Enabling content-based search of objects in an image database with reduced matching

• US 7,317,816 B2
• Filed: 08/19/2003
• Issued: 01/08/2008
• Est. Priority Date: 08/19/2003
• Status: Active Grant

• Current Assignee: Indian Institute of Technology Delhi, Intel Corporation
• Original Assignee: Indian Institute of Technology Delhi, Intel Corporation
• Inventors: Mishra, Ranjit K., Ray, Ajoy K., Acharya, Tinku
• Primary Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US10/643,672
• Publication Number:
  US 20050041863A1
• Time in Patent Office: 1,603 Days
• Field of Search: 382/103, 382/118, 382/199, 382/190, 382/305, 709/204, 709/205, 355/75, 715/759, 715/753, 370/260, 379/202.01, 341/17
• US Class Current: 382/118
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

In one embodiment, the present invention includes a method to use active contours to detect image boundaries of a first view and a second view of a human face, and mark a first set of fiducial points on the first view and a second set of fiducial points on the second view. Using the fiducial points, feature vectors may be calculated and stored in a database for later comparison with a query image.

Claims

1. A method comprising:

using active contours to detect image boundaries of a first view and a second view of a human face; and

marking a first set of fiducial points on the first view and a second set of fiducial points on the second view.

2. The method of claim 1, including:

determining a first feature set using the first set of fiducial points, and determining a second feature set using the second set of fiducial points.

3. The method of claim 2, further comprising:

normalizing distances in the first feature set in terms of a distance between two preselected fiducial points of the first set of fiducial points.

4. The method of claim 2, wherein the first feature set and the second feature set each comprise less than ten distances.

5. The method of claim 2, further comprising storing the first feature set and the second feature set in a database.

6. The method of claim 5, further comprising:

partitioning said database based on a feature vector of one of the first feature set and the second feature set.

7. The method of claim 6, further comprising:

in response to a query image, querying the database using reference image content corresponding to the feature vector.

8. The method of claim 1, wherein the active contours comprise snake contours and the first and second views comprise a front view and a side view.

9. The method of claim 8, wherein the snake contours for the front view comprise at least two of a face boundary, an eye boundary, a brow boundary, a nose boundary, and a lip boundary.

10. A method comprising:

obtaining feature set information including fiducial points from a first view of a human face and a second view of the human face; and

storing the feature set information in a database having a hierarchical tree structure.

11. The method of claim 10, wherein the database includes feature set information for a plurality of individuals.

12. The method of claim 11, further comprising searching the database for at least one search result corresponding to a query image.

13. The method of claim 10, further comprising determining whether to store the feature set information in a first branch or a second branch of the hierarchical tree structure based upon a value in the feature set information.

14. The method of claim 13, wherein the value corresponds to a metric distance function.

15. The method of claim 13, further comprising searching the first branch or the second branch for a search result corresponding to a query image based on a metric distance function of the query image.

16. An article comprising a machine-readable storage medium containing instructions that if executed enable a system to:

obtain feature set information including fiducial points from a first view of a human face and a second view of the human face; and

store the feature set information in a database having a hierarchical tree structure.

17. The article of claim 16, further comprising instructions that if executed enable the system to determine whether to store the feature set information in a first branch or a second branch of the hierarchical tree structure based upon a value in the feature set information.

18. The article of claim 17, further comprising instructions that if executed enable the system to search the first branch or the second branch for a search result corresponding to a query image based on a metric distance function of the query image.

19. The article of claim 16, further comprising instructions that if executed enable the system to search the database for at least one search result corresponding to a query image.

20. A system comprising:

a dynamic random access memory containing instructions that if executed enable the system to use active contours to detect image boundaries of a first view and a second view of a human face, and to mark a first set of fiducial points on the first view and a second set of fiducial points on the second view; and

a processor coupled to the dynamic random access memory to execute the instructions.

21. The system of claim 20, further comprising instructions that if executed enable the system to determine a first feature set using the first set of fiducial points, and determine a second feature set using the second set of fiducial points.

22. The system of claim 21, further comprising instructions that if executed enable the system to store the first feature set and the second feature set in a database.

23. The system of claim 22, further comprising instructions that if executed enable the system to partition the database based on a feature vector of one of the first feature set and the second feature set.

24. The system of claim 22, further comprising a display coupled to the processor to display a query image and at least one search result image obtained from the database in response to a similarity query based on at least one feature vector.