Robust face recognition using fusion of multiscale experts
Imran Naseem, Muhammad Affan Alim · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2012
This paper proposes an efficient and robust technique for face recognition. The proposed technique includes the Daubechie's wavelet transform D10, Principal Component Analysis (PCA) and Multiscale fusion for face recognition. Features are extracted using the PCA on original and multiscale images. The multiscale fusion is used to combine the results of PCA and wavelet transformed PCA to achieve better performance. The main idea is to utilize the discriminant information of various subbands rather than relying on a single scale. Multiscale experts are finally fused using the sum rule. Extensive experimental results on the AT&T database show that recognition performance is improved by the proposed method.