Dual 2DLDA Based on Gabor Features for Face Recognition

Subspace learning techniques for face recognition have been attached growing attention. Two dimensional linear discriminate analysis (2DLDA) is a popular face recognition technique, which learns two interrelated subspace in an iterative manner. DialLDA is proposed as a complementary method to 2DLDA. Motivated by the success of 2DLDA and DialLDA for face recognition, in the paper we develop an innovative algorithm named Dual-2DLDA method, and it integrates the two complementary methods 2DLDA and DialLDA to learn multiple subspaces by utilizing Gabor features extracted from the training data.The proposed method can fully extract the information in the training data effectively. We have experimentally compared our method to other popular feature extraction methods, such as 2DPCA, 2DLDA. Experimental results on ORL databases show that the proposed method is superior to the popular methods.