Towards hybrid 2D phase unwrapping using fuzzy clustering and neuro-fuzzy learning for SAR images: a case study on IFSAR phase image

In acknowledging that every phase unwrapping (PU) technique has its advantages and disadvantages, a hybrid PU scheme is proposed. There are three phases in this scheme. In phase I, phase images are segmented based on fuzzy clustering, in phase II, PU is performed on every segment and its results are evaluated, and in phase III, neurofuzzy training is used to map optimal PU technique to every cluster. In short, this method eliminates weakness of a PU technique while preserving its advantages. At the end of the whole process, a mapping knowledge that is able to tell which PU technique is best fit for which cluster, is produced. More significantly, issues and problems during the integration of validation of a number of clusters, spatial continuity incorporation, and clustering methods are justified and solved. An initial experimental result is presented. This result showed a significant improvement over single PU algorithm processing