Method for image processing and apparatus for implementing the same

Patent metadata

Field	Value
Publication number	US-11240521-B2
Application number	US-202117307364-A
Country	US
Kind code	B2
Filing date	May 4, 2021
Priority date	May 4, 2020
Publication date	Feb 1, 2022
Grant date	Feb 1, 2022

Abstract

A method of method of processing an image includes: determining estimates of parameters of an auto-regressive, parametric model of noise, according to which a current noise pixel is computed combining linear combination of previous noise pixels in a causal neighborhood of the current noise pixel weighted by respective model linear combination parameters with a generated noise sample corresponding to an additive Gaussian noise of model variance parameter; performing a convergence check loop, each iteration including: generating a noise template of noise pixels based on the estimated model parameters, the noise template having predetermined pixel size smaller than the image pixel size; estimating a noise template variance; if the estimated variance is below a first predetermined threshold or above a second predetermined threshold, proportionally decreasing the model linear combination parameters with a predetermined correcting factor, and performing another convergence check loop; otherwise exiting the convergence check loop.

First claim

The invention claimed is:

1. A method of processing an image, the method comprising:
determining estimates of parameters of an auto-regressive, AR, parametric model of noise contained in the image, according to which a current noise pixel is computed as a combination of a linear combination of P previous noise pixels in a causal neighborhood of the current noise pixel weighted by respective AR model linear combination parameters (φ 1 , . . . , φ P ) with a generated noise sample corresponding to an additive Gaussian noise of AR model variance parameter (σ);
performing a convergence check loop, wherein each iteration of the convergence check loop comprises:
generating a noise template of noise pixels based on the estimated AR model parameters, wherein the noise template is of a predetermined pixel size smaller than the pixel size of the image;
determining an estimate (σ P ) of a variance of the noise template;
if the estimated variance (σ P ) is below a first predetermined threshold (T min ) or above a second predetermined threshold (T max ), proportionally decreasing one or more of the AR model linear combination parameters (φ 1 , . . . , φ P ) with a predetermined correcting factor, and performing a new iteration of the convergence check loop;
otherwise exiting the convergence check loop.

2. The method according to claim 1 , wherein the AR model is configured to model grain contained in the image.

3. The method according to claim 1 , performed at an encoder configured for encoding the image, wherein the predetermined pixel size of the noise template is chosen corresponding to the pixel size of a noise synthesis template used at a decoder to synthesize film grain based on the AR model.

4. The method according to claim 1 , wherein the predetermined pixel size of the noise template is chosen to be 64×64 pixels.

5. The method according to claim 1 , wherein proportionally decreasing the AR model linear combination parameters (φ 1 , . . . , φ P ) with the predetermined correcting factor comprises dividing each of the AR model linear combination parameters (φ 1 , . . . , φ P ) by the predetermined correcting factor, wherein the predetermined correcting factor is greater than 1.

6. The method according to claim 1 , wherein the predetermined correcting factor is equal to 1.4.

7. The method according to claim 1 , wherein the first predetermined threshold (T min ) is defined based on the additive Gaussian noise of AR model variance parameter (σ).

8. The method according to claim 1 , wherein the second predetermined threshold (T max ) is defined based on the additive Gaussian noise of AR model variance parameter (σ).

9. The method according to claim 1 , wherein the first predetermined threshold (T min ) is defined as comprising the additive Gaussian noise of AR model variance parameter (σ) divided by a first predetermined scaling factor (K min ).

10. The method according to claim 1 , wherein the second predetermined threshold (T max ) is defined as comprising the additive Gaussian noise of AR model variance parameter (σ) multiplied by a second predetermined scaling factor (K max ).

11. The method according to claim 1 , wherein at least one of the AR model linear combination parameters (φ 1 , . . . , φ P ) is preset to zero.

12. The method according to claim 11 , wherein a number of AR model linear combination parameters that are preset to zero is chosen based on a pixel resolution of the image.

