Method and system for DNN based imaging

1. A method of DNN-based image processing by an imaging device, the method comprising:

obtaining, by the imaging device, at least one input image with a plurality of color channels;

extracting, by the imaging device and through a first branch at a first resolution, a semantic information from each of the color channels of the at least one input image;

extracting, by the imaging device and through a second branch at the first resolution, a color correlation information from each of the color channels of the at least one input image; and

generating, by the imaging device, at least one output image having an image quality that is enhanced compared to the input image, based on the semantic information that was extracted through the first branch and the color correlation information that was extracted through the second branch.

2. The method of claim 1 , further comprising performing, by the imaging device, a channel-wise fusion of the semantic information and the color correlation information.

3. The method of claim 2 , further comprising generating one or more semantic filters corresponding to each of the color channels, and

wherein the performing the channel-wise fusion comprises fusing each of the one or more semantic filters with the extracted color correlation information corresponding to each of the color channels.

4. The method of claim 2 , wherein the performing the channel-wise fusion of the semantic information and the color correlation information comprises:

performing, by the imaging device, a fusion of pixels of respective channels from the semantic information and the color correlation information for each of the channels; and

generating, by the imaging device, learned maps between the semantic information and the color correlation information for each of the channels based on the fusion.

5. The method of claim 2 , wherein the generating at least one output image comprises:

generating predictions, by the imaging device, based on the channel-wise fusion of the semantic information and the color correlation information;

correcting, by the imaging device, the at least one input image based on the predicted values generated from the channel-wise fusion operation; and

generating, by the imaging device, the at least one output image based on the correction.

6. The method of claim 1 , wherein the extracting the semantic information and the color correlation information comprises extracting the semantic information and the color correlation information independently by using a respectively separate neural network.

7. The method of claim 1 , wherein the extracting the semantic information comprises extracting the semantic information by using Depth-Wise convolution.

8. The method of claim 1 , wherein the extracting the color correlation information comprises extracting the color correlation information by using Point-Wise convolution.

9. An imaging device for DNN-based image processing, the imaging device comprising:

a memory;

a processor coupled to the memory and configured to:

obtain at least one input image with a plurality of color channels;

extract, through a first branch at a first resolution, a semantic information from each of the color channels of the at least one image extract, through a second branch at the first resolution, a color correlation information from each of the color channels of the at least one image; and

generate at least one output image having an image quality that is enhanced compared to the input image, based on the semantic information that was extracted through the first branch and the color correlation information that was extracted through the second branch.

10. The imaging device of claim 9 , wherein the processor is further configured to perform a channel-wise fusion of the semantic information and the color correlation information.

11. The imaging device of claim 10 , wherein the processor is further configured to generate at least one semantic filters corresponding to each of the color channels; and

wherein the processor is further configured to fuse each of the at least one semantic filters with the extracted color correlation information corresponding to each of the color channels.

12. The imaging device of claim 10 , wherein the processor is configured to

generate prediction, based on the channel-wise fusion of the semantic information and the color correlation information;

correct the at least one input image based on the predicted values generated from the channel-wise fusion operation; and

generate the at least one output image based on the correction.

13. The imaging device of claim 9 , wherein the processor is configured to extract the semantic information and the color correlation information independently by using a respectively separate neural network.

14. The imaging device of claim 9 , wherein the processor is configured to extract the semantic information by using Depth-Wise convolution.

15. The imaging device of claim 9 , wherein the processor is configured to extract the color correlation information by using Point-Wise convolution.

16. The imaging device of claim 9 , wherein the processor is configured to

perform a fusion of pixels of respective channels from the semantic information and the color correlation information for each of the channels; and

generate learned maps between the semantic information and the color correlation information for each of the channels based on the fusion.

17. A non-transitory computer-readable recording medium having an executable program recorded thereon, wherein the program, when executed by at least one processor, instructs a computer to perform:

obtaining, by an imaging device, at least one input image with a plurality of color channels;

extracting, by the imaging device and through a first branch at a first resolution, a semantic information from each of the color channels of the at least one input image;

extracting, by the imaging device and through a second branch at the first resolution, a color correlation information from each of the color channels of the at least one input image; and

generating, by the imaging device, at least one output image having an image quality that is enhanced compared to the input image, based on the semantic information that was extracted through the first branch and the color correlation information that was extracted through the second branch.