Face Recognition with Statistical Local Line Binary Pattern

In this paper, we present an improved method for face recognition named statistical Local Line Binary Pattern (SLLBP), which is based on Local Binary Pattern (LBP). The basic idea of the algorithm is to first obtain the local binary code along with horizontal and vertical direction separately and the gray-level difference between neighbor pixels, which is also important to describe a face feature and often been ignored, is then computed. To deal with the gray-level difference we firstly take a statistical analysis on the probability distribution of the gray-level difference values, and then a mapping function is used to encode the range of these values. Our algorithm is tested on the public ORL face database. The experimental results show that the method improves the performance significantly.