Face description based on adaptive local weighted Gabor comprehensive histogram feature

Face recognition is an extensively research topic in pattern recognition and image processing fields due to its broad application prospects in many areas such as counter terrorism, access identification, electronic passport, e-government affairs, etc. Inspired by the Gabor phase information and local image information content, a novel face description algorithm using adaptive Local Weighted Gabor Comprehensive Histogram (LWGCH) is proposed. It consists of two components: Local Gabor Comprehensive Histogram (LGCH) and Contribution Map (CM). Actually, the adaptive local weighted Gabor comprehensive histogram is generated with LGCH weighed by CM calculated by information content model. Finally, Extensive experiments on ORL, YALE, CMUPIE and Yale B face databases validate the effectiveness of the proposed methods under various conditions of partial occlusion, complex illumination, different expressions and poses. Experimental results indicate that the proposed algorithm is a competitive and robust method compared with several state-of-the-art approaches.