Human Face Recognition Using Combined Approaches PCA and ICA

For a human vision system not much effort is required to recognize a face from a given image but it is a challenging task to make a computer system understand it. In this paper we propose to design and implement a system to recognize a human face using a hybrid technique with holistic-based face recognition (PCA & ICA) approach. For a face to be recognized we use whole information of a face patch and perform some transformations on this patch to get a compact representation for face recognition. The transformations are usually obtained from statistics.  Essential information may be contained in the higher order relationships between all pixels. To have a better basis images a method sensitive to these high-order statistics for which ICA is used Experiments are proposed to be carried out using Indian, FERET and ORL face database. Test cases are considered for performance evaluation and its experiments were conducted with same face given in training, slightly variation in face and more variation in facial image. Performances of both ICA and PCA have a lesser accuracy though ICA is precise enough to consider a solution.