Face recognition by Zero-Ratio based LGBP features

LGBP has been successfully applied in face recognition, which combines Gabor decomposition and LBP descriptor. However, in this method LBP histograms from multi-scale and multi-orientation Gabor energy images are concatenated to form the feature vector for input image, so the dimension of generated features is very high. In this paper we propose a novel face recognition method called Zero-Ratio based LGBP, which uses a feature selection method called Zero-Ratio to reduce the dimension of LGBP features effectively. We evaluate our method on AR face database, the experimental results show that our method can greatly reduce dimension of LGBP features without significantly affecting recognition accuracy.