Throughput improvement for CABAC coefficient level coding

Patent metadata

Field	Value
Publication number	US-9191670-B2
Application number	US-201213710212-A
Country	US
Kind code	B2
Filing date	Dec 10, 2012
Priority date	Jan 17, 2012
Publication date	Nov 17, 2015
Grant date	Nov 17, 2015

Abstract

This disclosure proposes various techniques for limiting the number of bins that are coded using an adaptive context model with context adaptive binary arithmetic coding (CABAC). In particular, this disclosure proposes to limit the number of bins that use CABAC for coding level information of transform coefficients in a video coding process.

First claim

What is claimed is:

1. A method of coding transform coefficients in a video coding process, the method comprising:
coding a significance map flag for transform coefficients in a chunk of transform coefficients, wherein the significance map flag indicates whether or not a particular transform coefficient has an absolute value greater than zero;
coding a greater-than-one flag for a first M1 transform coefficients in an inverse scan order in the chunk indicated as having an absolute value larger than zero by the significance map flag, wherein the greater-than-one flag indicates whether or not the particular transform coefficient has an absolute value greater than one, and wherein M1 is less than a maximum number of transform coefficients in the chunk;
coding a greater-than-two flag for transform coefficients in the chunk indicated as having an absolute value larger than one by the greater-than-one flag, wherein the greater-than-two flag indicates whether or not the particular transform coefficients has an absolute value greater than two; and
coding a level remaining value for transform coefficients in the chunk,
wherein the level remaining value represents the absolute value of a corresponding coefficient minus three for transform coefficients in the chunk having a coded greater-than-two flag that indicates that the particular transform coefficient has an absolute value greater than two, and
wherein the level remaining value represents the absolute value of a corresponding coefficient minus one for transform coefficients in the chunk that are non-zero but do not have a coded greater-than-one flag.

2. The method of claim 1 , wherein coding the greater-the-two flag comprises coding the greater-than-one flag for a first N transform coefficients in the inverse scan order in the chunk indicated as having an absolute value larger than one by the greater-than-one flag, wherein N is less than a maximum number of transform coefficients in the chunk, and
wherein the level remaining value represents the absolute value of a corresponding coefficient minus two for transform coefficients in the chunk having a coded greater-than-one flag that indicates that the particular transform coefficient has an absolute value greater than one, but the greater-than-two flag is not coded.

3. The method of claim 2 , wherein the value of M1 is 8.

4. The method of claim 1 ,
wherein the level remaining value represents the absolute value of a corresponding coefficient minus one for transform coefficients in the chunk having a coded significance map flag, but for which the greater-than-one flag is not coded.

5. The method of claim 4 , further comprising:
coding a sign flag for transform coefficients in the chunk, wherein the sign flag indicates the sign of the particular transform coefficient.

6. The method of claim 5 , wherein the significance map flag, the greater-than-one flag, and the greater-than-two flag are coded using context adaptive binary arithmetic coding (CABAC) with an adaptive context model, and wherein the level remaining value is coded in CABAC bypass mode, and wherein the sign flag is coded in CABAC bypass mode.

7. The method of claim 6 , further comprising:
coding the level remaining value using a Golomb parameter, and wherein the Golomb Parameter is determined based on at least one of a quantization parameter, a transform unit size, a transform unit depth, a prediction mode, a color component type, and coefficients statistics of neighboring chunks of transform coefficients.

8. The method of claim 5 , further comprising coding the significance map flag first, the greater-than-one flag second, the greater-than-two flag third, and the level remaining value after the greater-than-two flag.

9. The method of claim 8 , further comprising coding each of the significance map flag, the greater-than-one flag, the greater-than-two flag, and the level remaining value for the transform coefficients in the chunk according to the inverse scan order.

10. The method of claim 9 , wherein the inverse scan order is an inverse diagonal scan order.

11. The method of claim 1 , wherein the chunk is a subblock of a transform unit.

12. The method of claim 1 , wherein the chunk is a subblock of 16 transform coefficients of a transform unit.

13. The method of claim 1 , wherein the chunk is a transform unit.

14. The method of claim 1 , wherein the chunk is a number of consecutive transform coefficients along the inverse scan order within a transform unit.

15. The method of claim 14 , wherein the scan order is an inverse diagonal scan order.

16. The method of claim 1 , wherein the value of M1 is 8.

17. The method of claim 1 , wherein the value of M1 is based on at least one of a quantization parameter, a transform unit size, a transform unit depth, a color component type, a position of the chunk in a transform unit, a presence of a last significant coefficient in the chunk, and coefficient statistics of a neighboring chunk.

18. The method of claim 1 , wherein the video coding process is a video encoding process, the method further comprising:
encoding pixel data to produce residual video data;
transforming the residual video data to produce the chunk of transform coefficients; and
signaling the coded significance map flag, the coded greater-than-one flag, and the coded greater-than-two flag in an encoded video bitstream.

19. The method of claim 1 , further comprising:
signaling the value of M1 in an encoded video bitstream.

20. The method of claim 19 , wherein the value of M1 is signaled in at least one of a sequence parameter set, a picture parameter set, an adaptation parameter set, and a slice header.

21. The method of claim 1 , wherein the video coding process is a video decoding process, the method further comprising:
receiving the significance map flag, the greater-than-one flag, and the greater-than-two flag in an encoded video bitstream;
inverse transforming the chunk of transform coefficients to produce residual video data; and
decoding the residual video data to produce pixel data.

22. An apparatus configured to code transform coefficients in a video coding process, the apparatus comprising:
a memory configured to store transform coefficients; and
a video coder configured to:
code a significance map flag for the transform coefficients in a chunk of transform coefficients, wherein the significance map flag indicates whether or not a particular transform coefficient has an absolute value greater than zero;
code a greater-than-one flag for a first M1 transform coefficients in an inverse scan order in the chunk indicated as having an absolute value larger than zero by the significance map flag, wherein the greater-than-one flag indicates whether or not the particular transform coefficient has an absolute value greater than one, and wherein M1 is less than a maximum number of transform coefficients in the chunk;
code a greater-than-two flag for transform coefficients in the chunk indicated as having an absolute value larger than one by the greater-than-one flag, wherein the greater-than-two flag indicates whether or not the particular transform coefficients has an absolute value greater than two; and
code a level remaining value for transform coefficients in the chunk,
wherein the level remaining value represents the absolute value of a corresponding coefficient minus three for transform coefficient in the chunk having a coded greater-than-two flag that indicates that the particular transfor

Assignees

• Qualcomm Inc

Inventors

• Karczewicz Marta
• Chen Jianle
• Chien Wei-Jung
• Joshi Rajan Laxman

Classifications

• H04N19/70 Primary

  characterised by syntax aspects related to video coding, e.g. related to compression standards · CPC title

• H04N19/60 Primary

  using transform coding · CPC title

• H04N19/00775 Primary

  Electricity · mapped topic

• H04N19/13

  Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC] · CPC title