Multi-views Face Recognition System
Xu Liu, Fan Yang, Michel Paindavoine, Jisheng Dang · 2007
This paper presents some development results on a new multi-views face recognition system. Our objective is to study an efficient method in order to improve recognition performance and achieve more secure information. To do so, we built a simple and low-cost acquisition system composed of five standard cameras, which together can simultaneously take five views of a human face from different angles. Then, a multi-views face database of 3600 images was constructed. We also performed a preliminary study on multi-views face recognition based on two statistical methods: PCA and ICA. Experimental results show that an average improvement of 5.5% for face recognition performance has been obtained using our multi-views system.