Method and system for processing one or more multi-channel images

Patent metadata

Field	Value
Publication number	US-9977998-B2
Application number	US-201514696631-A
Country	US
Kind code	B2
Filing date	Apr 27, 2015
Priority date	Apr 27, 2015
Publication date	May 22, 2018
Grant date	May 22, 2018

Abstract

Various aspects of a method and device to process one or more multi-channel images are disclosed herein. The method to process one or more multi-channel images is executed within an electronic device. A channel distance value between each channel of a target patch and corresponding channel of one or more neighbor patches of one or more multi-channel images is determined. The determination of the channel distance value is based on channel data in each channel of the one or more multi-channel images. Based on relative signal information in each channel of the target patch, a channel weight for each channel is dynamically determined. Based on the determined channel distance value and the dynamically determined channel weight for each channel, a patch matching score is determined.

First claim

What is claimed is:

1. A method for processing one or more multi-channel images, the method comprising:
in an electronic device:
determining a channel distance value between each channel of a target patch and a corresponding channel of one or more neighbor patches based on channel data in said each channel of said one or more multi-channel images;
dynamically determining a channel weight for said each channel based on relative signal information in said each channel of said target patch;
updating said dynamically determined channel weight based on a change in the relative signal information for said each channel;
determining a patch matching score based on said determined channel distance value and said dynamically determined channel weight for said each channel;
selecting at least one neighbor patch of said one or more neighbor patches based on corresponding said determined patch matching score of said at least one neighbor patch; and
denoising said target patch based on said selected at least one neighbor patch.

2. The method of claim 1 , wherein said target patch comprises one or more pixels, wherein each pixel of said one or more pixels comprises one or more components.

3. The method of claim 1 , wherein said each channel of said target patch and said one or more neighbor patches comprise a similar component from each pixel of said target patch and said one or more neighbor patches, respectively.

4. The method of claim 1 , wherein said relative signal information in said each channel of said target patch is complementary with respect to remaining channels of said target patch.

5. The method of claim 1 , further comprising selecting said one or more neighbor patches based on said corresponding patch matching scores as determined less than a threshold value.

6. The method of claim 1 , further comprising selecting a number of said at least one neighbor patch of said one or more neighbor patches,
wherein corresponding patch matching scores of each of said number of said at least one neighbor patch are lower than a remaining of said one or more neighbor patches.

7. The method of claim 6 , further comprising computing an average of a first center pixel of said target patch and one or more second center pixels of said selected at least one neighbor patch.

8. The method of claim 7 , wherein said denoising on said target patch is further based on said averaging.

9. The method of claim 1 , wherein said target patch and said one or more neighbor patches comprise at least a portion of an image from said one or more multi-channel images.

10. The method of claim 1 , further comprising determining a first image and a consecutive second image from said one or more multi-channel images, wherein said first image comprises said target patch and said consecutive second image comprises said one or more neighbor patches.

11. The method of claim 1 , further comprising selecting one of said one or more neighbor patches with a lowest patch matching score.

12. The method of claim 11 , further comprising computing a vector difference between said target patch and said selected one or more neighbor patches.

13. The method of claim 12 , further comprising determining motion vector of said target patch based on said computed vector difference.

14. The method of claim 1 , wherein said determination of said channel distance value is based on one of a sum of absolute difference (SAD) function, a sum of squared difference (SSD) function, a weighted sum of absolute difference (WSAD) function, or a weighted sum of squared difference (WSSD) function.

15. The method of claim 1 , further comprising normalizing said dynamically determined channel weights for said each channel.

16. The method of claim 1 , further comprising computing said relative signal information in said target patch based on signal power, noise power, or signal-to-noise power ratio of said each channel of said target patch.

17. The method of claim 16 , wherein said signal power, said noise power, or said signal-to-noise power ratio of said each channel of said target patch is based on a local mean or a local variance of said each channel of said target patch.

18. A system to process one or more multi-channel images, the system comprising:
one or more processors operable to:
determine a channel distance value between each channel of a target patch and a corresponding channel of one or more neighbor patches based on channel data in said each channel of said one or more multi-channel images;
dynamically determine a channel weight for said each channel based on relative signal information in said each channel of said target patch;
update said dynamically determined channel weight based on a change in the relative signal information for said each channel;
determine a patch matching score based on said determined channel distance value and said dynamically determined channel weight for said each channel;
select at least one neighbor patch of said one or more neighbor patches based on corresponding said determined patch matching score of said at least one neighbor patch; and
denoise said target patch based on said selected at least one neighbor patch.

19. The system of claim 18 , wherein said one or more processors are further operable to select said one or more neighbor patches based on said corresponding patch matching scores as determined less than a threshold value.

20. The system of claim 18 ,
wherein said one or more processors are further operable to select a number of said at least one neighbor patch of said one or more neighbor patches, and
wherein corresponding patch matching scores of each of said number of said at least one neighbor patch are lower than a remaining of said one or more neighbor patches.

21. The system of claim 20 , wherein said one or more processors are further operable to compute an average of a first center pixel of said target patch and one or more second center pixels of said selected at least one neighbor patch.

22. The system of claim 21 , wherein said one or more processors are further operable to denoise said target patch based on said computed average.

23. The system of claim 18 , wherein said target patch and said one or more neighbor patches comprise at least a portion of an image from said one or more multi-channel images.

24. The system of claim 18 , wherein said one or more processors are further operable to determine a first image and a consecutive second image from said one or more multi-channel images,
wherein said first image comprises said target patch and said consecutive second image comprises said one or more neighbor patches.

25. The system of claim 24 , wherein said one or more processors are further operable to select one of said one or more neighbor patches with a lowest patch matching score.

26. The system of claim 25 , wherein said one or more processors are further operable to compute a vector difference between said target patch and said selected neighbor patch.

27. The system of claim 26 , wherein said one or more processors are further operable to determine a motion vector of said target patch based on said computed vector difference.

28. A non-transitory computer-readable medium having stored thereon, computer-executable instructions for causing a computer to execute operations, the operations comprising:
determining a channel distance value between e

Assignees

• Sony Corp

Inventors

• Wong Tak Shing
• Takatori Jiro
• Dong Xiaogang

Classifications

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06F18/24147

  Distances to closest patterns, e.g. nearest neighbour classification · CPC title

• G06V10/751 Primary

  Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06V10/56

  relating to colour · CPC title