An Adaptive Parameter Estimation in a BTV Regularized Image Super-Resolution Reconstruction

Access to the fine spatial resolution has always been a hotspot in digital imaging. One way to improve resolution is to use signal post-processing techniques. In this study, an improved multi-frame image super-resolution (SR) algorithm is proposed. The objective function should be minimized consists of a data error term, a regularization term and a regularization parameter. Based on the bilateral-total-variation (BTV) regularization, in the proposed method on one hand, the data error term incorporates frames with high accuracies in the reconstruction process, where an indicator weights each frame proportional to the frame error. On the other hand the regularization parameter is updated in each iteration based upon the Morozov's discrepancy principle. Iterative adjustment of the regularization parameter guarantees the SR solution to satisfy discrepancy principle. Visual evaluation and also quantitative measurements show that the performance of the proposed algorithm is better than of the several state-of-the-art methods.