Using an average motion vector for a motion search

• US 10,602,175 B2
• Filed: 12/21/2012
• Issued: 03/24/2020
• Est. Priority Date: 12/21/2012
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Chen, Jianjun, He, Xi, Yu, Xinyang, Fan, Hank
• Primary Examiner(s): Vazquez Colon, Maria E
• Application Number:
  US13/725,613
• Publication Number:
  US 20140177716A1
• Time in Patent Office: 2,650 Days
• CPC Class Codes: H04N 19/513 Processing of motion vectors
• CPC Class Codes: H04N 19/56 Motion estimation with initialisation of the vector search, e.g. estimating a good candidate to initiate a search

Abstract

A method for using an average motion vector in a motion vector search process. The method includes accessing an input frame for processing and reading average motion vector information from memory. The method further includes performing a motion vector search by using the average motion vector and a plurality of hints, calculating a winner motion vector based on the average motion vector and the plurality of hints, and storing the winner motion vector back into memory to create a new updated average motion vector. The method further includes finishing processing the input frame using the winning motion vector.

Claims

1. A method for using an average motion vector in a motion vector search process, comprising:

accessing an input frame for processing;

reading average motion vector information from memory, wherein the average motion vector comprises an average of a plurality of prior winning motion vectors computed when encoding a plurality of previous frames;

performing a motion vector search by using the average motion vector and a plurality of hints;

responsive to said motion vector search, calculating a winner motion vector based on the average motion vector and the plurality of hints;

storing the winner motion vector back into the memory to create a new updated average motion vector; and

finishing processing the input frame using the winner motion vector.

2. The method of claim 1, wherein the motion vector search implements a video compression process.

3. The method of claim 1, wherein the motion vector search is performed on a macro block.

4. The method of claim 1, wherein a plurality of winner motion vector values are calculated and an average motion vector hint is calculated before encoding a current frame.

5. The method of claim 1, wherein the motion vector search is implemented in a graphics processing unit.

6. The method of claim 1, wherein the average motion vector is scaled based on a distance between a current frame and a reference frame.

18. The method of claim 1, wherein the plurality of hints comprises:

a top neighbor motion vector, a left neighbor motion vector, and a collocated motion vector.

7. A non-transitory computer readable media having computer readable code which when executed by a computer system causes said computer system to implement a method for using an average motion vector in a motion vector search process, the method comprising:

accessing an input frame for processing;

reading average motion vector information from memory, wherein the average motion vector comprises an average of a plurality of prior winning motion vectors computed when encoding a plurality of previous frames;

performing the motion vector search process by using the average motion vector and a plurality of hints;

responsive to the motion vector search process, calculating a winner motion vector based on the average motion vector and the plurality of hints;

storing the winner motion vector back into the memory to create a new updated average motion vector; and

finishing processing the input frame using the winner motion vector.

8. The non-transitory computer readable media of claim 7, wherein the motion vector search process implements a video compression process.

9. The non-transitory computer readable media of claim 7, wherein the motion vector search process is performed on a macro block.

10. The non-transitory computer readable media of claim 7, wherein a plurality of winner motion vector values are calculated and an average motion vector hint is calculated before encoding a current frame.

11. The non-transitory computer readable media of claim 7, wherein the motion vector search process is implemented in a graphics processing unit.

12. The non-transitory computer readable media of claim 7, wherein the average motion vector is scaled based on a distance between a current frame and a reference frame.

19. The non-transitory computer readable media of claim 7, wherein the plurality of hints comprises:

a top neighbor motion vector, a left neighbor motion vector, and a collocated motion vector.

13. A computer system, comprising:

a computer system having a processor coupled to a computer readable storage media and executing computer readable code which causes the computer system to;

access an input frame for processing;

read average motion vector information from memory, wherein the average motion vector comprises an average of a plurality of prior winning motion vectors computed when encoding a plurality of previous frames;

perform a motion vector search process by using the average motion vector information and a plurality of hints;

responsive to said motion vector search process, calculate a winner motion vector based on the average motion vector and the plurality of hints;

store the winner motion vector back into the memory to create a new updated average motion vector; and

finish processing the input frame using the winner motion vector.

14. The computer system of claim 13, wherein the motion vector search process implements a video compression process.

15. The computer system of claim 13, wherein the motion vector search process is performed on a macro block.

16. The computer system of claim 13, wherein a plurality of winner motion vector values are calculated and an average motion vector hint is calculated before encoding a current frame.

17. The computer system of claim 13, wherein the average motion vector is scaled based on a distance between a current frame and a reference frame.

20. The computer system of claim 13, wherein the plurality of hints comprises:

a top neighbor motion vector, a left neighbor motion vector, and a collocated motion vector.