Method and apparatus for providing a robust object finder

• US 20040197011A1
• Filed: 04/05/2004
• Published: 10/07/2004
• Est. Priority Date: 04/04/2003
• Status: Active Grant

• Current Assignee: Princeton Identity Inc.
• Original Assignee: Sarnoff Corporation (SRI International, Inc.)
• Inventors: Wildes, Richard, Camus, Theodore A.
• Granted Patent: US 7,599,524 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 40/18 Eye characteristics, e.g. of the iris

Abstract

A method of quickly and robustly localizing a pair of concentric circular region boundaries in a digital image, such as the pupil and iris boundaries in an image of a human eye is disclosed. Specularities in the image are optionally filled. An object containing concentric regions, e.g., a pupil and iris, is found within the image.

Claims

1. A method of detecting an object in an image, comprising:

filling in at least one specularity in the image; and

finding the object within the image, wherein the object comprises concentric regions.

2. The method of claim 1, wherein finding the object within the image further comprises:

locating a plurality of candidate objects in the image;

modeling each candidate object as a first concentric region and a second concentric region; and

selecting the object from the plurality of candidate objects in accordance with a quality measure.

3. The method of claim 2, further comprising:

performing boundary fitting on the plurality of candidate objects.

4. The method of claim 3 wherein performing boundary fitting comprises performing a gradient search on the plurality of candidate objects, using the quality measure, until a local maximum is reached.

5. The method of claim 3, further comprising:

performing fit optimization on the selected object.

6. The method of claim 5 wherein performing fit optimization comprises performing a gradient search on the selected object.

7. The method of claim 2, wherein each candidate is modeled as a first concentric circle and a second concentric circle.

8. The method of claim 7, wherein at least one of the first concentric circle and second concentric is modeled as an ellipse.

9. The method of claim 5, further comprising performing pyramid processing on the image in order to reproduce the image in a plurality of resolutions.

10. The method of claim of claim 9, further comprising performing fit optimization on the selected object in a higher resolution image.

11. A method of finding an object within an image, comprising:

locating a plurality of candidate objects in the image;

modeling each candidate object as a first concentric region and a second concentric region;

performing boundary fitting on the plurality of candidate objects; and

selecting the object from the plurality of candidate objects in accordance with a quality measure.

12. The method of claim 11, wherein performing boundary fitting on the plurality of candidate objects further comprises:

configuring a plurality of rays into a rectangular representation for each candidate object; and

estimating, for each candidate object, a first radius corresponding to the first concentric region and a second radius corresponding to the second concentric region in order to determine the quality measure of each candidate object.

13. The method of claim 12, wherein the plurality of rays radiate from a proposed center of each candidate object.

14. The method of claim 13, wherein the plurality of rays represent pixel lines.

15. The method of claim 11, wherein selecting the object from the plurality of candidate objects further comprises selecting the candidate object having a maximum quality measure.

16. A method of finding an object within an image, comprising:

filling in at least one specularity in the image;

locating a plurality of candidate objects in the image;

modeling each candidate object as a first concentric region and a second concentric region;

performing boundary fitting on the plurality of candidate objects, where boundary fitting further comprises;

configuring a plurality of rays, representing pixel lines and radiating from a proposed center of each candidate object, into a rectangular representation for each candidate object;

estimating, for each candidate object, a first radius corresponding to the first concentric region and a second radius corresponding to the second concentric region in order to determine the quality measure of each candidate object; and

selecting the object from the plurality of candidate objects in accordance with a quality measure.

17. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to perform the steps of a method of detecting an object in an image, comprising of:

filling in at least one specularity in the image; and

finding the object within the image, wherein the object comprises concentric regions.

18. A device for detecting an object in an image, comprising:

means for filling in at least one specularity in the image; and

means for finding the object within the image, wherein the object comprises concentric regions.

19. A device for finding an object within an image, comprising:

means for locating a plurality of candidate objects in the image;

means for modeling each candidate object as a first concentric region and a second concentric region;

means for performing boundary fitting on the plurality of candidate objects; and

means for selecting the object from the plurality of candidate objects in accordance with a quality measure.

20. A device for finding an object within an image, comprising:

means for filling in at least one specularity in the image;

means for locating a plurality of candidate objects in the image;

means for modeling each candidate object as a first concentric region and a second concentric region;

means for performing boundary fitting on the plurality of candidate objects, where means for performing boundary fitting further comprises;

means for configuring a plurality of rays, representing pixel lines and radiating from a proposed center of each candidate object, into a rectangular representation for each candidate object;

means for estimating, for each candidate object, a first radius corresponding to the first concentric region and a second radius corresponding

to the second concentric region in order to determine the quality measure of each candidate object; and

means for selecting the object from the plurality of candidate objects in accordance with a quality measure.