IP Library Granted Patent US 9,794,588
Granted Patent B2
US 9,794,588 · App. 14/871,806 · Granted Oct 17, 2017

Image processing system with optical flow recovery mechanism and method of operation thereof

Inventors: Xiaogang Dong (Boyds, MD); Jiro Takatori (Tokyo, JP); Tak-Shing Wong (Pleasanton, CA)
Assignee: Sony Corporation
H04N19/521H04N19/573
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 9,794,588
App. No.
14/871,806
Granted
Oct 17, 2017
Kind
B2
Abstract

An image processing system, and a method of operation thereof, includes: an edge motion generation unit for detecting edges in a first image frame and a second image frame stored by a storage device or a memory and for generating edge motion vectors between the first image frame and the second image frame based on the edges; a motion vector list generation unit for extracting dominant motion vectors from a group of the edge motion vectors and for generating a motion vector list based on the dominant motion vectors; an image segmentation unit for generating a segmentation of the first image frame; an initial motion generation unit for generating initial motion vectors based on the segmentation and the motion vector list; and a smooth motion generation unit for generating smooth motion vectors based on the initial motion vectors and for generating a dense optical flow field by combining the smooth motion vectors.

Claims (58)

1. A method of operation of an image processing system comprising:

detecting edges in a first image frame and a second image frame stored by a storage device or a memory;

generating edge motion vectors between the first image frame and the second image frame based on the edges;

extracting dominant motion vectors from a group of the edge motion vectors;

generating a motion vector list based on the dominant motion vectors;

generating a segmentation of the first image frame;

generating initial motion vectors based on the segmentation and the motion vector list;

generating smooth motion vectors based on the initial motion vectors; and

generating a dense optical flow field by combining the smooth motion vectors.

2. The method as claimed in claim 1 wherein generating the initial motion vectors includes generating the initial motion vectors for image segments instead of individual pixels.

3. The method as claimed in claim 1 further comprising detecting a backward similar region in the first image frame by a backward motion search to determine an unreliable edge motion.

4. The method as claimed in claim 1 wherein generating the smooth motion vectors includes generating the smooth motion vectors based on a multiplier and the initial motion vectors.

5. The method as claimed in claim 1 further comprising:

generating the first image frame by down-sampling an original image frame having a finer scale; and

wherein:

detecting the edges includes detecting the edges in the first image frame having a coarsest scale with a resolution smaller than the finer scale of the original image frame.

6. A method of operation of an image processing system comprising:

detecting edges in a first image frame and a second image frame stored by a storage device or a memory;

generating edge motion vectors between the first image frame and the second image frame based on the edges;

extracting dominant motion vectors from a group of the edge motion vectors;

generating a motion vector list based on the dominant motion vectors;

generating a segmentation to generate image segments in the first image frame;

generating initial motion vectors based on the segmentation and the motion vector list;

generating smooth motion vectors based on the initial motion vectors; and

generating a dense optical flow field by combining the smooth motion vectors.

7. The method as claimed in claim 6 wherein generating the initial motion vectors includes generating the initial motion vectors for the image segments instead of individual pixels, the image segments generated based on luminance, chrominance, or texture of the first image frame.

8. The method as claimed in claim 6 further comprising:

detecting a backward similar region in the first image frame by a backward motion search to determine an unreliable edge motion; and

calculating the dominant motion vectors based on the backward similar region and reliable edge motion vectors.

9. The method as claimed in claim 6 further comprising:

generating an initial similarity distance with the image segments transformed by the initial motion vectors; and

generating a neighbor similarity distance with the image segments transformed by a neighbor motion vector, the neighbor motion vector generated based on a shared boundary of the image segments; and

wherein:

generating the smooth motion vectors includes generating the smooth motion vectors by comparing the neighbor similarity distance and a product of a multiplier and the initial motion vectors with the multiplier greater than one.

10. The method as claimed in claim 6 further comprising:

generating the first image frame by down-sampling an original image frame having a finer scale; and

wherein:

detecting the edges includes detecting the edges in the first image frame having a coarsest scale with a resolution smaller than the finer scale of the original image frame; and

generating the edge motion vectors includes generating the edge motion vectors based on the edges, the edge motion vectors having at least two values.

11. An image processing system comprising:

an edge motion generation unit for detecting edges in a first image frame and a second image frame stored by a storage device or a memory and for generating edge motion vectors between the first image frame and the second image frame based on the edges;

a motion vector list generation unit for extracting dominant motion vectors from a group of the edge motion vectors and for generating a motion vector list based on the dominant motion vectors;

an image segmentation unit for generating a segmentation of the first image frame;

an initial motion generation unit for generating initial motion vectors based on the segmentation and the motion vector list; and

a smooth motion generation unit for generating smooth motion vectors based on the initial motion vectors and for generating a dense optical flow field by combining the smooth motion vectors.

12. The system as claimed in claim 11 wherein the initial motion generation unit is for generating the initial motion vectors for image segments instead of individual pixels.

13. The system as claimed in claim 11 wherein the edge motion generation unit is for detecting a backward similar region in the first image frame by a backward motion search to determine an unreliable edge motion.

14. The system as claimed in claim 11 wherein the smooth motion generation unit is for generating the smooth motion vectors based on a multiplier and the initial motion vectors.

15. The system as claimed in claim 11 wherein the edge motion generation unit is for generating the first image frame by down-sampling an original image frame having a finer scale and for detecting the edges in the first image frame having a coarsest scale with a resolution smaller than the finer scale of the original image frame.

16. The system as claimed in claim 11 wherein the image segmentation unit is for generating the segmentation to generate image segments in the first image frame.

17. The system as claimed in claim 16 wherein the initial motion generation unit is for generating the initial motion vectors for the image segments instead of individual pixels, the image segments generated based on luminance, chrominance, or texture of the first image frame.

18. The system as claimed in claim 16 wherein:

the edge motion generation unit is for detecting a backward similar region in the first image frame by a backward motion search to determine an unreliable edge motion; and

the motion vector list generation unit is for calculating the dominant motion vectors based on the backward similar region and reliable edge motion vectors.

19. The system as claimed in claim 16 wherein the smooth motion generation unit is for generating an initial similarity distance with the image segments transformed by the initial motion vectors, for generating a neighbor similarity distance with the image segments transformed by a neighbor motion vector, the neighbor motion vector generated based on a shared boundary of the image segments, and for generating the smooth motion vectors by comparing the neighbor similarity distance and a product of a multiplier and the initial motion vectors with the multiplier greater than one.

20. The system as claimed in claim 16 wherein:

the edge motion generation unit is for generating the first image frame by down-sampling an original image frame having a finer scale and for detecting the edges in the first image frame having a coarsest scale with a resolution smaller than the finer scale of the original image frame; and

the motion vector list generation unit is for generating the edge motion vectors based on the edges, the edge motion vectors having at least two values.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 19, 2015
From: DONG, XIAOGANG; TAKATORI, JIRO; WONG, TAK-SHING
To: SONY CORPORATION
Reel/Frame 036825/0202 →
Continuity (1)
Related Publication 20170094310A1 · Mar 30, 2017