Robustness and expression independence in 3D face recognition

摘要

We describe a robust method for 3D face recognition under variance of facial expressions. The method utilizes the identical areas on two facial images as the measurement of distance. To segment the identical area, partial shape matching is performed using closest point registration and level set method. The segmentation problem is formalized into an Eikonal equation, which can be efficiently solved by fast marching method. The presented 3D face recognition method shows a very promising recognition rate.