Method of Adaptive Motion Vector Resolution for Video Coding

• US 20170295370A1
• Filed: 09/30/2015
• Published: 10/12/2017
• Est. Priority Date: 09/30/2014
• Status: Active Grant

• Current Assignee: HFI Innovation, Inc. (MediaTek, Inc.)
• Original Assignee: HFI Innovation, Inc. (MediaTek, Inc.)
• Inventors: XU, Xiaozhong, ZHANG, Kai, LIU, Shan, AN, Jicheng, ZHANG, Xianguo
• Granted Patent: US 10,455,231 B2
• Time in Patent Office: Days
• CPC Class Codes: H04N 19/117 Filters, e.g. for pre-processing or post-processing sub-band filter banks H04N19/635
• CPC Class Codes: H04N 19/124 Quantisation
• CPC Class Codes: H04N 19/139 Analysis of motion vectors, e.g. their magnitude, direction, variance or reliability
• CPC Class Codes: H04N 19/176 the region being a block, e.g. a macroblock
• CPC Class Codes: H04N 19/52 by predictive encoding
• CPC Class Codes: H04N 19/523 with sub-pixel accuracy
• CPC Class Codes: H04N 19/86 involving reduction of coding artifacts, e.g. of blockiness

Abstract

A method of MVP (motion vector prediction) for video coding with adaptive motion vector resolution is disclosed. According to the present invention, the MVP coding is applied to the current MV or the current MV is stored depending on the current MV resolution, the reference MV resolution, or both the current MV resolution and the reference MV resolution. In one embodiment, when the current MV resolution corresponds to integer pixel resolution, MVP coding is then applied to the current MV using a modified temporal MV as a motion vector predictor for the current MV, where the modified temporal MV is generated by right-shifting the temporal MV. In another embodiment, when the current MV resolution corresponds to integer pixel resolution, the current MV is left-shifted before it is stored in a memory.

Claims

1. A method of MVP (motion vector prediction) for video data, the method comprising:

receiving input data associated with a current MV (motion vector) for a current block in a current slice;

determining current MV resolution for the current MV, reference MV resolution for a reference MV associated with a reference block in a reference picture, or both the current MV resolution and the reference MV resolution; and

applying MVP coding to the current MV or storing the current MV depending on the current MV resolution, the reference MV resolution, or both the current MV resolution and the reference MV resolution.

2. The method of claim 1, wherein the reference MV associated with the reference block in the reference picture corresponds to a temporal MV associated with a temporal reference block in the reference picture.

3. The method of claim 2, wherein when the current MV resolution corresponds to integer pixel resolution, said applying MVP coding to the current MV uses a modified temporal MV as a motion vector predictor for the current MV, wherein the modified temporal MV is generated by right-shifting the temporal MV.

4. The method of claim 3, wherein an offset is added to the temporal MV before the temporal MV is right-shifted to generate the modified temporal MV.

5. The method of claim 2, wherein when the current MV resolution corresponds to integer pixel resolution, the current MV is left-shifted before it is stored in a memory.

7. The method of claim 5, wherein the current MV stored in the memory is right-shifted before it is used as a motion vector predictor for another block in a current picture containing the current slice.

8. The method of claim 2, wherein when the current MV resolution corresponds to integer pixel resolution, the current MV is left-shifted before it is used in determination of boundary strength used for a deblocking process.

10. The method of claim 2, wherein when the current MV resolution corresponds to non-integer pixel resolution and the reference MV resolution corresponds to integer pixel resolution, said applying MVP coding to the current MV uses a modified temporal MV as a motion vector predictor for the current MV, wherein the modified temporal MV is generated by left-shifting the temporal MV.

11. The method of claim 2, wherein when the current MV resolution is different from the reference MV resolution, said applying MVP coding to the current MV disables the MVP coding for the current block.

12. The method of claim 2, wherein when the current MV resolution is different from the reference MV resolution, said applying MVP coding to the current MV disregards the reference picture for the MVP coding of the current block.

13. The method of claim 2, wherein when a shift operation is applied to the current MV or the temporal MV due to the current MV resolution or the reference MV resolution respectively, the current MV shifted or the temporal MV shifted is clipped to a valid range.

14. The method of claim 2, wherein the current MV has different ranges for different current MV resolutions or the temporal MV has different ranges for different reference MV resolutions.

