Super-resolution via recapture and Bayesian effect modeling

This paper presents Bayesian edge inference (BEI), a single-frame super-resolution method explicitly grounded in Bayesian inference that addresses issues common to exist- ing methods. Though the best give excellent results at mod- est magnification factors, they suffer from gradient stepping and boundary coherence problems by factors of 4x. Cen- tral to BEI is a causal framework that allows image cap- ture and recapture to be modeled differently, a principled way of undoing downsampling blur, and a technique for incorporating Markov random field potentials arbitrarily into Bayesian networks. Besides addressing gradient and boundary issues, BEI is shown to be competitive with exist- ing methods on published correctness measures. The model and framework are shown to generalize to other reconstruc- tion tasks by demonstrating BEI's effectiveness at CCD de- mosaicing and inpainting with only trivial changes.