Face Recognition Algorithm Based On Monogenic Binary Coding

Face recognition has become a research focus in pattern recognition, statistical learning theory, applied mathematics and information safety, cause of its many applications in various domains.Local feature based face recognition could achieve state-of-the-art results in large-scale face database such as FERET, because its robustness on the variousness of illumination, expression and occlusion. In this paper, we present a locale statistics feature:Monogenic Binary Coding based face recognition algorithm. The main contribution of this paper is as following:First, we begin with the one-dimensional analytic signal; study its two-dimension generalization, monogenic signal representation. Then we obtain the monogenic representation of the face.Second, we adopt a different strategy to get the binary quantification of the monogenic representation of the face. And we add monogenic local imagery intensity encoding to the local variation encodes. This encoding method increases the divisibility of the image, thereby enhancing the recognition result.Third, we apply modular PCA and LDA algorithm to the obtained MBC feature map to reduce the dimensions of the features and the redundant information. Then we weighted each block based on its contribution to the final result by Logistic regression these weights are used to gain the final similarity.Last, the algorithm designed in this paper is tested on FERET face database. when do the recognition experiment, the result are 99.7%,99.5%,93.6%,91.5%rank one recognition rates on the subsets Fb, Fc, DupⅠ and DupⅡ. Then we make a standalone verification test on Dupl, we achieve VR=93.21% when FAR=0.001.