Recognizing Faces by Weakly Orthogonalizing Against Perturbations

Abstract

In this paper, we address the problem of face recognition under drastic changes of the imaging processes through which the facial images are acquired. A new method is proposed. Unlike the conventional algorithms that use only the face features, the present method exploits the statistical information of the variations between the face image sets being compared, in addition to the features of the faces themselves. To incorporate the face and perturbation features for recognition, a technique called weak orthogonalization of the two subspaces has been developed that transforms the two overlapped subspaces such that the volume of the intersection of the resulting two subspaces is minimized. Matching is performed in the transformed face space that has thus been weakly orthogonalized against perturbation space. Results using real pictures of the frontal faces from drivers' licenses demonstrate the effectiveness of the new algorithm.