Noise suppression in low light images
A low light noise reduction mechanism may perform denoising prior to demosaicing, and may also use parameters determined during the denoising operation for performing demosaicing. The denoising operation may attempt to find several patches of an image that are similar to a first patch, and use a weighted average based on similarity to determine an appropriate value for denoising a raw digital image. The same weighted average and similar patches may be used for demosaicing the same image after the denoising operation.
1. A method comprising:
receiving a raw digital image comprising pixel information;
identifying a first patch comprising a plurality of pixels;
identifying a plurality of patches similar to said first patch;
for each of said plurality of patches, determining a similarity between each of said plurality of patches and said first patch;
determining a weighted average for said plurality of patches based on said similarity; and
using said weighted average to determine a denoised version of said raw digital image.
2. The method of claim 1 further comprising:
using said weighted average to determine a demosaiced version of said denoised version of said raw digital image.
3. The method of claim 1 , said first patch being within a low light area of said image.
4. The method of claim 3 further comprising:
identifying a portion of said image being a low light portion and performing said method on said low light portion.
5. The method of claim 4 , said plurality of patches being within said low light portion.
6. The method of claim 4 , said method not being performed on a portion of said image outside of said low light portion.
7. The method of claim 1 , said first patch being a circular shaped patch.
8. The method of claim 1 , said first patch being a rectangular shaped patch.
9. The method of claim 1 , said first patch being greater than 16 pixels in size.
10. The method of claim 1 , said first patch being greater than 50 pixels in size.
11. A system comprising:
an image management system that includes one or more processors coupled to a memory, the one or more processors configured to receive a raw digital image comprising pixel information; and
an image enhancement system that:
identifies a first patch comprising a plurality of pixels within said raw digital image;
identifies a plurality of patches similar to said first patch;
for each of said plurality of patches, determines a similarity between each of said plurality of patches and said first patch;
determines a weighted average for said plurality of patches based on said similarity; and
uses said weighted average to determine a denoised version of said raw digital image.
12. The system of claim 11 , said image enhancement system that further:
uses said weighted average to determine a demosaiced version of said denoised version of said raw digital image.
13. The system of claim 11 , said image enhancement system that further:
analyzes said raw digital image to identify a low light area of said image.
14. The system of claim 13 , said first patch being within the low light area of said image.
15. The system of claim 14 , said image enhancement system that does not perform said identifies a first patch, identifies a plurality of patches similar to said first patch, determines a similarity, determines a weighted average, and uses said weighted average to determine a denoised version, on a portion of said image outside of said low light area.
16. The system of claim 11 , said image enhancement system that further:
determines that at least one of said plurality of patches does not have a similarity within a predefined similarity threshold and removes said at least one of said plurality of patches from said weighted average.
17. The system of claim 16 , said image enhancement system that further:
determines that an insufficient number of said plurality of patches are within said predefined similarity threshold and does not perform said denoising for said first patch, wherein the insufficient number is determined as being less than a predetermined minimum number.
18. A method comprising:
receiving a raw digital image comprising pixel information, said pixel information being Bayer filtered pixel information;
identifying a first patch comprising at least 16 pixels;
identifying a plurality of patches similar to said first patch;
for each of said plurality of patches, determining a similarity between each of said plurality of patches and said first patch;
determining a weighted average for said plurality of patches based on said similarity;
using said weighted average to determine a denoised version of said raw data; and
using said weighted average to determine a demosaiced version of said denoised version of said raw digital image.
19. The method of claim 18 further comprising:
for each of said plurality of patches, comparing said similarity to a predefined similarity threshold and removing those patches that do not meet said predefined similarity threshold.
20. The method of claim 19 further comprising:
determining that a minimum number of said plurality of patches do not meet said predefined similarity threshold and not denoising said first patch.