Face Recognition Fusion Multi-feature And Local Binary Pattern

In recent years, with the development of biometric technology, face recognition has been widely used in finance, education, security and other related areas, for its friendly, easily collection and non-invasive. There are great of face recognition designed by commercial companies have been commercialized. However, there are still many problems to be solved. such as the complexity of face structure and the effect of occlusion, expression, light and other factors, face recognition.Face recognition system consists of face detection, feature extraction, classification and identification three parts, Facial feature is good or bad, it will affect the performance of face recognition. Local Binary Pattern(LBP) is an effective texture description method, But the extracted features too simple, do not adequately describe the facial features. In this paper, monogenic filtering and image gradient information and local binary patterns are combined into a new feature, used to extract richer face information. The main work is as follows:(1) a new method of face recognition based on monogenic features and CS-LBP is proposed, Firstly, center-symmetric local binary pattern(CS-LBP)is adopted to encode the monogenic magnitude, The monogenic phase is quantified into four regions and the horizontal direction and vertical direction is encoded, Combination of these three parts to extract image feature, And get facial feature by block histogram, Finally the recognition is performed by using the nearest neighbor classifier. Experimental results show that the algorithm is an outstanding method under different illumination condition, different facial expression condition and partial occlusion condition.(2)a novel method of face recognition based on bidirectional gradient center-symmetric local binary pattern(BGCSBP)is proposed. Firstly horizontal gradient and vertical gradient of face image are calculated, and center-symmetric local binary pattern(CS-LBP)is proposed to encode the gradient. Secondly the proposed BGCSBP is the combination of the CS-LBP of horizontal gradient and vertical gradient. BGCSBP feature maps are divided into several blocks and the concatenated histogram features calculated over all blocks is used for the feature descriptor of face recognition, and the recognition is performed by using the histogram cross. Experimental results show that the algorithm is an outstanding method.