International Journal of Computer Applications
Expression variations pose a serious problem to automatic face recognition. In this paper a robust expression invariant face recognition algorithm is proposed using a single image for training. Discrete Wavelet Transform (DWT) using Coif-lets as the basis function is applied to the face images. Then SIFT descriptors of the combination of approximation and detail sub-bands are computed for each of the face images. Matching is done using angle distance similarity measure. The algorithm has been tested on the Essex grimace database and only a single image has been used for training.