Research On Face Recognition Based On LBP Features

Face Recognition is one of the most important research topics in the field of pattern recognition, which has very important theoretical research value and a broad range of potential applications. In recent years, researchers in the field of face recognition have achieved significant progress. A large number of face recognition algorithms have been proposed. On the basis of these algorithms, a number of commercial face recognition systems have been developed.Local Binary Pattern (LBP) is a kind of texture description methods. In recent years, researchers have successfully applied LBP features for face recognition. LGBP have been developed on the basis of original LBP descriptor. It combines multi-scale and multi-orientation Gabor decomposition with LBP descriptor, and therefore its representation power is further enhanced. Although LGBP algorithm has produced good performances in face recognition, it still has some drawbacks, such as very high feature dimension, and lack of robustness in complex lighting conditions. This paper mainly addresses these problems and the research contents are as follows:1. The distribution of LGBP features is analyzed. A face recognition algorithm by Zero-Ratio based LGBP features is proposed based on it. The method utilizes Zero-Ratio conception to select effective features from LGBP features. The method can significantly compress LGBP features without significantly affecting the recognition accuracy.2. A face recognition algorithm is proposed based on the Local Normalization and LGBP operator. Before using LGBP features to describe the face image, Local Normalization technique which is robust to illumination variant is utilized to preprocess the original face image. The method can effectively improve the robustness of LGBP operator to illumination variant.