Image processing

• US 7,630,561 B2
• Filed: 05/25/2005
• Issued: 12/08/2009
• Est. Priority Date: 05/28/2004
• Status: Expired due to Fees

• Current Assignee: Sony United Kingdom Limited (Sony Group Corp.)
• Original Assignee: Sony United Kingdom Limited (Sony Group Corp.)
• Inventors: Haynes, Simon Dominic, Living, Jonathan, Gillard, Clive Henry, Porter, Robert Mark Stefan, Rambaruth, Ratna
• Primary Examiner(s): Carter; Aaron W
• Application Number:
  US11/139,404
• Publication Number:
  US 20050265603A1
• Time in Patent Office: 1,658 Days
• Field of Search: 382/218, 382/103, 382/115, 382/118, 382/219, 382/220, 382/209, 348/155, 348/169
• US Class Current: 382/218
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries
• CPC Class Codes: G06V 40/165 using facial parts and geometric relationships
• CPC Class Codes: G06V 40/167 using comparisons between temporally consecutive images
• CPC Class Codes: G06V 40/173 face re-identification, e.g. recognising unknown faces across different face tracks

Abstract

An image processing method for selecting a subset of n (where n is two or more) images representative of a test set of images. The method includes detecting a degree of similarity between images in the test set of images so that the subset of images comprises those n images having substantially the lowest similarity with other images in the set.

Claims

1. An image processing method performed by an image processing apparatus for selecting a subset of n (where n is two or more) images representative of a test set of images, said method comprising:

detecting, by a processor of said image processing apparatus, a degree of similarity between images of said test set of images, respective ones of said test set of images being divided into one or more test regions, each of which being compared with one or more reference regions in one or more other images of said test set, identifying for each respective test region a reference region that most closely corresponds thereto, measuring the similarity between an image and a reconstruction of the image from the identified reference regions;

selecting, by the processor of said image processing apparatus, a subset of images comprising those n images having substantially the lowest similarity with respect to other images of said set;

populating a test subset with n images from said set;

detecting, by the processor of said image processing apparatus, mutual similarity between said n images in said test subset; and

for each other image in said set in turn;

detecting, by the processor of said image processing apparatus, a degree of similarity between each respective image and said n images in said subset; and

if an image is less similar to said n images in said test subset than one of said images in said subset is to the remainder of said test subset, replacing said least dissimilar image in said subset with the one of said images.

2. A method according to claim 1, in which, if said test set of images comprises fewer than n images, a subset of representative images is not selected.

3. A method according to claim 1, further comprising:

detecting the presence of an object in successive pictures of a video sequence;

deriving a sequence of picture portions containing said detected object, that sequence of picture portions forming said test set of images.

4. A method according to claim 3, in which said sequence of picture portions forming said test set of images is derived from a temporally contiguous group of pictures in said video sequence.

5. A method according to claim 3, further comprising:

processing said pictures to detect picture features indicative of the presence of said object; and

subsequently detecting said object with a greater likelihood than when detected by said processing;

in which said sequence of picture portions containing said detected object is derived from said processing, with zero or more picture portions being derived from said subsequent detection subject to constraints on distribution of picture portions within said sequence which can be derived using said or each subsequent detection.

6. A method according to claim 5, further comprising:

detecting colour properties of a picture indicative of the presence of said object.

7. A method according to claim 5, further comprising:

predicting the position of said object within a picture on the basis of the detected position of said object in one or more other pictures.

8. A method according to claim 5, further comprising:

excluding any picture portions derived using a subsequent detection from said test set of images.

9. A method according to claim 1, in which said objects are faces.

10. A computer readable storage medium including computer executable code which causes the image processing apparatus to implement a method according to claim 1.

11. An image processing apparatus operable to select a subset of n (where n is two or more) images representative of a test set of images, said apparatus comprising:

a processor configured to detect a degree of similarity between images in said test set of images, wherein respective ones of said test set of images are divided into one or more test regions, each of which being compared with one or more reference regions in one pg 29. or more other images of said test set, for each respective test region, a reference region being identified that most closely corresponds thereto, the similarity between an image and a reconstruction of the image from the identified reference regions being measured; and

wherein a subset of images is selected which comprises those n images having substantially the lowest similarity with respect to other images of said set, a test subset with n images from said set are populated to detect mutual similarities between said n images and said test subset, for each other image in said set intern to detect a degree of similarities between each respective image and said n images in said subset so that if an image is less similar to said n images and said test subset then one of set images in said subset is to the remainder of said test subset, replacing said at least this similar image in said subset with the one of said images.

12. Video conferencing apparatus comprising image processing apparatus according to claim 11.

13. A Surveillance apparatus comprising image processing apparatus according to claim 11.