Non-stationary t-distribution prior for image source separation from blurred observations

We propose a non-stationary spatial image model for blind image separation problem. Our model is defined on first order image differentials. We model the image differentials using t-distribution with space varying scale parameters. This prior image model has been used in the Bayesian formulation and the image source are estimated using a Langevin sampler method. We have tested the proposed model on astrophysical image mixtures and obtained better results regarding to stationary model.