Multi-view Face Recognition based on Low-rank Features and Sparse Representation Residual Ratio Comparison

Abstract

Concerning the problem that the existing face recognition algorithm fails to exhibit good robustness when dealing with multi-view mixed face data, a multi-view face recognition method based on low-rank features and auxiliary dictionary is proposed. The method is designed to select low-rank decomposition models with two different regular terms and complete the feature extraction process respectively, and reduce the difference of same people faces in different views. The obtained low-rank feature can enhance the discriminating information of different types of faces in the same view, and remove the influence of common parts on the recognition effect. Then, using the external data to learn auxiliary dictionary which can simulate view variable and a residual ratio comparison model based on the minimum residual and the secondary minimum residual is designed. Finally, the final classification results are determined based on the results of auxiliary dictionary learning and residual ratio comparison. Experiments on the multi-view face database CMU-PIE with or without auxiliary dictionary show that the improved method can obtain low-rank face discrimination information and eliminate the gesture interference effectively. The proposed face recognition algorithm has more efficient recognition rate and robustness in the experimental environment of multi-view mixed data.