Methods and Systems for Processing Film Grain Noise

• US 20080152296A1
• Filed: 07/10/2007
• Published: 06/26/2008
• Est. Priority Date: 12/21/2006
• Status: Active Grant

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Lei, Shawmin, Oh, Byung Tae
• Granted Patent: US 8,213,500 B2
• Time in Patent Office: Days
• US Class Current: 386/47
• CPC Class Codes: H04N 19/80 Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation H04N19/635, H04N19/86 take precedence

Abstract

Embodiments of the present invention comprise systems and methods for image noise characterization, parameterization, removal, reconstruction and/or re-combination with a denoised image.

Claims

1. A method for encoding image data and noise parameters, said method comprising:

a) receiving a video sequence comprising a first frame and a second frame, wherein said first frame comprises a first region and said second frame comprises a second region, which is static in relation to said first region;

b) calculating a noise image region by determining a difference between said first region and said second region;

c) measuring at least one noise parameter of said noise image region;

d) detecting a smooth region of a frame of said video sequence; and

e) denoising said smooth region using said at least one noise parameter as an a priori constraint in a total variation minimization process.

2. A method as described in claim 1 further comprising sending an encoded image resulting from said denoising and said at least one noise parameter to a video decoder.

3. A method as described in claim 1 wherein said measuring at least one noise parameter comprises determining the variance of said noise image region.

4. A method as described in claim 1 wherein said measuring at least one noise parameter comprises determining an auto-correlation value of said noise image region.

5. A method as described in claim 1 wherein said measuring at least one noise parameter comprises determining a cross-color correlation value of said noise image region.

6. A method for encoding image data and noise parameters, said method comprising:

a) receiving a video sequence comprising a first frame and a second frame, wherein said first frame comprises a first region and said second frame comprises a second region, which is static in relation to said first region;

b) calculating a noise image region by determining a difference between said first region and said second region;

c) measuring at least one noise parameter of said noise image region;

d) detecting a smooth region of a frame of said video sequence; and

e) denoising said smooth region using said at least one noise parameter as an a posteriori constraint in a MAP process.

7. A method as described in claim 6 further comprising sending an encoded image resulting from said denoising and said at least one noise parameter to a video decoder.

8. A method as described in claim 6 wherein said measuring at least one noise parameter comprises determining the variance of said noise image region.

9. A method as described in claim 6 wherein said measuring at least one noise parameter comprises determining an auto-correlation value of said noise image region.

10. A method as described in claim 6 wherein said measuring at least one noise parameter comprises determining a cross-color correlation value of said noise image region.

11. A method as described in claim 6 wherein said detecting a smooth region comprises edge detection.

12. A method for estimating a denoised image from an observed image, said method comprising:

a) determining a first-channel image of a denoised image by minimizing an energy function, wherein said minimization is subject to a first constraint and a second constraint,

b) wherein said first constraint is related to the difference between said first-channel image of said denoised image and a first-channel image of an observed image; and

c) wherein said second constraint is related to the difference between said first-channel image of said denoised image and a first-channel image of an observed image and the difference between a second-channel image of said denoised image and a second-channel image of said observed image.

13. A method as described in claim 12 wherein energy function is a gradient magnitude.

14. A method as described in claim 12 wherein said observed image is an RGB image, said first channel is a green color channel and said second channel is a color channel selected from the group consisting of a red color channel and a blue color channel.

15. A method as described in claim 12 wherein said first constraint is minimizing the norm of the difference between said first-channel image of said denoised image and said first-channel image of said observed image.

16. A method as described in claim 12 wherein said second constraint is minimizing the norm of the difference between said first-channel image of said denoised image and said first-channel image of said observed image and minimizing the norm of the difference between said second-channel image of said denoised image and said second-channel image of said observed image.

17. A method for image edge detection, said method comprising:

a) filtering an image with a first filter to create a first edge energy image;

b) binarizing said first edge energy image to create a first binarized edge energy image;

c) performing a directionality check on said first binarized edge energy image to create a first directionality image;

d) assigning edge probability values to said first edge energy image to create a first edge energy map;

e) assigning edge probability values to said first directionality image to create a first directionality-based edge map

f) combining said first edge energy map with said first directionality-based edge map to create a combined edge map.

18. A method as described in claim 17 further comprising:

a) filtering said image with a second filter to create a second edge energy image;

b) binarizing said second edge energy image to create a second binarized edge energy image;

c) performing a directionality check on said second binarized edge energy image to create a second directionality image;

d) assigning edge probability values to said second edge energy image to create a second edge energy map;

e) assigning edge probability values to said second directionality image to create a second directionality-based edge map;

f) combining said first edge energy map and said second edge energy map a combined edge energy map;

g) combining said first directionality-based edge map with said second directionality-based edge map to create a combined directionality-based edge map; and

h) wherein said combining said first edge energy map with said first directionality-based edge map is replaced with combining said combined edge energy map with said combined directionality-based edge map to create a combined edge map.

19. A method as described in claim 17 wherein said performing a directionality check comprises calculating auto-correlation values in 8 directions.