A modularly vectorized two dimensional LDA for face recognition

In this paper, a modularly vectorized 2DLDA (Mv2DLDA) is proposed for face recognition. First, the original images are divided into modular blocks. Then, each sub-block is transformed into a vector. By using column vector to represent each modular block, we can obtain a two dimensional matrix representation for image. Finally 2DLDA is applied directly on these 2D matrices. Experimental results on ORL, Yale B and PIE databases show that the proposed method can achieve better recognition performance in comparison with RLDA, 2DPCA and 2DLDA.