Face Recognition with Combined PCA-Based Datasets

• US 20080031498A1
• Filed: 08/02/2007
• Published: 02/07/2008
• Est. Priority Date: 08/02/2006
• Status: Active Grant

• Current Assignee: FotoNation Limited (Adeia, Inc.)
• Original Assignee: FotoNation Vision Limited (Adeia, Inc.)
• Inventors: Corcoran, Peter, Costache, Gabriel N.
• Granted Patent: US 7,515,740 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06F 18/2135 based on approximation criteria, e.g. principal component analysis
• CPC Class Codes: G06F 18/253 of extracted features
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 10/806 of extracted features
• CPC Class Codes: G06V 40/169 Holistic features and representations, i.e. based on the facial image taken as a whole

Abstract

A face recognition method for working with two or more collections of facial images is provided. A representation framework is determined for a first collection of facial images including at least principle component analysis (PCA) features. A representation of said first collection is stored using the representation framework. A modified representation framework is determined based on statistical properties of original facial image samples of a second collection of facial images and the stored representation of the first collection. The first and second collections are combined without using original facial image samples. A representation of the combined image collection (super-collection) is stored using the modified representation framework. A representation of a current facial image, determined in terms of the modified representation framework, is compared with one or more representations of facial images of the combined collection. Based on the comparing, it is determined which, if any, of the facial images within the combined collection matches the current facial image.

Claims

1. A face recognition method for working with two or more collections of facial images, comprising:

(a) determining a representation framework for a first collection of facial images including at least principle component analysis (PCA) features;

(b) storing a representation of said first collection using said representation framework

(c) determining a modified representation framework based on statistical properties of original facial image samples of a second collection of facial images and the stored representation of the first collection;

(d) combining the first and second collections without using original facial image samples;

(e) storing a representation of the combined image collection (super-collection) using said modified representation framework

(f) comparing a representation of a current facial image, determined in terms of said modified representation framework, with one or more representations of facial images of the combined collection; and

(g) based on the comparing, determining whether one or more of the facial images within the combined collection matches the current facial image.

2. The method of claim 1, further comprising, prior to (b), updating said first collection by combining the first collection with a third collection.

3. The method of claim 2, wherein the updating comprises adding one or more new data samples to the first collection.

4. The method of claim 1, further comprising re-projecting data samples of the first collection into a new eigenspace without using original facial images of the first collection.

5. The method of claim 4, wherein the re-projecting uses existing PCA features.

6. The method of claim 1, further comprising combining training data from said first and second collections.

7. A method according to claim 1 which does not use original data samples of the first collection.

8. The method of claim 1, wherein the first and second collections contain samples of different dimension, and the method further comprises:

(i) choosing a standard size of face region for the new collection, and

(ii) re-sizing eigenvectors using interpolation to the standard size.

9. The method of claim 8, wherein said samples of different dimension were analyzed using different standard sizes of face region.

10. The method of claim 1, further comprising generating the modified representation framework according to the following:

11. A face recognition method for working with two or more collections of facial images, comprising:

(a) determining different representation frameworks for first and second collections of facial images each including at least principle component analysis (PCA) features;

(b) storing different representations of said first and second collections using said different representation frameworks;

(c) determining a modified representation framework based on the different representations of the first and second collections, respectively;

(d) combining the first and second collections without using original facial image samples;

(e) storing a representation of the combined image collection (super-collection) using said modified representation framework

(f) comparing a representation of a current facial image, determined in terms of said modified representation framework, with one or more representations of facial images of the combined collection; and

(g) based on the comparing, determining whether one or more of the facial images within the combined collection matches the current facial image.

12. The method of claim 11, further comprising updating the PCA features based on first eigenvectors and first sets of eigenvalues for each of the first and second collections.

13. The method of claim 12, wherein the updating comprises calculating a new set of eigenvectors from the previously calculated first eigenvectors of the first and second collections.

14. A method according to claim 11 which does not use original data samples of the first and second collections.

15. The method of claim 11, wherein the first and second collections contain samples of different dimension, and the method further comprises:

(i) choosing a standard size of face region for the new collection, and

(ii) re-sizing eigenvectors using interpolation to the standard size.

16. The method of claim 15, wherein said samples of different dimension were analyzed using different standard sizes of face region.

17. The method of claim 11, further comprising generating the modified representation framework according to the following:

18. One or more processor readable media having program code embodied therein for programming one or more processors to perform a face recognition method for working with two or more collections of facial images, wherein the method comprises:

(a) determining a representation framework for a first collection of facial images including at least principle component analysis (PCA) features;

(b) storing a representation of said first collection using said representation framework

(c) determining a modified representation framework based on statistical properties of original facial image samples of a second collection of facial images and the stored representation of the first collection;

(d) combining the first and second collections without using original facial image samples;

(e) storing a representation of the combined image collection (super-collection) using said modified representation framework

(f) comparing a representation of a current facial image, determined in terms of said modified representation framework, with one or more representations of facial images of the combined collection; and

(g) based on the comparing, determining whether one or more of the facial images within the combined collection matches the current facial image.

19. The one or more media of claim 18, wherein the method further comprises, prior to (b), updating said first collection by combining the first collection with a third collection.

20. The one or more media of claim 19, wherein the updating comprises adding one or more new data samples to the first collection.

21. The one or more media of claim 18, wherein the method further comprises re-projecting data samples of the first collection into a new eigenspace without using original facial images of the first collection.

22. The one or more media of claim 21, wherein the re-projecting uses existing PCA features.

23. The one or more media of claim 18, wherein the method further comprises combining training data from said first and second collections.

24. A one or more media according to claim 18, wherein the method does not use original data samples of the first collection.

25. The one or more media of claim 18, wherein the first and second collections contain samples of different dimension, and the method further comprises:

(i) choosing a standard size of face region for the new collection, and

(ii) re-sizing eigenvectors using interpolation to the standard size.

26. The one or more media of claim 25, wherein said samples of different dimension were analyzed using different standard sizes of face region.

27. The one or more media of claim 18, wherein the method further comprises generating the modified representation framework according to the following:

28. One or more processor readable media having program code embodied therein for programming one or more processors to perform a face recognition method for working with two or more collections of facial images, wherein the method comprises:

(a) determining different representation frameworks for first and second collections of facial images each including at least principle component analysis (PCA) features;

(b) storing different representations of said first and second collections using said different representation frameworks;

(c) determining a modified representation framework based on the different representations of the first and second collections, respectively;

(d) combining the first and second collections without using original facial image samples;

(e) storing a representation of the combined image collection (super-collection) using said modified representation framework

(f) comparing a representation of a current facial image, determined in terms of said modified representation framework, with one or more representations of facial images of the combined collection; and

(g) based on the comparing, determining whether one or more of the facial images within the combined collection matches the current facial image.

29. The one or more media of claim 28, wherein the method further comprises updating the PCA features based on first eigenvectors and first sets of eigenvalues for each of the first and second collections.

30. The one or more media of claim 29, wherein the updating comprises calculating a new set of eigenvectors from the previously calculated first eigenvectors of the first and second collections.

31. A one or more media according to claim 28, wherein the method does not use original data samples of the first and second collections.

32. The one or more media of claim 28, wherein the first and second collections contain samples of different dimension, and the method further comprises:

(i) choosing a standard size of face region for the new collection, and

(ii) re-sizing eigenvectors using interpolation to the standard size.

33. The one or more media of claim 32, wherein said samples of different dimension were analyzed using different standard sizes of face region.

34. The one or more media of claim 28, wherein the method further comprises generating the modified representation framework according to the following: