Method and component for image recognition

• US 8,335,355 B2
• Filed: 04/21/2010
• Issued: 12/18/2012
• Est. Priority Date: 12/29/2004
• Status: Expired due to Fees

• Current Assignee: FotoNation Limited (Adeia Inc.), National University of Ireland Galway
• Original Assignee: DigitalOptics Corporation Europe Limited (Adeia Inc.), National University of Ireland Galway
• Inventors: Costache, Gabriel, Corcoran, Peter, Mulryan, Rhys, Steinberg, Eran
• Primary Examiner(s): Le, Brian
• Application Number:
  US12/764,650
• Publication Number:
  US 20100202707A1
• Time in Patent Office: 972 Days
• Field of Search: 382/118, 382/159, 382/224, 382/305
• US Class Current: 382/118
• CPC Class Codes: G06F 16/5854 using shape and object relationship
• CPC Class Codes: G06V 20/30 in albums, collections or shared content, e.g. social network photos or video
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions

Abstract

A method and system for image recognition in a collection of digital images includes training image classifiers and retrieving a sub-set of images from the collection. For each image in the collection, any regions within the image that correspond to a face are identified. For each face region and any associated peripheral region, feature vectors are determined for each of the image classifiers. The feature vectors are stored in association with data relating to the associated face region. At least one reference region including a face to be recognized is/are selected from an image. At least one classifier on which said retrieval is to be based is/are selected from the image classifiers. A respective feature vector for each selected classifier is determined for the reference region. The sub-set of images is retrieved from within the image collection in accordance with the distance between the feature vectors determined for the reference region and the feature vectors for face regions of the image collection.

Claims

1. A digital image acquisition device, including a lens, an image sensor and a processor, and having an operating system including a component embodied within a processor-readable medium for programming the processor to perform an image recognition method a) training a plurality of image classifiers, including:

for a plurality of images in the collection, identifying one or more regions corresponding to a face region;

for each image identified as having multiple face regions, for each of a plurality of image classifiers, determining combination feature vectors corresponding to the multiple face regions; and

storing said combination feature vectors in association with certain recognizable data relating to at least one of the multiple face regions, and

b) retrieving a sub-set of images from said collection or a different collection that includes one or more images including both a face associated with certain recognizable data and a second face, or a subset of said collection, or a combination thereof, including;

selecting from said plurality of image classifiers at least one classifier on which said retrieving is to be based, said at least one classifier being configured for programming the processor to select images containing at least two reference face regions including a first face to be recognized and a second face;

determining, for said at least two reference face regions, a respective feature vector for one or more selected classifiers; and

retrieving said sub-set of images from within said collection or said different collection that includes one or more images including both said face associated with certain recognizable data and said second face, or said subset of said collection, or said combination thereof, in accordance with the distance between the feature vectors determined for said reference region and the feature vectors for face regions of said image collection; and

wherein said determining comprises;

a) for each face region, extracting respective features representative of the region;

b) for each of said plurality of image classifiers, determining respective basis vectors according to said extracted features; and

c) for the extracted features for each region, for each classifier, determining said feature vectors, based on each determined basis vector.

14. A device as claimed in claim 1, wherein the method further comprises calculating for each set of distance measures, mean and variance values.

16. A device as claimed in claim 1, wherein each basis vector for a classifier comprises a selected set of attributes and respective reference values for these attributes.

17. A device as claimed in claim 1, wherein said determining feature vectors for said reference region, is responsive to determining that feature vectors have previously been determined for said reference region for said classifier, for retrieving said feature vectors from storage.

18. A device as claimed in claim 1, wherein retrieving said sub-set of images comprises, for each classifier, comparing feature vectors for the selected face region with feature vectors for face regions in the image collection to provide a set of distance measures.

2. A method for image recognition in a collection of digital images comprising:

a) training a plurality of image classifiers, including;

for a plurality of images in the collection, identifying one or more regions corresponding to a face region;

for each image identified as having multiple face regions, for each of a plurality of image classifiers, determining combination feature vectors corresponding to the multiple face regions; and

storing said combination feature vectors in association with certain recognizable data relating to at least one of the multiple face regions, and

b) retrieving a sub-set of images from said collection or a different collection that includes one or more images including both a face associated with certain recognizable data and a second face, or a subset of said collection, or a combination thereof, including;

selecting from said plurality of image classifiers at least one classifier on which said retrieving is to be based, said at least one classifier being configured for programming the processor to select images containing at least two reference face regions including a first face to be recognized and a second face;

