Abstract

We proposed a deformable patches based method for sin-gle image super-resolution. By the concept of deforma-tion, a patch is not regarded as a fixed vector but a flexi-ble deformation flow. Via deformable patches, the dictio-nary can cover more patterns that do not appear, thus be-coming more expressive. We present the energy functionwith slow, smooth and flexible prior for deformation model.During example-based super-resolution, we develop the de-formation similarity based on the minimized energy func-tion for basic patch matching. For robustness, we utilizemultiple deformed patches combination for the final recon-struction. Experiments evaluate the deformation effectiveness and super-resolution performance, showing that the deformable patches help improve the representation accuracy and perform better than the state-of-art methods.