Gabor-2DLDA: Face Recognition Using Gabor Features and 2D Linear Discriminant Analysis

An effective face recognition method is described in the proposed paper, which is based on Gabor wavelets and 2D linear discriminant analysis (Gabor-2DLDA). Although Gabor features has been recognized as one of the most successful face representations, its huge number of features often brings about the problem of curse of dimensionality. In this paper, we use Gabor feature matrix to represent the facial features, and then apply 2DLDA to derive subspaces from Gabor feature matrix, thus effectively addressing the issue of dimensional disaster and avoiding the singularity problem of linear discriminant analysis method. Finally, support vector machine (SVM) is applied to classify the extracted face features. Experimental results on ORL database and subset of CAS-PEAL database show that the combination of Gabor-2DLDA with SVM can achieve promising results.