Face recognition based on fusion of multilevel centre-symmetric local derivative pattern

In order to extract richer face texture feature, multi-level centre-symmetric local derivative pattern (CS-LDP) algorithm is proposed. Firstly, CS-LDP is utilized to extract the first level features from the original face image; Secondly, features are extracted from the processed image by the former step using CS-LDP again; lastly, we can get multi-level texture features. The multi-level features can be fused into a whole feature vectors. Experimental results on the ORL and Yaleb faces databases demonstrate that the algorithm of fusing the multi-level CS-LDP features can improve the face recognition accuracy.