Claims:WE CLAIM:
1. A method for de-noising images, the method comprising:
creating, by an image de-noising device, a single dimensional vector for an image through a multi-layer neural network;
converting, by the image de-nosing device, the single dimensional vector into a multi-dimensional matrix based on number of layers in the multi-layer neural network;
generating, by the image de-nosing device, a feature hierarchy based on the multi-dimensional matrix, wherein the feature hierarchy comprises a plurality of levels, and wherein each level in the plurality of levels comprises at least one feature associated with the image;
creating, by the image de-nosing device, a plurality of segments for the image based on the feature hierarchy, wherein each of the plurality of segments comprises a set of features associated with the image; and
removing, by the image de-nosing device, each segment comprising noise from the plurality of segments to generate a de-noised image.

2. The method of claim 1, wherein the feature hierarchy comprises a root node representing the image at a first level in the plurality of levels, and wherein features in remaining plurality of levels combine to form the image at the first level.

3. The method of claim 1, wherein creating the plurality of segments for the image comprises determining semantic information associated with each of a plurality of features associated with the image.

4. The method of claim 3 further comprising grouping semantically related features to form each of the plurality of segments.

5. The method of claim 1 further comprising classifying each of the plurality of segments as one of a segment comprising noise and a segment not comprising noise, wherein each segment comprising noise is removed and each segment not comprising noise is retained.

6. The method of claim 1 further comprising validating the de-noised image by comparing vector representation of the de-noised image with vector representation of an original image.

7. The method of claim 6 further comprising:
creating a single dimensional vector for the de-noised image through the multi-layer neural network;
converting the single dimensional vector for the de-noised image into a multi-dimensional matrix based on number of layers in the multi-layer neural network;
generating a feature hierarchy based on the multi-dimensional matrix, wherein the feature hierarchy comprises a plurality of levels, and wherein each level in the plurality of levels comprises at least one feature associated with the de-noised image;
creating a plurality of segments for the de-noised image based on the feature hierarchy, wherein each of the plurality of segments comprises a set of features associated with the de-noised image; and
removing each segment comprising noise from the plurality of segments.

8. The method of claim 7, wherein the de-noised image is iteratively validated to completely remove noise from the image.

9. The method of claim 1, wherein removing each segment comprising noise from the plurality of segments comprises up-sampling the remaining plurality of segments after removal of each segment comprising noise.

10. The method of claim 1 further comprising providing the de-noised image to a user as an output image.

11. An image de-noising device for removing noise from an image, the system image de-noising device comprising:
a processor; and
a memory communicatively coupled to the processor, wherein the memory stores processor instructions, which, on execution, causes the processor to:
create, by an image de-noising device, a single dimensional vector for an image through a multi-layer neural network;
convert, by the image de-nosing device, the single dimensional vector into a multi-dimensional matrix based on number of layers in the multi-layer neural network;
generate, by the image de-nosing device, a feature hierarchy based on the multi-dimensional matrix, wherein the feature hierarchy comprises a plurality of levels, and wherein each level in the plurality of levels comprises at least one feature associated with the image;
create, by the image de-nosing device, a plurality of segments for the image based on the feature hierarchy, wherein each of the plurality of segments comprises a set of features associated with the image; and
remove, by the image de-nosing device, each segment comprising noise from the plurality of segments to generate a de-noised image.

12. The image de-noising device of claim 11, wherein creating the plurality of segments for the image comprises determining semantic information associated with each of a plurality of features associated with the image.

13. The image de-noising device of claim 12, wherein processor instructions further cause the processor to group semantically related features to form each of the plurality of segments.

14. The image de-noising device of claim 11, wherein processor instructions further cause the processor to classifying each of the plurality of segments as one of a segment comprising noise and a segment not comprising noise, wherein each segment comprising noise is removed and each segment not comprising noise is retained.

15. The image de-noising device of claim 11, wherein processor instructions further cause the processor to validate the de-noised image by comparing vector representation of the de-noised image with vector representation of an original image.

16. The image de-noising device of claim 15, wherein processor instructions further cause the processor to:
create a single dimensional vector for the de-noised image through the multi-layer neural network;
convert the single dimensional vector for the de-noised image into a multi-dimensional matrix based on number of layers in the multi-layer neural network;
generate a feature hierarchy based on the multi-dimensional matrix, wherein the feature hierarchy comprises a plurality of levels, and wherein each level in the plurality of levels comprises at least one feature associated with the de-noised image;
create a plurality of segments for the de-noised image based on the feature hierarchy, wherein each of the plurality of segments comprises a set of features associated with the de-noised image; and
remove each segment comprising noise from the plurality of segments.

17. The image de-noising device of claim 16, wherein the de-noised image is iteratively validated to completely remove noise from the image.

18. The image de-noising device of claim 11, wherein processor instructions further cause the processor to remove each segment comprising noise from the plurality of segments comprises up-sampling the remaining plurality of segments after removal of each segment comprising noise.

19. The image de-noising device of claim 11, wherein processor instructions further cause the processor to provide the de-noised image to a user as an output image.

Dated this 15th day of May, 2018

Swetha SN
Of K&S Partners
Agent for the Applicant
IN/PA-2123
, Description:Technical Field
This disclosure relates generally to processing images, and more particularly to method and device for de-noising images.