基于CS-LDP和LCCP特征融合的人脸识别算法

Face Recognition Based on Feature Fusion of CS-LDP and LCCP

人脸识别是当前模式识别和人工智能的研究热点,论文提出中心对称的局部二阶微分模式(center-symmetriclocal derivative pattern,CS-LDP)和中心对称二阶局部二值模式(local center-symmetric pattern,LCCP)特征融合的算法。该算法对图像分别提取CS-LDP特征和LCCP特征,并将两个特征融合得到最终的特征向量,最后通过计算直方图欧式距离来得到人脸识别结果。实验结果表明,CS-LDP提取图像的二阶微分特征,LCCP提取图像的凹凸特征,融合两种特征得到更为有效的图像特征的识别信息,在ORL、AR和Yale B人脸数据库上实验,相对于CS-LDP算法和LCCP算法,识别率均得到提高。

Face recognition is an active research area in the artificial intelligence. In this paper,a hybrid method of center-symmetric local derivative pattern（CS-LDP）and local center-symmetric pattern（LCCP）algorithm are proposed. The algorithm of the image is extract CS-LDP feature information and LCCP feature information. Further more,the features calculated by CS-LDP and LCCP are combined to obtain the final texture features. Finally,The results of face recognition are obtained by calculating the histogram of Euclidean distance. The results show that CS-LDP extracts the second order differential feature and LCCP extracts the irregularities feature of images. Then effective image feature recognition information by integration of two features is gotten.Then experiments on ORL,AR and Yale B face database and done. Compared to CS-BP algorithm and LBP algorithm,the recognition rate is improved.