IMAGE DENOISING METHOD

• US 20110069902A1
• Filed: 07/02/2010
• Published: 03/24/2011
• Est. Priority Date: 09/21/2009
• Status: Active Grant

• Current Assignee: PixArt Imaging Incorporated
• Original Assignee: PixArt Imaging Incorporated
• Inventors: Yang, Shu Sian, Yao, Wen Han, Lee, Chuan Hsin
• Granted Patent: US 8,885,965 B2
• Time in Patent Office: Days
• US Class Current: 382/260
• CPC Class Codes: G06T 2207/20012 Locally adaptive
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing

Abstract

An image denoising method includes the steps of: sequentially selecting a pixel in an image as a current pixel; dynamically determining a current search block and a strength parameter; pre-denoising the comparison block of each pixel in the current search block; comparing the comparison block of the pre-denoised neighborhood pixel and the comparison block of the pre-denoised current pixel to obtain a similarity between each neighborhood pixel and the current pixel in the current search block; determining a weighting of each neighborhood pixel related to the current pixel according to the strength parameter, and a distance and the similarity between each neighborhood pixel and the current pixel in the current search block; and weighted averaging each neighborhood pixel and the current pixel in the current search block according to the weighting to obtain a reconstruction value of the current pixel.

Claims

1. An image denoising method, comprising the steps of:

sequentially selecting a pixel in an image as a current pixel, wherein the pixels around the current pixel are defined as neighborhood pixels;

dynamically determining a current search block enclosing the current pixel and a strength parameter and determining a comparison block for each of the pixels in the current search block, wherein the comparison block encloses the each pixel;

pre-denoising the comparison blocks for the pixels in the current search block;

comparing the pre-denoised comparison blocks for the neighborhood pixels with the pre-denoised comparison block for the current pixel to obtain a similarity between each of the neighborhood pixels and the current pixel in the current search block;

determining a weighting for each of the neighborhood pixels related to the current pixel according to the strength parameter, the similarity and a distance between each of neighborhood pixels and the current pixel in the current search block; and

weighted averaging each of the neighborhood pixels and the current pixel in the current search block according to the weighting to obtain a reconstruction value of the current pixel.

2. The image denoising method as claimed in claim 1, further comprising:

determining whether the reconstruction values for all the pixels in the image have be obtained.

3. The image denoising method as claimed in claim 1, wherein the pre-denoising is performed by a mean filter or a neighborhood filter.

4. The image denoising method as claimed in claim 1, wherein the similarity is a sum of the squares of the differences between the pixels in the pre-denoised comparison block for the neighborhood pixel and the corresponding pixels in the pre-denoised comparison block for the current pixel.

5. The image denoising method as claimed in claim 1, wherein the step of dynamically determining a current search block enclosing the current pixel and a strength parameter comprises:

determining a maximal search block enclosing the current pixel and a comparison block for each of the pixels in the maximal search block, wherein the comparison block encloses the each pixel;

calculating the ratios of the edge pixels for the sub-search blocks in the maximal search block; and

selecting the sub-search block that has the smallest ratio of the edge pixels as the current search block and determining the strength parameter according to the ratio of the edge pixels of the current search block.

6. The image denoising method as claimed in claim 5, wherein the step of calculating the ratios of the edge pixels for the sub-search blocks in the maximal search block comprises:

pre-denoising the comparison blocks for the pixels in the maximal search block;

calculating a sum of the absolute values of the differences between the pixels in the comparison blocks not to be pre-denoised for the pixels in the maximal search block and the corresponding pixels in the pre-denoised comparison blocks;

determining the pixel to be an edge pixel when the sum of the absolute values of the differences is greater than a predetermined threshold value; and

calculating the ratios of the edge pixels for the sub-search blocks in the maximal search block according to the determined result.

7. The image denoising method as claimed in claim 6, wherein the pre-denoising is performed by a mean filter or a neighborhood filter.

8. The image denoising method as claimed in claim 6, wherein the sub-search blocks have sizes of (n−

2m)×

(n−

2m), n is the side length of the maximal search block and n is zero or an integer.

9. The image denoising method as claimed in claim 8, wherein the smallest sub-search block has a size of 3×

3 and the largest sub-search block is equal to the maximal search block.

10. The image denoising method as claimed in claim 5, wherein the largest sub-search block has a size of 7×

7.

11. The image denoising method as claimed in claim 5, wherein when the ratio of edge pixels is higher, the strength parameter is smaller, and when the ratio of edge pixels is lower, the strength parameter is larger.

12. The image denoising method as claimed in claim 1, wherein the current search block and comparison block have a shape of square, rectangle, rhombus, circle or ellipse.

13. An image denoising method, comprising the steps of:

sequentially selecting a pixel in an image as a current pixel, wherein the pixels around the current pixel are defined as neighborhood pixels;

determining a maximal search block enclosing the current pixel and a comparison block for each of the pixels in the maximal search block, wherein the comparison block encloses the each pixel;

pre-denoising the comparison blocks for the pixels in the maximal search block;

determining whether the pixels are edge pixels according to the comparison blocks not to be pre-denoised and the pre-denoised comparison blocks for the pixels in the maximal search block;

calculating the ratios of the edge pixels for the sub-search blocks in the maximal search block;

selecting the sub-search block that has the smallest ratio of the edge pixels as a current search block and determining a strength parameter according to the ratio of the edge pixels of the current search block;

comparing the pre-denoised comparison blocks for the neighborhood pixels with the pre-denoised comparison block for the current pixel to obtain a similarity between each of the neighborhood pixels and the current pixel in the current search block;

determining a weighting for each of the neighborhood pixels related to the current pixel according to the strength parameter, the similarity and a distance between each of neighborhood pixels and the current pixel in the current search block; and

weighted averaging each of the neighborhood pixels and the current pixel in the current search block according to the weighting to obtain a reconstruction value of the current pixel.

14. The image denoising method as claimed in claim 13, wherein the pre-denoising is performed by a mean filter or a neighborhood filter.

15. The image denoising method as claimed in claim 13, wherein the step of determining whether the pixels are edge pixels according to the comparison blocks not to be pre-denoised and the pre-denoised comparison blocks for the pixels in the maximal search block comprises:

calculating a sum of the absolute values of the differences between the pixels in the comparison blocks not to be pre-denoised for the pixels in the maximal search block and the corresponding pixels in the pre-denoised comparison blocks; and

determining the pixel to be an edge pixel when the sum of the absolute values of the differences is greater than a predetermined threshold value.

16. The image denoising method as claimed in claim 13, wherein when the ratio of edge pixels is higher, the strength parameter is smaller, and when the ratio of edge pixels is lower, the strength parameter is larger.

17. The image denoising method as claimed in claim 13, wherein the sub-search blocks have sizes of (n−

2m)×

(n−

2m), n is the side length of the maximal search block and n is zero or an integer.

18. The image denoising method as claimed in claim 17, wherein the smallest sub-search block has a size of 3×

3 and the largest sub-search block is equal to the maximal search block.

19. The image denoising method as claimed in claim 13, wherein the similarity is a sum of the squares of the differences between the pixels in the pre-denoised comparison block for the neighborhood pixel and the corresponding pixels in the pre-denoised comparison block for the current pixel.

20. The image denoising method as claimed in claim 13, further comprising:

determining whether the reconstruction values for all the pixels in the image have be obtained.

21. The image denoising method as claimed in claim 13, wherein the maximal search block, comparison block, sub-search block and current search block have a shape of square, rectangle, rhombus, circle or ellipse.

22. The image denoising method as claimed in claim 13, wherein the strength parameter is a mathematical function or a numeric value.