SDF Matched Filter Based on Gabor Wavelet Transform for Face Recognition

A novel face recognition method is proposed based on synthetic discriminant function(SDF) matched filters using Gabor wavelet features.In this method,the generic training face database is used,and the corresponding SDF matched filters are generated in the Gabor feature space.Each test image is projected onto those non-orthogonal basis vectors to yield a group of correlative eigenvectors,and the similarity between two face images is measured using the distance between the corresponding eigenvectors.The adoption of Gabor transform,SDF matched filter and class-dependence feature analysis (CFA) enables the proposed method to be robust to the variations of illumination and expression.In addition,this method is of good generalizability.Experimental comparison on FERET face database verifies the validity of the proposed method.