Mean-field annealing for phase unwrapping

Use of mean-field annealing theory is proposed for solving the phase-unwrapping (PU) problem. PU is formulated as a constrained optimization problem for the field of integer corrections to be added to the wrapped gradient field. A deterministic algorithm is described to provide an approximation of the average of the correction field over the global minima of the cost function. The proposed algorithm can be applied for any choice of the cost function. Using a cost function based on second-order differences, we obtain results close to those from simulated annealing and spend less computational time.