Model based faced coding and decoding using feature detection and eigenface coding
First Claim
Patent Images
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.
1 Assignment
0 Petitions
Accused Products
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.
187 Citations
16 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. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
Specification