A Complete Fisher Discriminant Analysis for Based Image Matrix and Its Application to Face Biometrics

Risco Mulwani Mutelo, Wai Lok Woo, Satnam Singh Dlay

Abstract: This paper presents a Complete Orthogonal Image discriminant (COID) method and its application to biometric face recognition. The novelty of the COID method comes from 1) the derivation of two kinds of image discriminant features, image regular and image irregular, in the feature extraction stage and 2) the development of the Complete OID (COID) features-based on the fusion of the two kinds of image discriminant features used in classification. Firstly, the COID method first derives a feature image of the face image with reduced dimensionality of the image matrix by means of two dimensional principal component analysis and then performs discriminant analysis in a double discriminant subspaces in order to derive the image regular and irregular features making it more suitable for small sample size problem. Finally combines the image regular and irregular features which are complementary for achieving better discriminant features. The feasibility of the COID method has been successfully tested using the ORL images where it was 73.8% more superior to 2DFLD method on face recognition.