Regularized Image Restoration

Abstract

The scope of this work focusses on non-blind image restoration where the point spread function (PSF) of the blur convolutional kernel is known. Blind deconvolution is, by its nature, a more challenging problem, Haykin (1994); Kundur & Hatzinakos (1996). However with effective and efficient PSF estimation techniques, Fergus et al. (2006); Joshi et al. (2008); Krahmer et al. (2006); Nayar & Ben-Ezra (2004); Oliveira et al. (2007), the research trend has been to handling blind deconvolution in two steps, with PSF estimation as the first step and image estimation as the second step, Levin et al. (2009). This motivates us to focus on efficient algorithms for image restoration where the blur convolutional kernel is known.