Super resolution through alternative optimization using sparsity and PSF prior

Existing sparse representation model uses image statistics in the form of neighborhood correlation, learning algorithm for use of redundant dictionary, etc. The ill-posed nature of the problem means that there is no exact solution so any solution is an approximate of the actual solution and this often leads to discrepancy in the form of degradation as global smoothing of the final high resolution image. In our paper we propose overcome this drawback by using point spread function (PSF) or blur prior which will remove the degradations to give us an final super enhanced high resolution image. The PSF prior is integrated in to the SRM thereby preserving the computational complexity. The experimental results using the proposed method is compared with the existing state of the art methods for performance comparison.