A new image division for LBP method to improve face recognition under varying lighting conditions

Local binary patterns (LBP) is one of the most used methods in face recognition. This paper presents a different way of obtaining the regions that are used to construct the LBP histograms, in order to improve its performance in front of illumination problems. The proposed method takes into account the shape of the face to build a triangular mesh in which a better description of the face image through LBP is achieved. Experimental results conducted on Yale B database show that under varying lighting conditions, the proposal improves the performance of the traditional rectangular division of the LBP method.