Null Space based Kernel Fisher Discriminant Analysis for Face Recognition 01301558英文电子书.pdf

Null Space-based Kernel Fisher Discriminant Analysis for Face Recognition 1 1 2 1 Wei Liu , Yunhong Wang , Stan Z. Li , Tieniu Tan 1 National Laboratory of Pattern Recognition, Institute of Automation, Ch inese Academy of Sciences, Beijing, P. R. China, 100080 E-mails: {wliu, wangyh, tnt}@ 2 Microsoft Research Asia, Beijing Sigma Center, Beijing, P. R. China, 100080 E-mail: szli@ guaranteed not more than the rank of Sw so as to obtain a Abstract full-rank within-class scatter matrix. Then standard LDA is used to extract and represent facial features. All these The null space-based LDA takes full advantage of the methods above do not consider the importance of null null space while the other methods remove the null space. space of the within-class scatter matrix, and remove the It proves to be optimal in performance. From the null space to make the resulting within-class scatter theoretical analysis, we present the NLDA algorithm and full-rank.