13. An apparatus, the apparatus comprising a processor and a memory operatively coupled to the processor, wherein the processor is configured to perform a method of processing an image, the method comprising:
determining estimates of parameters of an auto-regressive, AR, parametric model of noise contained in the image, according to which a current noise pixel is computed as a combination of a linear combination of P previous noise pixels in a causal neighborhood of the current noise pixel weighted by respective AR model linear combination parameters (φ 1 , . . . , φ P ) with a generated noise sample corresponding to an additive Gaussian noise of AR model variance parameter (σ);
performing a convergence check loop, wherein each iteration of the convergence check loop comprises:
generating a noise template of noise pixels based on the estimated AR model parameters, wherein the noise template is of a predetermined pixel size smaller than the pixel size of the image;
determining an estimate (φ P ) of a variance of the noise template;
if the estimated variance (φ P ) is below a first predetermined threshold (T min ) or above a second predetermined threshold (T max ), proportionally decreasing one or more of the AR model linear combination parameters (φ 1 , . . . , φ P ) with a predetermined correcting factor, and performing a new iteration of the convergence check loop;
otherwise exiting the convergence check loop.

14. A video encoder, configured to encode video content comprising a plurality of images, the video encoder comprising an apparatus according to claim 13 configured to process images of an input video.

15. The apparatus according to claim 13 , wherein the AR model is configured to model grain contained in the image.

16. The apparatus according to claim 13 , wherein the first predetermined threshold (T min ) is defined based on the additive Gaussian noise of AR model variance parameter (σ).

17. The apparatus according to claim 13 , wherein the second predetermined threshold (T max ) is defined as comprising the additive Gaussian noise of AR model variance parameter (σ) multiplied by a second predetermined scaling factor (K max ).

18. A non-transitory computer-readable medium encoded with executable instructions which, when executed, causes an apparatus comprising a processor operatively coupled with a memory, to perform a method of processing an image, the method comprising:
determining estimates of parameters of an auto-regressive, AR, parametric model of noise contained in the image, according to which a current noise pixel is computed as a combination of a linear combination of P previous noise pixels in a causal neighborhood of the current noise pixel weighted by respective AR model linear combination parameters (φ 1 , . . . , φ P ) with a generated noise sample corresponding to an additive Gaussian noise of AR model variance parameter (σ);
performing a convergence check loop, wherein each iteration of the convergence check loop comprises:
generating a noise template of noise pixels based on the estimated AR model parameters, wherein the noise template is of a predetermined pixel size smaller than the pixel size of the image;
determining an estimate (σ P ) of a variance of the noise template;
if the estimated variance (σ P ) is below a first predetermined threshold (T min ) or above a second predetermined threshold (T max ), proportionally decreasing one or more of the AR model linear combination parameters (φ 1 , . . . , φ P ) with a predetermined correcting factor, and performing a new iteration of the convergence check loop;
otherwise exiting the convergence check loop.

19. The non-transitory computer-readable medium according to claim 18 , wherein the AR model is configured to model grain contained in the image.

20. The non-transitory computer-readable medium according to claim 18 , wherein the first predetermined threshold (T min ) is defined based on the additive Gaussian noise of AR model variance parameter (σ).

Assignees

• Ateme

Inventors

• Guionnet Thomas
• Herbreteau Sébastien

Classifications

• H04N19/85 Primary

  using pre-processing or post-processing specially adapted for video compression · CPC title

• H04N19/117

  Filters, e.g. for pre-processing or post-processing (sub-band filter banks H04N19/635) · CPC title

• H04N19/46 Primary

  Embedding additional information in the video signal during the compression process (H04N19/517, H04N19/68, H04N19/70 take precedence) · CPC title

• H04N19/177

  the unit being a group of pictures [GOP] · CPC title

• H04N19/179

  the unit being a scene or a shot · CPC title