作者: Xiaoyuan Jing , Sheng Li , Yongfang Yao , Lusha Bian , Jingyu Yang
关键词:
摘要: In this paper, a kernel uncorrelated adjacent-class discriminant analysis (KUADA) approach is proposed for image recognition. The optimal nonlinear vector obtained by can differentiate one class and its adjacent classes, i.e., nearest neighbor constructing the specific between-class within-class scatter matrices in space using Fisher criterion. manner, KUADA acquires all vectors class. Furthermore, makes every satisfy locally statistical constraints corresponding part of most classes. Experimental results on public AR CAS-PEAL face databases demonstrate that outperforms several representative methods.