Removing Boundary Effect of a Patch-Based Super-Resolution Algorithm

SUMMARY In a patch-based super-resolution algorithm, a low-resolution patch is influenced by surrounding patches due to blurring. Wepropose to remove this boundary effect by subtracting the blur from thesurrounding high-resolution patches, which enables more accurate sparserepresentation. We demonstrate improved performance through experi-mentation. The proposed algorithm can be applied to most of patch-basedsuper-resolution algorithms to achieve additional improvement. key words: super-resolution, dictionary, sparse representation, patch 1. Introduction High-resolution images contain high frequency detail anddemand for them has increased. However, spatial resolutionis limited by an image sensor or an image acquisition sys-tem[1]. Super-resolution (SR) algorithms can overcome thephysical limitation of the imaging system at low-cost.In a SR algorithm based on a sparse representation,an image patch is represented by a sparse linear combina-tion from an overcomplete dictionary. Assuming that a low-resolution(LR)patchandthecorrespondinghigh-resolution(HR) patch have the same sparse representation vector, weestimate a HR patch for each LR patch[2]. Yang