Robust Discriminant Analysis of Gabor Feature for Face Recognition

This paper introduces a robust discriminant analysis of Gabor feature (RGF) method for face recognition. The RGF method apply a novel robust Fisher linear discriminant model (RFM) to the low dimensional Gabor feature defined by principle component analysis. The robustness of the RGF method comes form the new RFM which improves the generalization capability of the FLD by robust estimate of the within-class scatter matrix. The feasibility of the RGF method has been successfully tested on face recognition using 1400 images from ORL and FERET database. The effectiveness of the RGF method is shown in term of both the excellent accuracy (99% and 97.75%) and the comparative performance against some popular face recognition schemes such as the eigenfaces, the Fisherfaces, and the Gabor-Fisher classifier method.