Realistic Face Modeling by Registration of a 3-D Mesh Model and Multi-View Color Images

Realistic 3-D face modeling is a challenging research topic that involves tight integration of computer vision and computer graphics techniques. Acquisition of surface textures is an indispensable step in realistic 3-D face modeling. In this paper, we propose a camera-calibration based method to register 3-D mesh models with multi-view color images. Then a global texture map is synthesized from texture patches corresponding to each image. A main difficulty here is that the texture-mapping parameters are sensitive to initially selected correspondences between the mesh model and images. To solve this problem, we developed an iterative optimization algorithm for the parameter estimation through automatic adjustment of the selected correspondences. Experimental results are given to demonstrate the performance of our method.