Accelerating phase unwrapping based on integer linear programming by processing of subgraphs

Unwrapping is a crucial step when analyzing ground displacements with InSAR techniques. Formulations of unwrapping as integer linear programming (ILP) problem suitable for PSInSAR were given by Costantini et al. under the name redundant integration of finite differences and by Shanker et al. under the name edge list algorithm. They allow the treatment of "4D"-graphs, in the sense of making use of the redundancy both in the PS-net and temporal graph, in a consistent way. This important advantage is paid for by a high computational load, which for large graphs even might prevent the algorithm from finishing in an acceptable time. The approach presented in this paper suggests subdividing the graph into spatio-temporal regions, thereby maintaining the advantage of redundancy, and solving the corresponding ILPs. This results in a considerable acceleration. Besides explaining our approach we report on results obtained for simulated and real data.