MAP Estimators for Self-Similar Sparse Stochastic Models

We consider the reconstruction of multi-dimensional signals from noisy samples. The problem is formulated within the framework of the theory of continuous-domain sparse stochastic processes. In particular, we study the fractional Laplacian as the whitening operator specifying the correlation structure of the model. We then derive a class of MAP estimators where the priors are confined to the family of infinitely divisible distributions. Finally, we provide simulations where the derived estimators are compared against total-variation (TV) denoising.