A Novel Image Super-Resolution

Super-resolution (SR) image reconstruction is the process ofcombining several low resolution images into a single higherresolution image or Generating high resolution image from a single low resolution image. There is a driving need for digital images ofhigher resolutions and quality.In order to improvethe resolution and perceptual quality of such web image/video, webring forward a sensible solution which mixes adaptive regularization and learning-based super-resolution. The contributionof this work is twofold. First, we tend to propose to research the imageenergy modification characteristics during the iterative regularizationprocess, i.e., the energy change magnitude relation between primitive (e.g., edges, ridges and corners) and non-primitive fields supported therevealed convergence property of the energy modification magnitude relation, acceptable regularization strength will then be determined to wellbalance compression artifacts removal and primitive elementspreservation. Second, we tend to verify that this adaptive regularizationcan steady and greatly improve the try matching accuracy inlearning-based super-resolution.