A Blind Superresolution Method that Works for Arbitrary Motions and Point Spread Functions

Few studies exist in the field of blind image super-resolution (SR). Most of these studies assume that all of the observed low resolution images have gone through a common Point Spread Function (PSF) and the warp operators are pure translations. In this study, changing the order of the warp and blur operators in the observation model, which links the observed low resolution images to the unobserved high resolution image, commonly used in the literature is proposed to eliminate these two non-realistic assumptions. Conditions under which the change in the observation model is valid are derived. Simulation results demonstrate that the existing blind image SR algorithms can be extended to non-common PSFs and general warp operators case under the proposed model change.