| US 7,415,152 B2 | ||
| Method and system for constructing a 3D representation of a face from a 2D representation | ||
| Dalong Jiang, Beijing (China); Hong-Jiang Zhang, Beijing (China); Lei Zhang, Beijing (China); Shuicheng Yan, Hong Kong (China); and Yuxiao Hu, Urbana, Ill. (US) | ||
| Assigned to Microsoft Corporation, Redmond, Wash. (US) | ||
| Filed on Apr. 29, 2005, as Appl. No. 11/118,116. | ||
| Prior Publication US 2006/0245639 A1, Nov. 02, 2006 | ||
| Int. Cl. G06K 9/00 (2006.01) | ||
| U.S. Cl. 382—154 [382/118; 382/280; 345/473] | 16 Claims |
| 1. A method in a computer system for generating a 3D image of an object, the method comprising:
providing a 3D model of an object, the 3D model including a 3D image of a standard object and parameters indicating variations
of an individual object;
receiving a 2D image of the object;
locating feature points of the 2D image of the object;
calculating parameters that map the located feature points of the 2D image to feature points of a 2D representation of the
3D image of the standard object;
computing 3D feature points of the 3D image of the object using the provided 3D model and the calculated parameters;
after computing the 3D feature points, interpolating 3D non-feature points of the 3D image from the computed 3D feature points;
and
after the 3D non-feature points of the 3D image is interpolated, texture mapping the 2D image of the object to the 3D image
with the computed 3D feature points and interpolated 3D non-feature points
wherein the 3D model is represented by the following equation:
S′=S+Pα where S=(X1, Y1, Z1, X2, . . . Yn, Zn)T ∈3n, X, Y, and Z are coordinates of the n points, S represents the standard object, P ∈3n×m is the matrix of the first m eigenvectors of S, and α=(α1, α2, . . . , αm)T ∈m are the parameters.
|