Discriminant feature extraction using dual-objective optimization model

In this paper, we establish a dual-objective optimization (DOO) model for discriminating feature extraction, in the sense that the optimizations of between-class scatter and within-class scatter are taken into investigation separately rather than simultaneously through a quotient like Fisher criterion. Based on the various solutions of the proposed model, we outline the optimization strategy of null space of within-class scatter matrix and the framework of complex PCA for classification purpose. We test the performance of the proposed algorithms on the ORL and Yale face databases. The experimental results show that the proposed algorithms are effective. Particularly, the complex PCA enhanced by complex LDA appears to be the best among the considered algorithms in terms of recognition performance and is robust against noises.