Face Recognition under Varying Illumination Based on Gradientface and Local Features

Illumination variation is among the several bottlenecks in a practical face recognition system. Extracting illumination-invariant features, such as the gradient-based descriptor, is an effective method to deal with this problem and shows outstanding performance. In this paper, a novel illumination-invariant, histogram-based descriptor, namely local Gradientface XOR and binary pattern (LGXBP), is proposed to enhance the recognition performance of gradient-based method under varying lighting conditions. To this end, first, this approach transforms a face image into the logarithm domain and obtains a pair of illumination-insensitive components. Then the two components are encoded further by local XOR patterns and local binary patterns, respectively, to get compact descriptors and histogram representations, which not only preserve illumination invariance but also gain more discriminating power. Finally, the complementary histograms are integrated for face recognition. Comparative experimental results verify the effectiveness of our approach and show that LGXBP achieves superior recognition rates and a high degree of stability under varying illumination conditions, outperforming most of the state-of-the-art methods. (c) 2014 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.