Abstract: We proposed a face recognition representation method, called multi scale block Local Binary Pattern. The Local Binary Pattern(LBP) was proved to be effective for image representation,but it was too local to be robust. In MS-BLBP, the computation was done based on average values of block subregions, instead of individual pixels. The face area was first divided into small regions from which Block Local Binary Patterns (BLBP) with different weights histograms were extracted and concatenated into a single, spatially enhanced feature histogram efficiently representing the face image. The classification was performed using a nearest neighbour classifier in the computed feature space with Chi square as a dissimilarity measure. Experiments in face databases show that the proposed MS-BLBP method outperforms other LBP based face recognition algorithms.

Key words: Local binary pattern, Block local binary pattern, Face recognition, Dissimilarity measure