Abstarct: One of the biometric ways for identifying people could be using visage’s features. As in this way physical contacts with persons are not needed, so it is very noticed socially. Two dimensional modern identifying systems, do well in controlled and steady conditions, although changes in image’s lightings and visage’s different gestures are going to be it’s problems. But, three dimensional images can show objects regardless light level clearly. This study aims to improve function of face’s identifications systems in improper conditions such as lighting, gesture and uncontrolled positions by using three dimensional images information. This study is done on the baxse of data station (GAVAB) with images included different improper kinds of positions in identifying visages such as laughing, up and down glance, turning head to left and right and also accidental positions of face. Methods of this study are categorized into two holistic and component baxsed methods. In holistic methods we have used three characteristics of depth, mean curvature and Gaussian curvature. The best conclusion in holistic methods due to the mean curvatures features with 91.9 percent accuracy. Simplicity and high sped are the privileges of this method. In component baxsed methods we have used features of complete local binary patterns (CLBP), binary patterns baxsed on variance (LBPV), binary patterns baxsed on Gaussian curvature (LBPK) and binary patterns baxsed on mean curvature (LBPH). Among these component baxsed methods, best conclusion goes to LBP_S/M/C method with 96.47 percent accuracy. Also in this study two specific new extraction methods (LBPH, LBPK) are introduced.