16. The method of claim 2, wherein the current MV resolution is indicated by a MV resolution flag in a slice header and all blocks within a corresponding slice share the MV resolution flag.

17. The method of claim 2, wherein the current MV resolution is indicated by a MV resolution flag in a sequence level and all blocks within a corresponding sequence share the MV resolution flag.

6. (canceled)

9. (canceled)

15. (canceled)

19. The method of claim 15, wherein when the MV resolution corresponds to integer resolution, the current MV and the neighboring MV are left-shifted by 2 to become a shifted current MV and a shifted neighboring MV, and the shifted current MV and the shifted neighboring MV are included in determination of boundary strength used for said deblocking.

18. A method of deblocking for reconstructed video data, the method comprising:

receiving input data associated with a current reconstructed block in a current slice;

determining MV (motion vector) resolution associated with the current slice;

determining a current MV associated with the current reconstructed block;

determining a neighboring MV associated with a neighboring reconstructed block in the current slice and adjacent to a block boundary of the current reconstructed block; and

deblocking the block boundary depending on the MV resolution.

22. The method of claim 18, wherein when the MV resolution corresponds to integer resolution, a first absolution difference in a vertical component between the current MV and the neighboring MV and a second absolution difference in a horizontal component between the current MV are compared to a threshold value of one instead of four to determine boundary strength used for said deblocking.

20. (canceled)

21. (canceled)

23. (canceled)

24. (canceled)

25. A method of video decoding for color video data, wherein the color video data includes multiple video components and encoding process includes color-space transform, the method comprising:

receiving coded data associated with a current coding block;

determining qPs (quantization parameters) for color components and a color-space-transform-flag from the coded data for the current coding block;

if the color-space-transform-flag indicates that the color-space transform is applied to the current coding block;

generating valid adjusted qPs from the qPs, wherein said generating the valid adjusted qPs comprises modifying the qPs to adjusted qPs to account for the color-space transform and setting the adjusted qPs to equal to or greater than zero if the adjusted qPs are smaller than zero;

de-quantizing quantized transform coefficients associated with the current coding block using the valid adjusted qPs to generate decoded transform coefficients;

applying inverse transform to the decoded transform coefficients to generate a first intermediate reconstructed coding block;

applying inverse color-space transform to the first intermediate coding block or processed first intermediate coding block to generate a second intermediate reconstructed coding block; and

further processing the second intermediate reconstructed coding block to generate a final reconstructed coding block.

26. The method of claim 25, wherein the multiple video components correspond to YCrCb color components, and said generating the valid adjusted qPs from the qPs corresponds to generating a valid adjusted quantization parameter qPX′

for one color component according to qPX′

=Max(0, qPX−

n_x) if the color-space-transform-flag indicates that the color-space transform is applied to the current coding block and qPX′

=Max(0, qPX) if the color-space-transform-flag indicates that the color-space transform is not applied to the current coding block, wherein qPX correspond to quantization parameter for one color component, n_x is 5 for Y component and Cb component and n_x is 3 for Cr component, and Max( ) is a maximum function.

27. The method of claim 25, wherein said setting the adjusted qPs to equal to or greater than zero corresponds to clipping the adjusted qPs to a range from minimum qPs to maximum qPs.

28. The method of claim 25, wherein the multiple video components correspond to YCrCb color components, and said generating the valid adjusted qPs from the qPs corresponds to generating a valid adjusted quantization parameter qPX′

from quantization parameter qPX for one color component by clipping (qPX−

n_x) to a range from MinQPX to MaxQPX if the color-space-transform-flag indicates that the color-space transform is applied to the current coding block, wherein MinQPX and MaxQPX correspond to a valid minimum quantization parameter and a valid maximum quantization parameter for one color component respectively, n_x is 5 for Y component and Cb component and n_x is 3 for Cr component.

29. The method of claim 28, wherein MinQPX is 0 and MaxQPX is 51 for the Y component, the Cb component and Cr component.

30. The method of claim 25, wherein said generating the valid adjusted qPs from the qPs corresponds to generating a valid adjusted quantization parameter qPX′

from quantization parameter qPX for one color component according to a function of qPX if the color-space transform is applied to the current coding block, wherein the function of qPX is always greater than or equal to zero.