Elastic registration of noisy and expression 3D faces

Non-rigid registration of 3D faces finds a variety of applications in computer vision, human-computer interaction and graphics. However, automatic non-rigid registration is a non-trival problem. In this paper, we introduce an elastic deformation based method for the non-rigid registration of 3D faces that are subject to large deformations due to expressions and that are contaminated by sensor noise. Registration is based on the matching of the geometrical and textural similarities between two faces. Mismatches give rise to a force field, and the resulting motion must be regularized. We use a non-linear elastic model and we realize it in 2D space after mapping faces from the 3D domain. Partial correspondences due to the missing data are handled with extrapolation. Experiments show that the algorithm works successfully and efficiently.