Dual-kernel based 2D linear discriminant analysis for face recognition

Abstract

This paper proposes a new image feature extraction method for face recognition, called dual-kernel based two dimensional linear discriminant analysis (D-K2DLDA), by integrating multiple kernel discriminant analysis with the existing K2DFDA method. The proposed method deals with a face image directly as a matrix, instead of a stacked vector from rows or columns of the image. Moreover, we separately perform an iterative scheme for kernel parameter optimization for each of the two kernels, based on the maximum margin criterion and the damped Newton’s method, followed by a fusion procedure of the two kernels. Experimental results on the ORL and UMIST face databases show the effectiveness of D-K2DLDA.