An improved local binary pattern for texture classification

This paper proposes an improved robust descriptor: the Joint Local Binary Patterns with Weber-like responses (JLBPW). The JLBPW descriptor mainly solves two problems of the existing Local Binary Pattern (LBP) method. The existing LBP methods discard the local intensity differences by holding their binary information. And the correlation of patterns under different scales is also ignored when the multi-resolution technique is used by the LBP methods. Aiming at the two problems, the JLBPW descriptor organizes the ignored local intensity differences according to the Weber's law that is discovered from psychological experiments. And the organized differences are set as responses of the corresponding joint patterns. The proposed descriptor supplies more information to describe local structures of images. Experiments on two representative databases demonstrate that the proposed descriptor is superior to other modern approaches.