Three-level face features for face recognition based on center-symmetric Local Binary Pattern

Abstract

Recently, Local Binary Pattern (LBP) operator has been applied into face recognition successfully. Center-symmetric Local Binary Pattern is a variation operator of LBP, which can not only reduce the dimensionality of LBP features, but also capture the gradient information better than basic LBP. In this paper, an algorithm is proposed for face recognition based on center-symmetric LBP. The main idea of the algorithm is using the center-symmetric LBP operator three times for extracting three-level face features with the assumption that a gray face image encoded by center-symmetric LBP operator is still a gray image. Since different level face features make different contributions to face recognition, we put appropriate weighting on each level face feature. The experimental results show that the proposed algorithm performances superiority over the comparative algorithms on ORL and the subset of FERET face databases.