Model based faced coding and decoding using feature detection and eigenface coding

• US 6,044,168 A
• Filed: 11/14/1997
• Issued: 03/28/2000
• Est. Priority Date: 11/25/1996
• Status: Expired due to Term

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Tuceryan, Mihran, Flinchbaugh, Bruce E.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Werner, Brian P.
• Application Number:
  US08/970,301
• Time in Patent Office: 865 Days
• Field of Search: 382/103, 382/115, 382/118, 382/173, 382/201, 382/217, 382/218, 382/232, 382/282, 382/285, 382/286, 382/215, 382/283, 345/473, 345/425, 345/430, 345/433-439, 340/825.34, 348/15, 348/25, 348/169, 348/384, 348/390
• US Class Current: 382/118
• CPC Class Codes: G06T 9/001 Model-based coding, e.g. wire frame
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions

Abstract

The method uses a three-dimensional face model and a technique called eigenface decomposition to analyze the video at one end. The facial feature locations and eigenface coding of the face image are sent to a decoder. The decoder synthesizes the face image at the receiving end. Eigenface decoding is used to texture map a three-dimensional model warped by detected feature locations.

Claims

1. A method of sending a facial image comprising the steps of:

using a set of multiscale templates locating a point on said facial image with the highest correlation coefficient at the best scale as the location of the face;

detecting facial features by template matching facial features to provide a set of points defining said facial feature locations;

encoding said facial image into eigenface parameters including the steps of providing a training database of eigenfaces and performing eigenface decomposition of an input image by finding eigenvectors and eigenvalues of points in free space represented by a training set of images;

transmitting and receiving said eigenvalues and said set of points defining said facial feature locations;

providing at a receiver a three-dimensional generic model of a face;

warping said generic face model using said set of points defining said facial feature locations;

said warping step includes using an affine transform for a portion of said face and bilinear transform to warp a different portion of said face;

decoding said facial image from eigenvalues to reconstruct the face image to define a texture image;

said decoding including the step of providing said training database of eigenfaces; and

mapping said texture image onto said warped three-dimensional face model to provide a synthesized facial image.

2. The method of claim 1 wherein said facial features include a corner of an eye.

3. The method of claim 2 wherein said facial features include a corner of a mouth.

4. The method of claim 3 wherein said facial features include a tip of nose.

5. The method of claim 4 wherein said facial features includes sides of said nose.

6. The method of claim 5 wherein said facial features includes a facial outline.

7. The method of claim 1 wherein said facial features include corner of a mouth.

8. The method of claim 1 wherein said facial features include tip of nose.

9. The method of claim 8 wherein said facial features includes sides of said nose.

10. The method of claim 1 wherein said facial features includes facial outline.

11. The method of claim 1 wherein said method of defining said feature location includes matching feature templates of an eye.

12. The method of claim 1 wherein said method of defining said feature location includes matching feature templates of a mouth.

13. The method of claim 1 wherein said method of defining said feature location includes matching feature templates of a nose.

14. The method of claim 1 wherein said method of defining said feature location includes matching feature templates of a facial outline.

15. The method of claim 1 wherein said method of defining said feature location includes matching feature templates of a facial outline, an eye, a nose and a mouth.

16. The method of claim 1, wherein said affine transform is used in the upper and lower portions of said face and said bilinear transform is used in the middle part of said face.