A class of global optimization problems as models of the phase unwrapping problem (Q1850828)

Let \(U\) be a discrete real function defined on \(N_1 \times N_2\) grid nodes and let \(W = [U]_{2 \pi}\), i.e., the modules \(2 \pi\) of \(U\). The phase unwrapping problem recovers \(U\) from some knowledge of \(W\). There are infinitely many functions \(U\) corresponding to the same function \(W\). A desirable \(U\) is obtained by minimizing certain merit functions. When a merit function is defined by \(l_p\)-norm with \(p > 1\), the problem becomes a nonlinear minimum cost flow problem. Algorithms for solving it are proposed and numerical results are presented. The phase unwrapping problem is the key problem in the SAR (Synthetic Aperture Radar) interferometry problem.