Feature extraction using two-dimensional local graph embedding based on maximum margin criterion

In this paper, we propose a novel method for image feature extraction, namely the two-dimensional local graph embedding, which is based on maximum margin criterion and thus not necessary to convert the image matrix into high-dimensional image vector and directly avoid computing the inverse matrix in the discriminant criterion. This method directly learns the optimal projective vectors from 2D image matrices by simultaneously considering local graph embedding and maximum margin criterion. The proposed method avoids huge feature matrix problem in Eigenfaces, Fisherfaces, Laplacianfaces, maximum margin criterion (MMC) and inverse matrix in 2D Fisherfaces, 2D Laplacianfaces and 2D Local Graph Embedding Discriminant Analysis (2DLGEDA) so that computational time would be saved for feature extraction. Experimental results on the Yale and the USPS databases show the effectiveness of the proposed method under various experimental conditions.