Method and device for processing an image of pixels, corresponding computer program product and computer-readable medium

• US 10,055,673 B2
• Filed: 08/28/2016
• Issued: 08/21/2018
• Est. Priority Date: 08/28/2015
• Status: Expired due to Fees

• Current Assignee: Magnolia Licensing LLC (IP Edge LLC)
• Original Assignee: Thomson Licensing (Vantiva SA)
• Inventors: Burgos, Xavier, Fleureau, Julien, Le Clerc, Francois
• Primary Examiner(s): Grant, II, Jerome
• Application Number:
  US15/249,449
• Publication Number:
  US 20170061254A1
• Time in Patent Office: 723 Days
• Field of Search: 382159
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06F 18/24323 Tree-organised classifiers
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/20076 Probabilistic image processing
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection
• CPC Class Codes: G06T 7/77 using statistical methods
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

A method for processing an image of pixels is described. The method includes determining a plurality of features representative of the image, obtaining both information representative of a probability of presence of an object in the image and information representative of a shape of the object by implementing a unique multivariate regression applied to at least a part of the features.

Claims

1. A method for processing an image, the method comprising:

determining a plurality of features representative of said image;

obtaining both information representative of a probability of presence of an object in said image and information representative of a shape of said object by implementing a unique multivariate regression applied to at least a part of said features;

wherein said features are shape-indexed by using a reference shape as an input to said method and wherein said reference shape is projected on at least one sliding window, said at least one sliding window delimiting an area of said image where said unique multivariate regression is applied.

2. The method for processing an image according to claim 1, wherein said method is iterative, implementing a plurality of iterations, wherein, from the second iteration, each iteration uses the result of the previous iteration.

3. The method for processing an image according to claim 1, wherein said unique multivariate regression implements a cascade of regressors.

4. The method for processing an image according to claim 1, wherein said cascade of regressors implements an additive update of both said information representative of a probability of presence of an object in said image and of said information representative of a shape of said object.

5. The method for processing an image according to claim 1, wherein said unique multivariate regression implements at least one random forest regressor.

6. The method for processing an image according to claim 5, wherein said training is iterative and wherein, at a current iteration, each training images of said dataset of training images is weighted for its use during a following iteration, said weighting taking into account the result of said current iteration.

7. The method for processing an image according to claim 6, wherein said weighting is exponential.

8. The method for processing an image according to claim 1, wherein said method comprises a previous training of said unique multivariate regression, said training being performed by using a dataset of training images.

9. The method for processing an image according to claim 8, wherein said dataset of training images comprises:

a first set of training images, which provides a positive classifying result when being processed by said multivariate regression, and

a second set of training images, which provides a negative classifying result when being processed by said multivariate regression.

10. The method for processing an image according to claim 9, wherein said second set of training images comprises images, which comprise no object, and/or images where object landmark annotation is erroneous.

11. A non-transitory computer-readable medium comprising a computer program product recorded thereon and capable of being run by a processor, including program code instructions for implementing the steps of the method for detecting at least one fitting error according to claim 1 when it is executed by a processor.

12. A device for processing an image, said device comprising a processor configured to control:

a module for determining a plurality of features representative of said image;

a module for obtaining both information representative of a probability of presence of an object in said image and information representative of a shape of said object by implementing a unique multivariate regression applied to at least a part of said features;

wherein said features are shape-indexed according to a reference shape used as input for determining features of said representative image and wherein said reference shape is projected on at least one sliding window, said at least one sliding window delimiting an area of said image where said unique multivariate regression is applied.

13. The device for processing an image according to claim 12, wherein the processing is iterative, implementing a plurality of iterations, wherein, from the second iteration, each iteration uses the result of the previous iteration.

14. The device for processing an image according to claim 12, wherein said unique multivariate regression implements a cascade of regressors.

15. The device for processing an image according to claim 12, wherein said cascade of regressors implements an additive update of both said information representative of a probability of presence of an object in said image and of said information representative of a shape of said object.

16. The device for processing an image according to claim 12, wherein said unique multivariate regression implements at least one random forest regressor.

17. The device for processing an image of according to claim 16, wherein said training is iterative and wherein, at a current iteration, each training images of said dataset of training images is weighted for its use during a following iteration, said weighting taking into account the result of said current iteration.

18. The device for processing an image according to claim 17, wherein said weighting is exponential.

19. The device for processing an image according to claim 12, wherein said processor comprises a previous training module for unique multivariate regression, said training being performed by using a dataset of training images.

20. The device for processing an image according to claim 19, wherein said dataset of training images comprises:

a first set of training images, which provides a positive classifying result when being processed by said multivariate regression, and

a second set of training images, which provides a negative classifying result when being processed by said multivariate regression.

21. The device for processing an image according to claim 20, wherein said second set of training images comprises images, which comprise no object, and/or images where object landmark annotation is erroneous.