LI Wen, CHEN Xi*, LIU Zengli, HUANG Qingsong
(School of Information Engineering and Automation, Kunming Science and Technology University, Kunming 650500, Yunnan, China)
Abstract:
In order to extract richer facial texture feature and improve the accuracy of face recognition, a hybrid method of local binary pattern (LBP) and center-symmetric local derivative pattern(CS-LDP) algorithm is proposed. The LBP is utilized to extract the first level features from the original face image. Then using CS-LDP method, the features are extracted from images processed by LBP.Furthermore, the features calculated by LBP and CS-LDP are combined to obtain the final texture features. The Histogram intersection distance is calculated to find the similarity degree between two texture features vectors.The results show that LBP extracts the first-order differential features of image, while CS-LDP extracts the second order differential feature. Fusing such two features can obtain richer texture information. The mean recognition accuracy reaches above 90% on ORL, YaleB and FERET face databases, which provides a feasible scheme for practical face recognition.
KeyWords:
LBP; CS-LDP; feature fusing; face recognition