determining, for said at least two reference face regions, a respective feature vector for one or more selected classifiers; and

retrieving said sub-set of images from within said collection or said different collection that includes one or more images including both said face associated with certain recognizable data and said second face, or said subset of said collection, or said combination thereof, in accordance with the distance between the feature vectors determined for said reference region and the feature vectors for face regions of said image collection; and

wherein said determining comprises;

a) for each face region, extracting respective features representative of the region;

b) for each of said plurality of image classifiers determining respective basis vectors according to said extracted features; and

c) for the extracted features for each region, for each classifier, determining said feature vectors, based on each determined basis vector.

3. A method as claimed in claim 2, wherein said determining further comprises:

a) for each associated peripheral region for said each face region, extracting respective features representative of the peripheral region;

b) for each of said plurality of image classifiers, determining respective basis vectors according to said extracted features; and

c) for the extracted features for each peripheral region, for each classifier, determining said feature vectors, based on each determined basis vector.

4. A method as claimed in claim 2 wherein each basis vector for a classifier comprises a selected set of attributes and respective reference values for these attributes.

5. A method as claimed in claim 2, wherein said determining feature vectors for said reference region, is responsive to determining that feature vectors have previously been determined for said reference region for said classifier, for retrieving said feature vectors from storage.

6. A method as claimed in claim 2, wherein retrieving said sub-set of images comprises, for each classifier, comparing feature vectors for the selected face region with feature vectors for face regions in the image collection to provide a set of distance measures.

7. A method as in claim 2, further comprising calculating for each set of distance measures, mean and variance values.

8. A component embodied within a non-transitory processor-readable medium for programming a processor to perform an image recognition method including image recognition in a collection of digital images, wherein the method comprises:

a) training a plurality of image classifiers, including;

for a plurality of images in the collection, identifying one or more regions corresponding to a face region;

for each image identified as having multiple face regions, for each of a plurality of image classifiers, determining combination feature vectors corresponding to the multiple face regions; and

storing said combination feature vectors in association with certain recognizable data relating to at least one of the multiple face regions, and

b) retrieving a sub-set of images from said collection or a different collection that includes one or more images including both a face associated with certain recognizable data and a second face, or a subset of said collection, or a combination thereof, including;

selecting from said plurality of image classifiers at least one classifier on which said retrieving is to be based, said at least one classifier being configured for programming the processor to select images containing at least two reference face regions including a first face to be recognized and a second face;

determining, for said at least two reference face regions, a respective feature vector for one or more selected classifiers; and

retrieving said sub-set of images from within said collection or said different collection that includes one or more images including both said face associated with certain recognizable data and said second face, or said subset of said collection, or said combination thereof, in accordance with the distance between the feature vectors determined for said reference region and the feature vectors for face regions of said image collection; and

wherein said determining comprises;

a) for each face region, extracting respective features representative of the region;

b) for each of said plurality of image classifiers, determining respective basis vectors according to said extracted features; and

c) for the extracted features for each region, for each classifier, determining said feature vectors, based on each determined basis vector.

9. A component as claimed in claim 8, wherein said determining further comprises:

d) for each associated peripheral region for said each face region, extracting respective features representative of the peripheral region;

e) for each of said plurality of image classifiers, determining respective basis vectors according to said extracted features; and

f) for the extracted features for each peripheral region, for each classifier, determining said feature vectors, based on each determined basis vector.

10. A component as claimed in claim 8, wherein each basis vector for a classifier comprises a selected set of attributes and respective reference values for these attributes.

11. A component as claimed in claim 8, wherein said determining feature vectors for said reference region, is responsive to determining that feature vectors have previously been determined for said reference region for said classifier, for retrieving said feature vectors from storage.

12. A component as claimed in claim 8, wherein retrieving said sub-set of images comprises, for each classifier, comparing feature vectors for the selected face region with feature vectors for face regions in the image collection to provide a set of distance measures.

13. A component as claimed in claim 8, wherein the method further comprises calculating for each set of distance measures, mean and variance values.

15. A component as claimed in claim 8, wherein said determining further comprises:

g) for each face region and any associated peripheral region, extracting respective features representative of the region;

h) for each of said plurality of image classifiers, determining respective basis vectors according to said extracted features; and

i) for the extracted features for each region, for each classifier, determining said feature vectors, based on each determined basis vector.