Illumination and pose variable face recognition via adaptively weighted ULBP_MHOG and WSRC

Illumination and pose variable face recognition is an important and challenging task in computer vision. To solve the problem that the accuracy of face recognition reduces with illumination and pose changing, this paper proposes an adaptively weighted ULBP_MHOG and WSRC method. In the proposed method, we first normalize illumination for face images, and extract Uniform Local Binary Pattern (ULBP) and multiple Histogram of Oriented Gradient (MHOG) features in each block which are called ULBP_MHOG features. Then we use information entropy to obtain adaptively weighted ULBP_ MHOG (WULBP_MHOG) features. Finally, test face images can be classified via weighted sparse representation (WSRC). The comparison experiments with different blocks, features, classifiers and the state-of-the-art methods have been conducted on the ORL, the Yale, the Yale B, the Extended Yale B and CMU-PIE databases. Experimental results show our method can improve the accuracy effectively in illumination and pose variable face recognition.