Gabor Feature-Based Face Recognition Using Median MSD

This paper introduces a novel Gabor-based median maximum scatter difference (GMMSD) method for face recognition. Maximum scatter difference (MSD) is a recently proposed linear discriminative method for dimensionality reduction. Its theory is similar to linear discriminative analysis (LDA). In this paper, we investigate its extension, called median MSD, in which the within-class mean is replaced with within-class median. The GMMSD method, which is robust to variations of illumination and facial expression, applies the MMSD to an augmented Gabor feature vector derived from the Gabor wavelet representation of face images. We performed comparative experiments of various face recognition schemes, including the proposed GMMSD method, PCA method, LDA method, MSD (also GMSD and MMSD) method, the combination of Gabor and PCA method (GPCA) and the combination of Gabor and LDA method (GLDA). Experimental results on CAS-PEAL database and FERET database show superior of the novel GMMSD method.