Adaptive Local Binary Patterns for 3D Face Recognition

Local binary patterns (LBP) method has been successfully applied into many texture classification areas. However, the most widely used uniform local binary patterns are defined according to the general texture micro-structures, which are not optimal for some specific application. In this paper we propose a novel adaptive local binary patterns (ALBP) method, which can adaptively choose the most suitable patterns according to its tasks. ALBP is tested on the challenging CASIA 3D face database in two situations: the adaptive patterns obtained from the 3D face image (global texture) and the adaptive patterns obtained from the Gabor features (local texture). Experimental results illustrate the effectiveness of our proposed method.