Facial model generation is an important issue in the model- based applications, such as MPEG-4 and the virtual reality. An effective and precise construction algorithm of 3D facial model from 2D images should be available for practical applications. To generate facial model usually requires stereoscopic view of the face in the pre-processing stage. Although facial model can be successfully estimated from two stereo facial images, the occlusion effect and imprecise location of the feature point prohibit us from obtaining an accurate facial model. In this paper, we proposed several facial model estimation algorithms to find the precise facial model form a stereo or mono image sequence. The information of head movement, which is recorded in the image sequence, in the temporal domain is utilized for the facial model estimation. Even though the a prior information about the 3D position of the head with respect to the camera and the rotation axis and angle of the head's movement are unknown, an accurate facial model (within 7.21%) error) can still be obtained by our schemes. In addition, our schemes do not require the precise camera parameters and avoid the tedious camera calibration such that the facial model generation is easily achieved.

Date Published: 28 June 2000
PDF: 12 pages
Proc. SPIE 4080, Input/Output and Imaging Technolgies II, (28 June 2000); doi: 10.1117/12.389454

Published in SPIE Proceedings Vol. 4080:
Input/Output and Imaging Technolgies II
Yung-Sheng Liu; Thomas S. Huang, Editor(s)