A reliable face recognition algorithm based on matching normalized gradients

Face recognition is a challenging problem that has attracted a number of researchers from academia and industry in recent years. One of the main challenges in face image recognition is due to illumination variations. Face matching is a crucial step for these problems. A reliable face recognition algorithm robust against illumination variation is proposed. The basic idea in our method is to accumulate the consistency measure of corresponding normalized gradients at the face contour locations from two comparing face images. The consistency measure is defined to be the inner product between two normalized gradient vectors at the corresponding locations in the two images. To compensate for lighting variations, three face images captured with very different lighting directions of each person are integrated into our face recognition method. We tested the proposed algorithm on the Yale face database B and compared with three other methods. Satisfactory recognition results are shown from face recognition experiments under different illumination conditions.