Systems and methods for processing low light images

1. A method of processing one or more low light images, the method comprising:

obtaining at least one short exposure image, the at least one short exposure image being associated with one or more shutter speeds;

obtaining at least one long exposure image, the at least one long exposure image being captured using a slower shutter speed than the one or more shutter speeds associated with the at least one short exposure image;

mapping color information from the at least one long exposure image to the at least one short exposure image by determining color-matched output values for pixels of the at least one short exposure image, a color-matched output value for a pixel in a short exposure image being determined by modifying a pixel value of the pixel in the short exposure image using one or more statistical representations of color values of the pixels of a long exposure image; and

generating an output image including the color-matched output values for the pixels of the at least one short exposure image.

2. The method of claim 1 , wherein mapping the color information from the at least one long exposure image to the at least one short exposure image includes applying an affine transform to the at least one short exposure image, the affine transform mapping the one or more statistical representations of the color values of the pixels of the long exposure image and one or more statistical representations of color values of the pixels of the short exposure image.

3. The method of claim 2 , wherein the one or more statistical representations of the color values of the pixels of the long exposure image and the one or more statistical representations of the color values of the pixels of the short exposure image are included in a three-dimensional tensor, the method further comprising applying one or more three-dimensional filters to the three-dimensional tensor.

4. The method of claim 2 , wherein the affine transform includes mean-variance matching, wherein the one or more statistical representations of the color values of the pixels of the short exposure image include a mean and a variance of the color values of the pixels of the short exposure image, and wherein the one or more statistical representations of the color values of the pixels of the long exposure image include a mean and a variance of the color values of the pixels of the long exposure image.

5. The method of claim 4 , further comprising generating a three-dimensional tensor, the three-dimensional tensor including a plurality of cells, each cell including at least one mean of a group of pixels from the short exposure image, at least one standard deviation of the group of pixels from the short exposure image, at least one mean of a group of pixels from the long exposure image, and at least one standard deviation of the group of pixels from the long exposure image.

6. The method of claim 5 , wherein the three-dimensional tensor includes a plurality of layers of cells, and wherein each layer of the three-dimensional tensor corresponds to a different range of luminance values.

7. The method of claim 1 , wherein the at least one short exposure image is single image captured using a shutter speed.

8. The method of claim 1 , wherein the at least one short exposure image is a noise-reduced short exposure image generated from one or more short exposure images captured using the one or more shutter speeds, and wherein mapping the color information from the at least one long exposure image to the at least one short exposure image includes mapping the color information from the at least one long exposure image to the noise-reduced short exposure image.

9. The method of claim 8 , further comprising:

generating the noise-reduced short exposure image, wherein generating the noise-reduced short exposure image includes processing the one or more short exposure images using a neural network trained to remove noise from the one or more short exposure images.

10. The method of claim 1 , further comprising:

receiving a plurality of short exposure images captured using the one or more shutter speeds; and

generating an averaged short exposure image, wherein generating the averaged short exposure image includes averaging pixels from the plurality of short exposure images.

11. The method of claim 10 , wherein the short exposure image is a noise-reduced short exposure image generated from the averaged short exposure image, and wherein mapping the color information from the at least one long exposure image to the at least one short exposure image includes mapping the color information from the at least one long exposure image to the noise-reduced short exposure image.

12. The method of claim 11 , further comprising:

generating the noise-reduced short exposure image, wherein generating the noise-reduced short exposure image includes processing the averaged short exposure image using a neural network trained to remove noise from the averaged short exposure image.

13. A method of processing one or more low light images, the method comprising:

obtaining one or more short exposure images, the one or more short exposure images being captured using one or more shutter speeds;

generating a noise-reduced short exposure image from the one or more short exposure images;

obtaining a long exposure image, the long exposure image being captured using a slower shutter speed than the one or more shutter speeds used to capture the one or more short exposure images;

mapping color information from the long exposure image to the noise-reduced short exposure image by determining color-matched output values for pixels of the noise-reduced short exposure image, a color-matched output value for a pixel in the noise-reduced short exposure image being determined by modifying a pixel value of the pixel in the noise-reduced short exposure image using one or more statistical representations of color values of the pixels of the long exposure image; and

generating an output image including the color-matched output values for the pixels of the noise-reduced short exposure image.

14. An apparatus for processing one or more low light images, comprising:

a memory configured to store data corresponding to one or more short exposure images and data corresponding to one or more long exposure images; and

a processor configured to:

obtain at least one short exposure image, the at least one short exposure image being associated with one or more shutter speeds;

obtain at least one long exposure image, the at least one long exposure image being captured using a slower shutter speed than the one or more shutter speeds associated with the at least one short exposure image;

map color information from the at least one long exposure image to the at least one short exposure image by determining color-matched output values for pixels in the at least one short exposure image, a color-matched output value for a pixel in a short exposure image being determined by modifying a pixel value of the pixel in the short exposure image using one or more statistical representations of color values of the pixels in a long exposure image; and

generate an output image including the color-matched output values for the pixels of the at least one short exposure image.

