MAP-MRF filtering of SAR interferometric phase fields

In this paper we present a new technique, based on a MAP approach, for filtering interferometric phase images. To describe prior knowledge about the unknown image, a MRF/Gibbs model is adopted, and its energy function is defined so as to impose two desirable constraints on the solution: the absence of residues (which leads to unambiguous phase unwrapping) and the smoothness of the phase field (which reduces noise). MAP estimation is carried out by simulated annealing, obtaining accurate results even for low values of the coherence.