Single image super resolution using local and non-local priors

Abstract

The task of single image super resolution (SR) is to generate a plausible high-resolution (HR) image from a given low-resolution (LR) measurement. This paper presents a reconstruction-based single image SR method. Firstly, an adaptive-shape non-local means (AS-NLM) model is proposed by taking the local structures around pixels into consideration. Afterwards, AS-NLM is utilized to further improve the existing non-local steering kernel regression (NLSKR) model, achieving a new model called I-NLSKR. To obtain superior performance, AS-NLM and I-NLSKR are combined, leading to a new SR algorithm named SRLNP (SR using local and non-local priors). Experimental results demonstrate that SRLNP outperforms many existing methods in both objective and subjective evaluations.