Using geometry modeling to find pose invariant features in face recognition

Face recognition has been an important topic in computer vision for the last two decades. While many algorithms have been developed to address this issue, one of the major challenges faced by them is variation in pose. One of the possible solutions is to find invariant features among different poses of a single person. In this paper a geometry mapping between a frontal face and its rotated pose is used to find invariant features for pose robust face recognition. This mapping is solely based on the angle of rotation and indicates mutual regions between a frontal view of a person and its rotated image. Invariant features based on the low frequency coefficients of these mutual regions are then extracted and used for recognition. Simulations performed on the ldquoYale Face DatabaseBrdquo show promising results for pose invariant face recognition.