Enhanced Two-Dimension Scatter Difference Discriminant Analysis for Face Recognition

Abstract

A novel model for image feature extraction and recognition called enhanced two-dimension scatter difference discriminant analysis (E2DSDD) is presented in the paper. 2DSDD can extract less coefficients than the traditional two-dimension scatter difference discriminant analysis (2DSDD) for image representation and lead to faster classification. In addition, a new feature selection scheme is suggested for the selection of the most discriminative features. Experiments on the ORL face databases show E2DSDD outperforms the current 2DSDD, 2DLDA and 2DPCA algorithms in its computation efficiency and recognition performance.