The proposed method for phase unwrapping is based on a global analysis of t
he interferometrical phase. The underlying principle is that the interferog
ram is partitioned such that the unwrapped-phase function on each element c
an be locally modelled by the mean values of the phase difference between n
eighbouring pixels in azimuth and range directions. Using this local inform
ation and a least-squares algorithm (Gauss-Seidel relaxation), an approxima
te model of the unwrapped phase is then generated and tested by calculating
the "residue image" defined as the difference between the original interfe
rogram and the model itself. If there are residual fringes, then the result
must be iteratively refined applying the method to the residue image. The
accuracy of the proposed estimation depends on the dimensions of the elemen
ts and the dynamic content of the phase, i.e. on the "roughness" of the gro
und surface.