MQI based face recognition under uneven illumination

Abstract

Face recognition has been applied in many fields, while face recognition under uneven illumination is still an open problem. Our approach is based on Morphological Quotient Image (MQI) for illumination normalization, and Dynamic Morphological Quotient Image (DMQI) is proposed to improve the performance. Before applying MQI, singularity noise should be removed, and after MQI operation, an effective scheme is used to wipe off the grainy noise as postprocessing. Weighted normalized correlation is adopted to measure the similarity between two images. Experiments on Yale Face Database B show that the proposed MQI method has a good performance of face recognition under various light conditions. Moreover, its computational cost is very low. © Springer-Verlag Berlin Heidelberg 2007.