A New Alternative Blind Super-Resolution Algorithm Based on Motion Estimation

In this paper, we propose a new alternative blind SR to estimate motion, blurring, and higher resolution (HR) image alternatively and iteratively in case the three are all unknown. The related ternary optimization is first divided into two alternate and iterative optimizations, in which the unitary one is about motion estimation and the dualistic one is about blurring and HR image estimation, and the dualistic one is divided into two alternate and iterative unitary optimizations again. Then in the related minimization functional, Total Variant regularization is used to conserve local discontinuous of the estimation, Half-quadratic regularization is used to solve the thus induced non-quadratic problem, and additive operator split is used to simplify the related computation. Test results at last show the validity of the proposed SR.