Single-frame image super-resolution using a Pearson type VII MRF

Image super-resolution restoration aims to recover a high resolution scene from its low resolution measurements. It is a difficult, ill-posed problem, with no consensus as to how best to formulate image models that can both impose smoothness and preserve the edges in the image. Here we develop a new image prior based on the Pearson type VII density integrated with a Markov Random Field model. This has desirable robustness properties and achieves state-of-the-art performance in terms of the mean square error, in a range of noise conditions. We develop a fully automated hyperparameter estimation procedure for this approach, which makes it advantageous in comparison with alternatives.