Simplified non-square quadtree transforms for video coding

Patent metadata

Field	Value
Publication number	US-9912944-B2
Application number	US-201313863266-A
Country	US
Kind code	B2
Filing date	Apr 15, 2013
Priority date	Apr 16, 2012
Publication date	Mar 6, 2018
Grant date	Mar 6, 2018

Abstract

In an example, a method of decoding video data includes determining a prediction partitioning structure for predicting pixel values associated with a block of video data. The method also includes determining a transform partitioning structure for applying one or more transforms to the predicted pixel values. Determining the transform split structure includes splitting a parent transform unit, upon determining the transform partitioning structure comprises splitting the parent transform unit into one or more square transforms, determining one or more square transforms such that each of the one or more square transforms correspond to exactly one prediction partition, and upon determining the transform partitioning structure comprises splitting the parent transform unit into one or more non-square transforms, determining whether to split the one or more non-square transforms based at least in part on the one or more non-square transforms being non-square.

First claim

What is claimed is:

1. A method of decoding encoded video data, the method comprising:
determining that residual quadtree splitting is enabled with respect to a block of the encoded video data;
determining a maximum residual quad tree depth with respect to the residual quadtree splitting of the block of the encoded video data;
determining that a prediction partitioning structure for reconstructing pixel values associated with the block of video data includes one or more asymmetric motion partition (AMP) boundaries;
determining a transform partitioning structure for applying one or more inverse transforms to the reconstructed pixel values, wherein determining the transform partitioning structure includes:
performing, in response to the residual quadtree splitting being enabled with respect to the block of the encoded video data and based on the residual quadtree splitting resulting in one or more transforms crossing at least one AMP boundary of the AMP boundaries, forced splitting of a parent transform unit associated with the block of the encoded video data, wherein the parent transform unit has an equal area with respect to the block of video data, wherein the forced splitting of the parent transform unit produces a plurality of square transforms, and wherein each square transform of the plurality of square transforms is at the maximum residual quad tree depth associated with the residual quadtree splitting; and

upon performing the forced splitting of the parent transform unit to form the plurality of square transforms, determining each respective square transform of the plurality of square transforms based on the one or more AMP boundaries such that each respective square transform of the plurality of square transforms corresponds to a maximum transform size that does not cross any AMP boundary of the one or more AMP boundaries.

2. The method of claim 1 , wherein performing the forced splitting of the parent transform unit to form the plurality of square transforms comprises splitting at least one transform that is applied across a respective AMP boundary of the one or more AMP boundaries and without splitting any transforms that are not applied across the respective AMP boundary.

3. The method of claim 1 , further comprising adaptively enabling the determination of the plurality of square transforms based on a coding characteristic associated with the block of the encoded video data, the coding characteristic comprising one of a block size, a picture type, a picture size, a prediction mode, a motion vector amplitude, or a reference index associated with the block of the encoded video data.

4. The method of claim 1 , further comprising applying differently sized transforms to the residual values.

5. The method of claim 4 , wherein the differently sized transforms include at least one transform having a size smaller than a smallest transform size indicated by a residual quad-tree structure associated with the forced splitting.

6. The method of claim 1 , wherein the prediction partitioning structure comprises prediction units sized one of nL×2N, nR×2N, 2N×nU, or 2N×nD.

7. The method of claim 1 , further comprising determining that each respective square transform of the plurality of square transforms can only be further split into further non-square transforms.

8. The method of claim 7 , wherein the parent transform unit comprises a parent luminance transform unit and wherein the further non-square transforms comprise one or more non-square luminance transforms, the method further comprising:
splitting a parent chrominance transform unit into a plurality of non-square chrominance transforms, wherein the parent chrominance transform unit corresponds to the parent luminance transform unit; and
splitting the one or more non-square chrominance transforms.

9. The method of claim 1 , further comprising applying one or more transform splitting restrictions based at least in part on one or more of a size of a coding unit corresponding to the parent transform unit, a picture type of a picture including the parent transform unit, a pixel resolution of the picture including the parent transform unit, and a block type for the coding unit corresponding to the parent transform unit.

10. The method of claim 1 , the method being executable on a wireless communication device, wherein the wireless communication device comprises:
data storage medium configured to store the block of video data;
a processor configured to execute instructions to process the block of video data stored in the data storage medium; and
a receiver to receive the block of video data.

11. The method of claim 10 , wherein the wireless communication device is a cellular telephone and the block of video data is received by the receiver and modulated according to a communication standard.

12. A method of encoding video data, the method comprising:
determining that a prediction partitioning structure for residual pixel values associated with a block of video data includes one or more asymmetric motion partition (AMP) boundaries;
determining a transform partitioning structure for applying one or more transforms to the residual pixel values, wherein determining the transform partitioning structure includes:
determining that a residual quadtree splitting of the block results in one or more transforms crossing at least one AMP boundary of the AMP boundaries;
performing, based on the transform splitting of the block resulting in the one or more transforms crossing the at least one AMP boundary of the AMP boundaries, forced splitting of a parent transform unit associated with the block of video data, wherein the parent transform unit has an equal area with respect to the block of video data, wherein the forced splitting of the parent transform unit produces a plurality of square transforms, and wherein each square transform of the plurality of square transforms is at a maximum residual quad tree depth associated with the residual quadtree splitting;
upon performing the forced splitting of the parent transform unit to form the plurality of square transforms, determining each respective square transform of the plurality of square transforms based on the one or more AMP boundaries such that each respective square transform of the plurality of square transforms corresponds to a maximum transform size that does not cross any AMP boundary of the one or more AMP boundaries;

encoding an indication that the transform splitting is enabled with respect to the block of video data;
encoding an indication of the maximum residual quad tree depth with respect to the transform splitting of the block of video data; and
encoding the residual pixel values based on the determined transform partitioning structure.

13. The method of claim 12 , wherein performing the forced splitting to form the plurality of square transforms comprises splitting at least one transform that is applied across a respective AMP boundary of the one or more AMP boundaries and without splitting any transforms that are not applied across the respective AMP boundary of the one or more AMP boundaries.

14. The method of claim 12 , further comprising adaptively enabling the determination of the plurality of square transforms based on a coding characteristic associated with the block of video data, the coding characteristic comprising one of a block size, a picture type, a picture size, a prediction mode, a motion vector amplitude, or a reference index associated with the block of video data.

15. The method of claim 12 , further comprising applying differently sized transforms to the residual values.

Assignees

• Qualcomm Inc

Inventors

• Guo Liwei
• Wang Xianglin
• Karczewicz Marta

Classifications

• H04N19/619

  the transform being operated outside the prediction loop · CPC title

• H04N19/96 Primary

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

• H04N19/119

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

• H04N19/159

  Prediction type, e.g. intra-frame, inter-frame or bidirectional frame prediction · CPC title

• H04N19/122

  Selection of transform size, e.g. 8x8 or 2x4x8 DCT; Selection of sub-band transforms of varying structure or type · CPC title