A new two-step face hallucination through block of coefficients

We introduce a two-step face hallucination frame work as one of classifying among sparse residual compensation model. In the first step, the optimal coefficients of the interpolated training images are used to construct a global face image. In the second step, a class of priors is computed based on mixing a set of linear priors related to dissimilar priors. The blocks of coefficients are considered to find the sparse mixing weights. In order to find the best improved information of the face image in the residual compensation of step-two, a sparse signal representation is considered over coefficients in a frame. Finally, we obtain a hallucinated face image by integrating these two steps. The extensive experiments on publicly available database show the effectiveness of the framework.