Face recognition using curvature Gabor features

This paper introduces a homogeneous Gabor feature based face recognition approach under uncontrolled conditions such as unexpected illumination changes, pose changes, blurring and facial expression changes. The system uses curvature Gabor features instead of conventional Gabor features, and the classifiers are obtained by applying PCLDA to the selected features. By combining some of the obtained classifiers using different fusion methods, good verification accuracies are achieved with low computational complexity. The system is tested on FRGC version 2.0 database, and it achieves 93.11% verification rate.