An effective neutrosophic set-based preprocessing method for face recognition

Face recognition (FR) is a challenging task in biometrics due to various illuminations, poses, and possible noises. In this paper, we propose to apply a novel neutrosophic set (NS)-based preprocesssing method to simultaneously remove noise and enhance facial features in original face images. We then employ the Tan and Triggs (TT) discriminant method, which applies kernel fisher linear discriminant analysis (KFDA) on the linear ternary pattern (LTP) features, on the NS-based preprocessed images to further improve FR accuracy. Our experiments on two databases (ORL and FEI) show that the NSbased preprocessing method is more effective than other preprocessing methods to improve FR accuracy of discriminative methods. It can also be integrated with other preprocessing methods to further improve FR accuracy.