Review on data uncertainty in face recognition using appearance-based methods

The face images should not be the completely accurate for representation and for an observation. To reducing the uncertainty for representation of the face images and to improving the accuracy of face recognition, more observation of the same person face images is required in the face recognition. In the real world face recognition system the uncertainty highly occurred because the limited number of available face images of subject and due to this there is high uncertainty is occurred. In this paper, develop the model which is to improve the accuracy in the face recognition by reducing the data uncertainty. The model is to reduce the uncertainty of face images representation by synthesizing the virtual training samples. Here, the useful training samples are selected, which are comparable to the test sample from the set of all the original training samples and synthesized virtual training sample.