Phase unwrapping using Markov Random Field model applied to segmented regions

This paper deals with a new phase unwrapping method based on the Markovian Random Field model. We estimate the 2π cycles of each homogeneous region in the interferogram. The first step in this approach is the segmentation of the phase image into homogeneous regions after locating the edges of the fringes using the Structural Feature Set (SFS) algorithm. Then, the phase image is modeled using MRF model where a potential function is defining. The true phase is computed by estimating the appropriate 2π cycle for each region. This number of cycles is estimated by minimising an energy function between neighboring regions. In this paper, we used the conjugate gradient algorithm to calculate this minimum. The proposed unwrapping approach is tested and validated on simulated and real interferograms provided from Envisat-2 satellite pair images captured over the region of Mahdia in Tunisia.