Region-based weighted-norm approach to video super-resolution with adaptive regularization

We propose a super-resolution (SR) algorithm that takes into account inaccurate estimates of the registration parameters. When frames obey the assumed global motion model, these inaccurate estimates, along with the additive Gaussian noise in the low-resolution image sequence, result in different noise level for each frame. However, in case of existence of local motion and/or occlusion, regions that have local motion and/or occlusion have different noise level. To cope with this problem, we propose to adaptively weight each region according to its reliability and the regularization parameter is simultaneously estimated for each region. The regions are generated by segmenting the reference frame using watershed segmentation. The experimental results using real video sequences show the effectiveness of the proposed algorithm compared to three state-of-the-art SR algorithms.