Component fusion for face detection

• US 20050001013A1
• Filed: 05/11/2004
• Published: 01/06/2005
• Est. Priority Date: 05/14/2003
• Status: Active Grant

• Current Assignee: Siemens Corp. (Siemens AG)
• Original Assignee: Siemens Corp. (Siemens AG)
• Inventors: Ramesh, Visvanathan, Comaniciu, Dorin, Simon, Markus, Xie, Binglong
• Granted Patent: US 7,602,941 B2
• Time in Patent Office: Days
• US Class Current: 228/103
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

A system and method for object detection are provided where the system includes a component detection unit for detecting components in an image, a component fusion unit in signal communication with the component detection unit for fusing the components into an object, and a CPU in signal communication with the detection and fusion units for comparing the fused components with a statistical model; and the method includes receiving observation data for a plurality of training images, forming at least one statistical model from the plurality of training images, receiving an input image having a plurality of pixels, detecting a plurality of components in the input image, determining a fusion of the detected components, comparing the fusion with the statistical model, and detecting an object in accordance with the comparison.

Claims

1. A method for object detection, the method comprising:

receiving observation data for a plurality of training images;

forming at least one statistical model from the plurality of training images;

receiving an input image having a plurality of pixels;

detecting a plurality of components in the input image;

determining a fusion of the detected components;

comparing the fusion with the statistical model; and

detecting an object in accordance with said comparison.

2. A method as defined in claim 1 wherein said images are indicative of at least one face.

3. A method as defined in claim 2 wherein said components are indicative of a left eye, a right eye, and a lower face, respectively.

4. A method as defined in claim 1 wherein said input image is a real-time image.

5. A method as defined in claim 1, further comprising calibrating said statistical model for heteroscedastic noise.

6. A method as defined in claim 1 wherein the the left and right eye components are each rectangular areas of 36 by 28 pixels, and the lower face component is a rectangular area of 52 by 40 pixels.

7. A method as defined in claim 1 wherein the faces are aligned to a 64 by 64 pixel image.

8. A method as defined in claim 1 wherein the detectors are responsive to at least the poses of frontal, turning left, turning right, tilting up, and tilting down.

9. A method as defined in claim 1, further comprising AdaBoosting to obtain better strong classifiers.

10. A system (100) for object detection, the system comprising:

a component detection unit (170) for detecting components in an image;

a component fusion unit (180) in signal communication with the component detection unit for fusing the components into an object; and

a CPU (102) in signal communication with said detection and fusion units for comparing the fused object with a statistical model.

11. A system as defined in claim 10, further comprising:

observation means for receiving observation data for a plurality of training images;

modeling means for forming at least one statistical model from the plurality of training images;

receiving means for receiving an input image having a plurality of pixels;

component detection means for detecting a plurality of components in the input image;

fusion means for determining a fusion of the detected components;

comparison means for comparing the fusion with the statistical model; and

object detection means for detecting an object in accordance with said comparison.

12. A system as defined in claim 10 wherein said images are indicative of at least one face.

13. A system as defined in claim 10 wherein said components are indicative of a left eye, a right eye, and a lower face, respectively.

14. A system as defined in claim 10 wherein said input image is a real-time image.

15. A system as defined in claim 10, further comprising calibration means for calibrating said statistical model for heteroscedastic noise.

16. A system as defined in claim 10 wherein the the left and right eye components are each rectangular areas of 36 by 28 pixels, and the lower face component is a rectangular area of 52 by 40 pixels.

17. A system as defined in claim 10 wherein the faces are aligned to a 64 by 64 pixel image.

18. A system as defined in claim 10 wherein the detectors are responsive to at least the poses of frontal, turning left, turning right, tilting up, and tilting down.

19. A system as defined in claim 10, further comprising AdaBoosting means for obtaining better strong classifiers.

20. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for object detection, the method steps comprising:

receiving observation data for a plurality of training images;

forming at least one statistical model from the plurality of training images;

receiving an input image having a plurality of pixels;

detecting a plurality of components in the input image;

determining a fusion of the detected components;

comparing the fusion with the statistical model; and

detecting an object in accordance with said comparison.

21. A program storage device as defined in claim 20 wherein said images are indicative of at least one face.

22. A program storage device as defined in claim 20 wherein said components are indicative of a left eye, a right eye, and a lower face, respectively.

23. A program storage device as defined in claim 20 wherein said input image is a real-time image.

24. A program storage device as defined in claim 20, further comprising a program step for calibrating said statistical model for heteroscedastic noise.

25. A program storage device as defined in claim 20 wherein the the left and right eye components are each rectangular areas of 36 by 28 pixels, and the lower face component is a rectangular area of 52 by 40 pixels.

26. A program storage device as defined in claim 20 wherein the faces are aligned to a 64 by 64 pixel image.

27. A program storage device as defined in claim 20 wherein the detectors are responsive to at least the poses of frontal, turning left, turning right, tilting up, and tilting down.

28. A program storage device as defined in claim 20, further comprising a program step for AdaBoosting to obtain better strong classifiers.