Abstract
The morphable modeling method of reconstructing 3D human face is based on a single given image, and the model fitting is ambiguous owing to the lack of depth information and geometry constraints between images. In this paper, we introduce a new face reconstruction method from multiple views based on the morphable model. The morphable face model is matched with multiple face images from different views. Using the multi-view geometry constraints, the ambiguity is efficiently eliminated and more realistic results are provided. We test our algorithm on the MPI face database and real images from our multi-camera system. The experimental results show the efficiency of the proposed method.