An Illumination-Insensitive Face Matching Algorithm

[[abstract]]Face matching is an essential step for face recognition and face verification. It is difficult to achieve robust face matching under various image acquisition conditions. In this paper, an illumination-insensitive face imagematching algorithm is proposed. This algorithm is based on an accumulated consistency measure of corresponding normalized gradients at face contour locations between two comparing face images under different lighting conditions. To solve the matching problem due to lighting changes between two face images, we first use a consistency measure, which is defined by the inner product between two normalized gradient vectors at the corresponding locations in the two images. Then we compute the sum of the individual consistency measures of the normalized gradients at all the contour pixels to be the robust matching measure between two face images. To better compensate for lighting variations, three face images with very different lighting directions for each person are used for robust face image matching. The Yale Face Database, which contains images acquired under three different lighting conditions for each person, are used to test the proposed algorithm. The experimental results show good recognition results under different lighting conditions by using the proposed illuminationinsensitive face matching algorithm.[[fileno]]2030206010016[[department]]資訊工程學