Scaled motion search section with parallel processing and method for use therewith

Patent metadata

Field	Value
Publication number	US-9420308-B2
Application number	US-64248209-A
Country	US
Kind code	B2
Filing date	Dec 18, 2009
Priority date	Dec 18, 2009
Publication date	Aug 16, 2016
Grant date	Aug 16, 2016

Abstract

A scaled motion search section can be used in a video processing device that processes a video input signal that includes a plurality of pictures. The scaled motion search section includes a downscaling module that downscales the plurality of pictures to generate a plurality of downscaled pictures. A reduced-scale motion search module generates a plurality of motion vector candidates at a downscaled resolution, based on the plurality of downscaled pictures. The reduced-scale motion search module includes a column buffer that stores a column of reference data and generates the plurality of motion vector candidates based on a parallel processing of the column of reference data for a group of adjacent macroblock pairs.

First claim

What is claimed is:

1. A scaled motion search section for use in processing a video input signal that includes a plurality of pictures, the scaled motion search section comprising:
a video processing device configured to provide:
a downscaling module that downscales the plurality of pictures to generate a plurality of downscaled pictures; and
a reduced-scale motion search module, coupled to the downscaling module, that generates a plurality of motion vector candidates at a downscaled resolution for at least one of the plurality of pictures, based on the plurality of downscaled pictures, the reduced-scale motion search module including a column buffer that stores a column of reference data for the at least one of the plurality of pictures;
wherein the reduced-scale motion search module generates the plurality of motion vector candidates based on a parallel processing of the column of reference data for a group of adjacent macroblock pairs in a row of the at least one of the plurality of pictures, each of the adjacent macroblock pairs in the group having a corresponding search region, wherein the parallel processing operates by contemporaneously processing the column of reference data for each macroblock or macroblock pair that contains that column of reference data in its corresponding search region.

2. The motion search section of claim 1 wherein the group of adjacent macroblock pairs horizontally span a horizontal dimension of a motion search range for one macroblock pair of the group of adjacent macroblock pairs.

3. The motion search section of claim 1 wherein the column buffer stores a first column of reference data; and wherein the reduced-scale motion search module parallel processes the first column of reference data for each macroblock pair in the group of adjacent macroblock pairs.

4. The motion search section of claim 3 wherein the column buffer stores a second column of reference data; and wherein the reduced-scale motion search module updates the group of adjacent macroblock pairs to form an updated group of adjacent macroblock pairs and parallel processes the second column of reference data for each macroblock pair in the updated group of adjacent macroblock pairs.

5. The motion search section of claim 4 wherein the second column of reference data is adjacent to the first column of reference data.

6. The motion search section of claim 4 wherein the updated group of adjacent macroblock pairs is formed by removing a first macroblock pair from the group of adjacent macroblock pairs and adding a second macroblock pair to the group of adjacent macroblock pairs.

7. The motion search section of claim 1 wherein a full-scale motion search module coupled to the scaled motion search module generates a plurality of motion search motion vectors at a full resolution, based on the plurality of pictures and further based on the plurality of motion vector candidates.

8. The motion search section of claim 1 wherein the downscaling module downscales the plurality of pictures in both a vertical direction and a horizontal direction.

9. The motion search section of claim 1 wherein reduced-scale motion search module operates on only a luma component of the plurality of downscaled pictures.

10. A method for use in a video processing device that processes a video input signal that includes a plurality of pictures, the method comprising:
downscaling the plurality of pictures to generate a plurality of downscaled pictures;
generating a plurality of motion vector candidates at a downscaled resolution for each macroblock pair included in a downscaled picture of the plurality of downscaled pictures by iteratively:
storing a column of reference data in a column buffer for the at least one of the plurality of pictures; and
parallel processing the column of reference data for a group of adjacent macroblock pairs in a row of the at least one of the plurality of pictures, each of the adjacent macroblock pairs in the group having a corresponding search region, wherein the parallel processing operates by contemporaneously processing the column of reference data for each macroblock or macroblock pair that contains that column of reference data in its corresponding search region; and

generating a plurality of motion search motion vectors at a full resolution, based on the plurality of pictures and further based on the plurality of motion vector candidates.

11. The method of claim 10 wherein the group of adjacent macroblock pairs horizontally span a horizontal dimension of a motion search range for one macroblock pair of the group of adjacent macroblock pairs.

12. The method of claim 10 wherein generating a plurality of motion vector candidates at a downscaled resolution for each macroblock pair included in a downscaled picture of the plurality of downscaled pictures includes:
storing a first column of reference data; and
parallel processing the first column of reference data for each macroblock pair in the group of adjacent macroblock pairs.

13. The method of claim 12 wherein generating a plurality of motion vector candidates at a downscaled resolution for each macroblock pair included in a downscaled picture of the plurality of downscaled pictures further includes:
storing a second column of reference data;
updating the group macroblock pairs to form an updated group of adjacent macroblock pairs; and
parallel processing the second column of reference data for each macroblock pair in the updated group of adjacent macroblock pairs.

14. The method of claim 13 wherein the second column of reference data is adjacent to the first column of reference data.

15. The method of claim 13 wherein the updated group of adjacent macroblock pairs is formed by:
removing a first macroblock pair from the group of adjacent macroblock pairs; and
adding a second macroblock pair to the group of adjacent macroblock pairs.

Assignees

• Young Eric
• Vixs Systems Inc

Inventors

• Young Eric

Classifications

• H04N19/53

  Multi-resolution motion estimation; Hierarchical motion estimation · CPC title

• H04N19/56

  Motion estimation with initialisation of the vector search, e.g. estimating a good candidate to initiate a search · CPC title

• H04N19/533

  Motion estimation using multistep search, e.g. two-dimensional [2D]-log search or one-at-a-time search [OTS] · CPC title

• H04N19/59 Primary

  involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution · CPC title