An image fusion technique for efficient face recognition

An image level fusion technique combines different form of images so that the combine image may contain more relevant information than the individual ones. An image fusion technique for face recognition is presented in this paper. Due to the topological structure of human face images, one can easily identify a face image by looking along the horizontal-vertical and also forward-backward directions. This indicates that there is enough relevant information available along those directions. To realize this, we generate the diagonal face image from the original face image and then combine them to get the fused image. Finally, we extract discriminant features from the fused images, which integrate the underlying discriminant information along those directions. This extracted feature vectors are applied on Radial Basis Function Neural Networks (RBF-NN) for classification and recognition. Experiments on the AT&T and UMIST face databases show that the proposed method can obtain a higher accuracy than some of the subspace-based methods.