Optimization of LBP parameters

Abstract

In this paper, we propose the optimal parameters of local binary patterns such as type of pattern, size of blocks in the feature space and distance measure for face recognition using a genetic algorithm. The genetic algorithm is able to optimize all these parameters quickly and to improve the recognition accuracy. We provide a comparative study of three types of local binary patterns (LBP, LGP and NRLBP) and four distance measures (L1, L2, χ2, EMD). The genetic algorithm is also used to optimize parameters such as dimension of histograms. Our results are tested on three different face databases which have the similar properties. We can set these optimal parameters into our face recognition system suitable for the next-generation of hybrid broadcast broadband television.