Generally, the centersymmetric local binary pattern (CSLBP) is used to extract features from the  face images only once, by which the extracted texture features are not adequate to represent the face images. Therefore, we employ CSLBP to extract more abundant and informative texture features for more times, and a new face recognition method is proposed which is on the basis of the fusion of multilevel centersymmetric local binary pattern features. In this method, first, the CSLBP is utilized to extract the first level features from the original face image; then, the second level features are extracted from the  feature image by CSLBP again; likewise, we can obtain the multilevel texture features and then fuse different level features to represent face images. The experimental results on the ORL and Yale face databases demonstrate that compared with one level face image features, the method of fusing the  multilevel CSLBP features can improve the face recognition accuracy obviously.