15. The apparatus of claim 14 , wherein mapping the color information from the at least one long exposure image to the at least one short exposure image includes applying an affine transform to the at least one short exposure image, the affine transform mapping the one or more statistical representations of the color values of the pixels in the long exposure image and one or more statistical representations of color values of the pixels in the short exposure image.

16. The apparatus of claim 15 , wherein the one or more statistical representations of the color values of the pixels in the long exposure image and the one or more statistical representations of the color values of the pixels in the short exposure image are included in a three-dimensional tensor, and wherein the processor is further configured to apply one or more three-dimensional filters to the three-dimensional tensor.

17. The apparatus of claim 15 , wherein the affine transform includes mean-variance matching, wherein the one or more statistical representations of the color values of the pixels in the short exposure image include a mean and a variance of the color values of the pixels in the short exposure image, and wherein the one or more statistical representations of the color values of the pixels in the long exposure image include a mean and a variance of the color values of the pixels in the long exposure image.

18. The apparatus of claim 17 , wherein the processor is further configured to generate a three-dimensional tensor, the three-dimensional tensor including a plurality of cells, each cell including at least one mean of a group of pixels from the short exposure image, at least one standard deviation of the group of pixels from the short exposure image, at least one mean of a group of pixels from the long exposure image, and at least one standard deviation of the group of pixels from the long exposure image.

19. The apparatus of claim 18 , wherein the three-dimensional tensor includes a plurality of layers of cells, and wherein each layer of the three-dimensional tensor corresponds to a different range of luminance values.

20. The apparatus of claim 14 , wherein the at least one short exposure image is single image captured using a shutter speed.

21. The apparatus of claim 14 , wherein the at least one short exposure image is a noise-reduced short exposure image generated from one or more short exposure images captured using the one or more shutter speeds, and wherein mapping the color information from the at least one long exposure image to the at least one short exposure image includes mapping the color information from the at least one long exposure image to the noise-reduced short exposure image.

22. The apparatus of claim 21 , wherein the processor is further configured to:

generate the noise-reduced short exposure image, wherein generating the noise-reduced short exposure image includes processing the one or more short exposure images using a neural network trained to remove noise from the one or more short exposure images.

23. The apparatus of claim 14 , wherein the processor is further configured to:

receive a plurality of short exposure images captured using the one or more shutter speeds; and

generate an averaged short exposure image, wherein generating the averaged short exposure image includes averaging pixels from the plurality of short exposure images.

24. The apparatus of claim 23 , wherein the short exposure image is a noise-reduced short exposure image generated from the averaged short exposure image, and wherein mapping the color information from the at least one long exposure image to the at least one short exposure image includes mapping the color information from the at least one long exposure image to the noise-reduced short exposure image.

25. The apparatus of claim 24 , wherein the processor is further configured to:

generate the noise-reduced short exposure image, wherein generating the noise-reduced short exposure image includes processing the averaged short exposure image using a neural network trained to remove noise from the averaged short exposure image.

26. The apparatus of claim 14 , wherein the apparatus comprises a mobile device.

27. The apparatus of claim 26 , further comprising one or more cameras for capturing the one or more short exposure images and the one or more long exposure images.

28. The apparatus of claim 27 , further comprising a display for displaying the output image.

29. An apparatus for processing one or more low light images, comprising:

a memory configured to store data corresponding to one or more short exposure images and data corresponding to one or more long exposure images; and

a processor configured to:

obtain one or more short exposure images, the one or more short exposure images being captured using one or more shutter speeds;

generate a noise-reduced short exposure image from the one or more short exposure images;

obtain a long exposure image, the long exposure image being captured using a slower shutter speed than the one or more shutter speeds used to capture the one or more short exposure images;

map color information from the long exposure image to the noise-reduced short exposure image by determining color-matched output values for pixels of the noise-reduced short exposure image, a color-matched output value for a pixel in the noise-reduced short exposure image being determined by modifying a pixel value of the pixel in the noise-reduced short exposure image using one or more statistical representations of color values of the pixels of the long exposure image; and

generating an output image including the color-matched output values for the pixels of the noise-reduced short exposure image.

30. The apparatus of claim 29 , wherein the apparatus comprises a mobile device, the mobile device including:

one or more cameras for capturing the one or more short exposure images and the one or more long exposure images; and

a display for displaying the output image.