基于多通道Gabor滤波与CS-LBP的人脸识别方法  

Face Recognition Method Based on Multi-channel Gabor Filtering and Center-Symmetric Local Binary Pattern

作　　者：何中市[1] 卢建云[1] 余磊[1] 

机构地区：[1]重庆大学计算机学院,重庆400030

出　　处：《计算机科学》2010年第5期261-264,共4页Computer Science

基　　金：国家863计划项目(2007AA01Z423);重庆市自然科学基金项目(CSTC;2007BB2134)资助

摘　　要：近来,局部二值模式(Local Binary Pattern,LBP)在人脸识别中取得了成功应用。然而,LBP提取的特征维数通常很高。而中心对称局部二值模式(Center-Symmetric Local Binary Pattern,CS-LBP)采用中心对称思想对图像进行编码,能够显著降低提取的特征的维数。为此,将CS-LBP应用于人脸图像特征提取,并结合多通道Gabor滤波,提出了基于多通道Gabor滤波与CS-LBP的人脸识别算法。在Yale,ORL,FETER标准人脸库上的实验结果表明,相比局部二值模式,CS-LBP以提取更少的特征维数取得了相当的识别率,并且,基于多通道Gabor滤波的CS-LBP能显著提高识别精度。In recent years, Local Binary Pattern （LISP） has been successfully applied to face recognition. However, the dimensionality of the feature vector extracted by LISP is usually very high. On the contrary, Center-Symmetric Local Binary Pattern （CS-LBP） encodes the image with the technique of Center Symmetry. As a result,CS-LBP can largely reduce the extracted feature dimension. Therefore, in this paper, CS-LISP was employed to extract features from facial images,and then a new face recognition algorithm based on multi channel Gabor filtering （MCGF） and CS-;BP was brought forward. Experimental results on Yale, ORL and FETER face databases demonstrate that compared with LBP,CS-LISP can achieve the comparable performance in terms of recognition rate with lower feature dimensionality. Additionally, CS-LBP based on MCGF can increase the accuracy obviously.

关 键 词：中心对称局部二值模式 多通道GABOR滤波 特征提取 人脸识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]