Image Super Resolution Using Expansion Move Algorithm

In multi-frame image super resolution (SR), graph-cut is an effective algorithm to minimize the energy function for SR. As a kind of graph-cut algorithm, aexpansion move algorithm can effectively minimize energy functions such as class F-2. However, the energy functions for SR established in Markov random field usually don't fall into this class and need some approximations, which may lead to poor results. In this paper, we propose a new method, with which we make the energy function for SR a form of class F-2 without approximation. Experimental results show that our motivation is valid and the proposed method is effective for not only synthetic low-resolution images but also real images.