An Exploration Through Various Optimization Methods for InSAR Images

Energy Minimization is a framework where the optimum assignment of values to the nodes of a graph is required. Assignment of values to the nodes of the graph results in a unary cost (for each single node) and higher order costs which is related to the relation between several nodes. Phase Unwrapping is one of those problems which can also be formulated in Energy Minimization framework. This problem only includes unary and pair-wise (second-order) terms. Phase Unwrapping is the problem of finding the continuous phase out of the wrapped phase which is a really important problem in InSAR images processing pipeline.

OpenGM (Open Graphical Model) is the result of a study investigating several inference methods for Energy Minimization and for different applications. It investigates MAP (Maximum A Posteriori) estimation from a Bayestian viewpoint. It includes the most recent updates of inference methods and uses specialized implementation for some of them. We have investigated different inference methods in OpenGM for phase unwrapping problem for InSAR images and we fine-tuned the compatible methods for this problem.