A Bayesian/maximum-entropy view to the spatial estimation problem