Facial Image Recognition Based on a Statistical Uncorrelated Near Class Discriminant Approach

• Published in: IEICE Transactions on Information and Systems Vol. E93.D; no. 4; pp. 934 - 937
• Main Authors: LI, Sheng, JING, Xiao-Yuan, BIAN, Lu-Sha, GAO, Shi-Qiang, LIU, Qian, YAO, Yong-Fang
• Format: Journal Article
• Language: English
• Published: Oxford The Institute of Electronics, Information and Communication Engineers 2010; Oxford University Press
• Subjects: Applied sciences; Artificial intelligence; Computer science; control theory; systems; Exact sciences and technology; face recognition; Image processing; Information systems. Data bases; Information, signal and communications theory; locally statistical uncorrelated constraints; Memory organisation. Data processing; near classes; Pattern recognition; Pattern recognition. Digital image processing. Computational geometry; Signal processing; Software; statistical uncorrelated near class discriminant (SUNCD); Telecommunications and information theory; Biometrics; Performance evaluation; Nearest neighbour; Image recognition; Face recognition; Image processing; statistical uncorrelated near class discriminant (SUNCD); Pattern recognition; near classes; Fisher information; locally statistical uncorrelated constraints; Database; Automatic recognition
• ISSN: 0916-8532; 1745-1361
• DOI: 10.1587/transinf.E93.D.934

In this letter, a statistical uncorrelated near class discriminant (SUNCD) approach is proposed for face recognition. The optimal discriminant vector obtained by this approach can differentiate one class and its near classes, i.e., its nearest neighbor classes, by constructing the specific between-class and within-class scatter matrices and using the Fisher criterion. In this manner, SUNCD acquires all discriminant vectors class by class. Furthermore, SUNCD makes every discriminant vector satisfy locally statistical uncorrelated constraints by using the corresponding class and part of its most neighboring classes. Experiments on the public AR face database demonstrate that the proposed approach outperforms several representative discriminant methods.