A multiframe super-resolution technique based on a nonlocal Bregman distance of bilateral total variation term

In this paper, we propose a variational multiframe super-resolution (SR) model based on a nonlocal regularization term using Bregman distances. Since the SR algorithms always skip the complex spatial interactions within images, we introduce a nonlocal form of the bilateral total variation (BTV) regularization term which can take into consideration these interactions and also efficiently preserve strong edges and contours of the reconstructed high resolution (HR) image for our model. In addition, to avoid contrast loss and smoothing gray values in the SR process, we introduce Bregman distances which produce a more consistent model. Moreover, to resolve the obtained SR algorithm, we use a first-order primal dual algorithm to ensure that the convergence to the desired HR image can be achieved in a fast way. As a result, the proposed algorithm shows improved performance compared to other SR methods.