有效用于人脸识别的光照不变特征表示算法  

Efficient illumination-robust features representation algorithm for face recognition

作　　者：孔锐[1,2] 揭英达 KONG Rui;JIE Yingda(College of Information and Science Technology, Jinan University, Guangzhou 510632, China;College of Electrical and Information, Jinan University, Zhuhai, Guangdong 519070, China)

机构地区：[1]暨南大学信息科学技术学院,广州510632 [2]暨南大学电气信息学院,广东珠海519070

出　　处：《计算机工程与应用》2017年第1期147-152,共6页Computer Engineering and Applications

基　　金：广东省学科建设专项资金项目-科技创新(No.2013KJCX0023);珠海市公共技术服务平台科技项目(No.2013D0501990013)

摘　　要：在光照变化环境下,人脸识别的鲁棒性是人脸识别系统中一大挑战。针对光照变化对人脸识别的影响,对经典光照不变特征表示算法进行了研究,提出一种基于局部标准差光照不变的人脸特征表示算法及其加权形式。结合完备线性鉴别分析(Complete-Linear Discriminant Analysis,C-LDA)算法提取特征,在Extended Yale-B与YALE人脸库中,与其他处理光照变化的经典方法相比,如多尺度Retinex(Multi Scale Retinex,MSR)、韦伯脸(Weber-Face,WF)和局部归一化(Local Normalization,LN),提出的算法能获得更高识别率。In the change illumination conditions which can not be controlled, the robustness of face recognition is achallenge in face recognition system. Concerning the illumination changes on the influence of the recognition rate in facerecognition, a new face recognition algorithm for varying illumination environment is proposed according to the classicillumination invariant feature representation algorithm, which is based on Local Standard Deviation and its weighted form.Compared with the existed representative approaches, such as WF(Weber-Face), MSR(Multi Scale Retinex)and LN(Local Normalization), the novel algorithm can obtain higher recognition rate on the Extended Yale-B Database and theYALE Face Database when combined with the analysis of complete linear discriminant(Complete-Linear DiscriminantAnalysis, C-LDA)algorithm for feature extraction.

关 键 词：光照不变特征表示 人脸识别 局部标准差 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]