PCA and LDA based face recognition under changing class sizes and non-uniform illumination conditions

Abstract

Face recognition is a vital research area with some important challenges mainly concerning images taken in non-ideal illumination conditions. In this paper the PCA and LDA based face recognition methods are considered, where the recognition performance of the two methods are compared for changing number of classes and for images acquired under non-uniform illumination conditions. The results show that the PCA based system is better for the databases of small number of classes and the LDA is better for the databases of large number of classes. Additionally, the effect of the non-uniform illumination conditions on the recognition performance is studied and the superiority of the LDA method over the PCA method has been shown under such conditions.