IMAGE OBJECT RETRIEVAL

• US 20130148903A1
• Filed: 12/08/2011
• Published: 06/13/2013
• Est. Priority Date: 12/08/2011
• Status: Active Grant

• Current Assignee: R2 Solutions LLC (Acacia Research Corporation)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Pueyo, Lluis Garcia, van Zwol, Roelof
• Granted Patent: US 9,870,517 B2
• Time in Patent Office: Days
• US Class Current: 382/224
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/757 Matching configurations of points or features

Abstract

Subject matter disclosed herein relates to image object recognition or retrieval.

Claims

1. A method comprising:

quantizing image descriptors for an image, the quantized image descriptors including interest point neighborhood state information; and

searching for matches between said quantized image descriptors and quantized image descriptors of reference images.

2. The method of claim 1, wherein said interest point neighborhood state information comprises state information regarding an interest point and one or more neighboring points of said interest point.

3. The method of claim 2, wherein said one or more neighboring points are determined among a plurality of points based, at least in part, on a geometrical distance between said one or more neighboring points and said interest point.

4. The method of claim 3, wherein said interest point and neighboring points include a scale;

and further comprising;

forming a signature between said interest points and neighboring points based at least in part on scale.

5. The method of claim 1, wherein said quantized image descriptors of reference images comprise components stored in an edge index.

6. The method of claim 5, wherein said components stored in said edge index comprise quantization identifications (IDs) of neighboring points of interest points in said reference images.

7. The method of claim 5, further comprising:

in response to a search query to search for said object of said image, forming N signatures for an interest point and N neighboring points of said interest point of said image; and

comparing said N signatures to said components stored in said edge index to determine a match between said object of said image and an object in said reference images.

8. The method of claim 7, further comprising:

forming a geometrical shape to enclose said object of said image.

9. The method of claim 1, wherein said quantized image descriptors are at least approximately scale invariant.

10. The method of claim 1, wherein the quantized image descriptors comprise interest points of said object, wherein relative positions of said interest points are substantially constant with respect to varying perspectives of said object.

11. An apparatus comprising:

a computing platform to discriminate two or more candidate interest points of an image based, at least in part, on quantization identifications (IDs) of respective neighboring points of said two or more candidate interest points.

12. The apparatus of claim 11, wherein said computing platform to rank said two more candidate interest points based at least in part on said quantization IDs.

13. The apparatus of claim 11, wherein said candidate interest points comprise quantized image descriptors.

14. The apparatus of claim 13, wherein said quantized image descriptors comprise SURF descriptors.

15. The apparatus of claim 11, wherein said computing platform to identify neighboring points based, at least in part, on geometric distance from said particular candidate interest point in said image.

16. An apparatus comprising:

a special purpose computing system, said special purpose computing system to;

transform a geometric arrangement of points of an object in an image;

quantize groups of said points; and

store the quantized groups of said points in an index.

17. The apparatus of claim 16, wherein said special purpose computing system to transform a geometric arrangement of points of an object in an image comprises said special purpose computing system to transform a geometric arrangement of points of an object in a reference image.

18. The apparatus of claim 16, wherein said index comprises an edge index.

19. The apparatus of claim 16, wherein said geometric arrangement of points comprises a polygon.