Locally adaptive regularization for robust multiframe super resolution reconstruction

Abstract

Super resolution reconstruction (SRR) is a post processing technique to correct the degradation of the acquired images due to warping, blur, downsampling and noise. In this paper, image is modeled as Markov random field (MRF) and we propose fuzzy logic filter based on gradient potential (FL) to distinguish between edge and noisy pixels. Based on pixel classification, Tikhonov regularization (TR) or bilateral total variation (BTV) is adopted as a prior in maximum a posteriori (MAP) estimation. Such priors are imperative to obtain a stable solution. Tukey's biweight norm (TBN) is adopted for removing the outliers. The proposed approach is demonstrated on standard test images. Experimental results indicate that the proposed approach performs quite well in terms of visual evaluation and quantitative measurements. © 2012 Springer-Verlag GmbH.