Integrated Denoising and Unwrapping of InSAR Phase Based on Markov Random Fields

In the traditional processing flow of interferometric synthetic aperture radar (SAR) technique, the processing of phase is conducted via two separated and successive steps, i.e., phase denoising and phase unwrapping. That is to say, first, wrapped phases without noise are generated, and then, the true phases without 2π-ambiguities are reconstructed (here and in the rest of this paper, true phase refers to the information-induced unwrapped phase without noise). Such separated steps will inevitably bring in extra estimation error because each step has necessary approximations and presumptions which do not always hold. On the contrary, in this paper, we treat phase denoising and unwrapping as a single problem of true phase recovery from observed ones. Following this methodology, an integrated phase denoising and unwrapping algorithm based upon Markov random fields (MRFs) is proposed. Taking a priori knowledge of interferometric phases into account, MRF is used to model the relationship between the elements in the random variable set including both true phases and their observations. After the model is built up, the energy function of this MRF is defined according to the local-independence property inferred from the MRF structure and then minimized to obtain the estimate of the true phase value. In the end of this paper, experiments on simulated and true phase data are conducted, and the comparison with several commonly used unwrapping methods is proposed to verify the efficiency of the proposed MRF algorithm.