A hybrid super resolution technique using adaptive sharpening algorithm based on steering kernel regression for restoration

Abstract

A conceptually simple hybrid Super Resolution (SR) algorithm is proposed using an adaptive edge sharpening algorithm. Most of the existing Super resolution algorithms are not robust to handle the high noisy conditions due to the ambiguity between the sharpening and denoising processes. The Low Resolution (LR) images are applied with the adaptive edge sharpening algorithm that is capable of capturing the local image statistics and adjusts the sharpening process accordingly. The restored LR images are then registered using Scale Invariant Feature Transform (SIFT) based registration to position all LR pixel values to a common spatial grid. The registered LR images are fused using Singular Value Decomposition (SVD) based Fusion algorithm. The experimental results show the efficacy of the developed algorithm, produces better results than the existing algorithms under high noisy conditions.