Hybride denoising of images and videos based on interest metrics

Patent metadata

Field	Value
Publication number	US-2018182072-A1
Application number	US-201715850547-A
Country	US
Kind code	A1
Filing date	Dec 21, 2017
Priority date	Dec 22, 2016
Publication date	Jun 28, 2018
Grant date	—

Abstract

Disclosed is a method for processing noise in a digital image having multiple image portions, including: (a) Predefining: criteria of interest for denoising selected details of any image portion of the digital image; a plurality of possible processing procedures to be applied to an image portion in order to denoise the selected details, each processing procedure having an efficiency related to an associated complexity level, the possible processing procedures being ordered by increasing complexity level; (b) For each portion of the image: analyzing the image portion to quantify the presence of one or more of the selected details in the image portion, and calculating an overall interest of the image portion as a function of respective quantifications of the presences of the selected details; comparing the overall interest at the complexity levels, in order to launch the processing procedure having the complexity level corresponding to the calculated overall interest.

First claim

1 . A method, implemented by computer means, for processing noise in a digital image having a plurality of image portions, the method comprising:
a) Predefining:
criteria of interest for denoising selected details of any image portion of the digital image;
a plurality of possible processing procedures to be applied to an image portion in order to denoise said selected details, each processing procedure having an efficiency related to an associated complexity level, said possible processing procedures being ordered by increasing complexity level;
b) For each portion of the image:
analyzing said image portion to quantify the presence of one or more of said selected details in the image portion, and calculating an overall interest of the image portion as a function of respective quantifications of the presences of said selected details,
comparing the overall interest at said complexity levels, in order to launch the processing procedure having the complexity level corresponding to the calculated overall interest.

2 . The method according to claim 1 , wherein the launched processing procedure is the one having the complexity level immediately above the calculated overall interest or having a maximum complexity level if the overall interest is greater than the maximum complexity level.

3 . The method according to claim 1 , wherein each processing procedure is defined by at least one processing type combined with a parameter set associated with the processing type, each possible combination of processing type and associated parameter set having a processing efficiency related to a complexity level pre-estimated in step a), said possible combinations being ordered by increasing complexity level in step a).

4 . The method according to claim 3 , wherein each processing procedure is defined by several combined processing types, each processing type having an associated parameter set, each possible combination of processing types having an overall processing efficiency related to a complexity level pre-estimated in step a), said possible combinations being ordered by increasing complexity level in step a).

5 . The method according to claim 1 , wherein, in step a), thresholds S 1 ,S 2 , . . . ,S |C|−1 are predefined, each threshold being associated with a processing procedure providing a desired denoising efficiency for an associated complexity level, and in step b), the overall interest is compared to these thresholds.

6 . The method according to claim 1 , wherein, in step b), K interest metrics M 1 ,M 2 , . . . ,M K , respectively associated with the selected details are calculated, each metric being specific to one of said criteria of interest predefined in step a).

7 . The method according to claim 6 , wherein, in step b), the overall interest is calculated as a function of the K interest metrics IG=f(M 1 , M 2 , . . . , M K ).

8 . The method according to claim 1 , wherein the digital image is an image divided into N regions having the same dimensions.

9 . The method according to claim 1 , wherein the digital image is an image divided into N regions having different dimensions.

10 . A non-transitory computer readable storage medium storing a computer-executable program comprising instructions for implementing the method according to claim 1 .

11 . A Noise processing device adapted to denoise a digital image comprising a plurality of image portions, the device comprising:
a processor configured to carry out the following operations: a) Predefining:
criteria of interest for denoising selected details of any image portion of the digital image;
a plurality of possible processing procedures to be applied to an image portion in order to denoise said selected details, each processing procedure having an efficiency related to an associated complexity level, said possible processing procedures being ordered by increasing complexity level;
b) For each portion of the image:
analyzing said image portion in order to quantify the presence of one or more of said selected details in the image portion, and calculating an overall interest of the image portion as a function of respective quantifications of the presences of said selected details,
comparing the overall interest at said complexity levels, in order to launch the processing procedure having the complexity level corresponding to the calculated overall interest.

Assignees

• Ateme

Inventors

• Mora Elie Gabriel
• Larbier Pierre

Classifications

• G06T7/11

  Region-based segmentation · CPC title

• G06T2207/20021

  Dividing image into blocks, subimages or windows · CPC title

• G06T2207/20012

  Locally adaptive · CPC title

• G06T2207/20164

  Salient point detection; Corner detection · CPC title

• G06T7/136

  involving thresholding · CPC title