Method and system for denoising and demosaicing artifact suppression in digital images

1. A method for processing a raw digital image, comprising:

employing a hardware processor to perform:

denoising the raw digital image, comprising:

(a) selecting a first pixel (p) of the raw digital image;

(b) selecting a second pixel (q) of the raw digital image;

(c) determining a first plurality of pixel patches, each pixel patch in the first plurality of pixel patches having a predetermined pixel patch shape and a predetermined pixel patch size and containing the first pixel (p);

(d) determining a second plurality of pixel patches, each pixel patch in the second plurality of pixel patches containing the second pixel (q), and corresponding in shape and size to a respective pixel patch of the first plurality of pixel patches;

(e) determining a patch distance between each pair of corresponding pixel patches in the first plurality of pixel patches and the second plurality of pixel patches;

(f) determining an effective distance between the first pixel (p) and the second pixel (q) based on determined patch distances for all pairs of corresponding pixel patches in the first plurality of pixel patches and the second plurality of pixel patches;

(g) repeating the steps (a) to (f) for the same first pixel (p) and other second pixels (q) until a predetermined number of second pixels (q) in the raw digital image is processed; and

(h) denoising the first pixel (p), comprising determining respective contributions of the second pixels (q) into a noise reduction of the first pixel (p), using respective effective distances of the second pixels (q);

and

removing high frequency artifacts from the first pixel (p), comprising:

determining an average for each pixel patch in the first plurality of pixel patches;

forming a multi-dimensional guide for the first pixel (p), comprising the average for the pixel patches as components of the multi-dimensional guide; and

applying a Bilateral Filter method with the multi-dimensional guide to the first pixel (p), thereby denoising and removing the high frequency artifacts.

2. The method of claim 1 , comprising repeating the steps (a) to (h) for other first pixels (p) until a predetermined number of first pixels (p) in the raw digital image is processed, thereby obtaining a denoised raw digital image.

3. The method of claim 2 , wherein:

the predetermined number of the second pixels (q) comprises all pixels (q) in the raw digital image outside of the first plurality of pixel patches; and

the predetermined number of the first pixels (p) comprises all pixels (p) in the raw digital image outside of the second plurality of pixel patches.

4. The method of claim 1 , wherein the predetermined pixel patch shape is one of the following:

a square;

a rectangle;

a circle;

an oval;

a hexagon;

an octagon; or

a polygon.

5. The method of claim 1 , wherein the predetermined pixel patch shape is a square, and a predetermined pixel patch size is from 2×2 pixels to 3×3 pixels.

6. The method of claim 1 , wherein the predetermined pixel patch shape is a square, and a predetermined pixel patch size is from about 2×2 pixels to about 8×8 pixels.

7. The method of claim 1 , wherein the predetermined pixel patch shape is a square, and a predetermined pixel patch size is from about 5×5 pixels to about 16×16 pixels.

8. The method of claim 1 , wherein the step (f) comprises determining the effective distance as an average patch distance among all patch distances.

9. The method of claim 1 , wherein the step (f) comprises determining the effective distance as a maximum patch distance among all patch distances.

10. The method of claim 1 , wherein the step of denoising (h) comprises applying one of:

a Non-Local Means method; or

a Bilateral Filter method.

11. The method of claim 1 , wherein:

the predetermined shape of the pixel patch is a rectangle;

a neighborhood of pixels surrounding the first pixel (p) used for denoising the first pixel (p) has a rectangular shape; and

the step (h) of denoising comprising comprises applying an integral images method.

12. The method of claim 1 , wherein the raw digital image is a Bayer's pattern image.

13. The method of claim 1 , wherein the steps (a) and (b) comprise selecting the first and second pixels of a same type.

14. A system for processing a raw digital image, comprising:

a processor and a non-transitory computer readable storage medium having computer readable instructions stored thereon for execution by the processor, causing the processor to:

denoise the raw digital image, causing the processor to:

(a) select a first pixel (p) of the raw digital image;

(b) select a second pixel (q) of the raw digital image;

(c) determine a first plurality of pixel patches, each pixel patch in the first plurality of pixel patches having a predetermined pixel patch shape and a predetermined pixel patch size and containing the first pixel (p);

(d) determine a second plurality of pixel patches, each pixel patch in the second plurality of pixel patches containing the second pixel (q), and corresponding in shape and size to a respective pixel patch of the first plurality of pixel patches;

(e) determine a patch distance between each pair of corresponding pixel patches in the first plurality of pixel patches and the second plurality of pixel patches;

(f) determine an effective distance between the first pixel (p) and the second pixel (q) based on determined patch distances for all pairs of corresponding pixel patches in the first plurality of pixel patches and the second plurality of pixel patches;

(g) repeat the steps (a) to (f) for the same first pixel (p) and other second pixels (q) until a predetermined number of second pixels (q) in the raw digital image is processed; and

(h) denoise the first pixel (p), comprising determining respective contributions of the second pixels (q) into a noise reduction of the first pixel (p), using respective effective distances of the second pixels (q);

and

remove high frequency artifacts from the first pixel (p), causing the processor to:

determine an average for each pixel patch in the first plurality of pixel patches;

form a multi-dimensional guide for the first pixel (p), comprising the average for the pixel patches as components of the multi-dimensional guide; and

apply a Bilateral Filter method with the multi-dimensional guide to the first pixel (p), thereby denoising and removing the high frequency artifacts.

15. The system of claim 14 , wherein the instructions cause the processor to repeat the steps (a) to (h) for other first pixels (p) until a predetermined number of first pixels (p) in the raw digital image is processed, thereby obtaining a denoised raw digital image.

16. The system of claim 15 , wherein:

the predetermined number of second pixels (q) comprises all pixels (q) in the raw digital image outside of the first plurality of pixel patches; and

the predetermined number of first pixels (p) comprises all pixels (p) in the raw digital image outside of the second plurality of pixel patches.

17. The system of claim 14 , wherein the predetermined pixel patch shape is one of the following:

a square;

a rectangle;

a circle;

an oval;

a hexagon;

an octagon; or

a polygon.

18. The system of claim 14 , wherein the predetermined pixel patch shape is a square, and a predetermined pixel patch size is from 2×2 pixels to 3×3 pixels.

19. The system of claim 14 , wherein the predetermined pixel patch shape is a square, and a predetermined pixel patch size is from about 2×2 pixels to about 8×8 pixels.

20. The system of claim 14 , wherein the predetermined pixel patch shape is a square, and a predetermined pixel patch size is from about 5×5 pixels to about 16×16 pixels.

21. The system of claim 14 , wherein the step (f) comprises determining the effective distance as an average patch distance among all patch distances.

22. The system of claim 14 , wherein the step (f) comprises determining the effective distance as a maximum patch distance among all patch distances.

23. The system of claim 14 , wherein the instructions to denoise the first pixel further cause the processor to apply one of:

a Non-Local Means method; or

a Bilateral Filter method.

24. The system of claim 14 , wherein:

the predetermined shape of the pixel patch is a rectangle;

a neighborhood of pixels surrounding the first pixel (p) used for denoising the first pixel (p) has a rectangular shape; and

the instructions to denoise the first pixel comprise applying an integral images method.

25. The system of claim 14 , wherein the raw digital image is a Bayer's pattern image.

26. The system of claim 14 , wherein the instructions cause the processor to select the first and second pixels of a same type in steps (a) and (b).