Dual-Stage Phase Unwrapping

Energy minimization on large scale problems can be expensive to compute. This paper presents an extension of a primal-dual algorithm to scale to large scale convex energy minimization problems, while guaranteeing the global optimum on convergence. This is done through the use of a so-called “Dual-Stage” approach that exposes a natural level of parallelism. Benchmarks on the Phase Unwrapping problem in InSAR show that using this approach can give significant improvements to computation-time and memory-time used.