CPC title

• H04N19/17

  the unit being an image region, e.g. an object · CPC title

• H04N19/14 Primary

  Coding unit complexity, e.g. amount of activity or edge presence estimation (H04N19/146 takes precedence) · CPC title

• H04N19/119 Primary

  Adaptive subdivision aspects, e.g. subdivision of a picture into rectangular or non-rectangular coding blocks · CPC title