Optimized parallelization of non-local means filter for image noise reduction of InSAR image

Non-local means filter is widely used for image noise reduction. Although the sequential implementation for this algorithm has been successfully applied to large scale InSAR problems, few parallel versions of the non-local InSAR filters have addressed well algorithm characteristics such as an unpredictable workload balance, a strong data dependency. These challenges engender many interesting design issues including domain decomposition, data access, management and workload balancing. This paper presents a sophisticated and improved parallel scheme for the iterative non-local Interferometric Synthetic Aperture Radar (InSAR) filters. The proposed algorithm gives a better speedup and high efficiency. Results for numerical simulations and performance measurements were obtained on SuperMUC.