Coupling strategies for multi-resolution deformable meshes: expanding the pyramid approach beyond its one-way nature

With higher resolutions, medical image processing operations like segmentation take more time to calculate per step. The pyramid technique is a common approach to solving this problem. Starting with a low resolution, a stepwise refinement is applied until the original resolution is reached.

Our work proposes a method for deformable model segmentation that generally utilizes the common pyramid technique with our improvement, to calculate and keep synchronized all mesh resolution levels in parallel. The models are coupled to propagate their changes. It presents coupling techniques and shows approaches for synchronization. The interaction with the models is realized using springs and volcanoes, and it is evaluated for the semantics of the operation to share them across the different levels.

The locking overhead has been evaluated for different synchronization techniques with meshes of individual resolutions. The partial update strategy has been found to have the least locking overhead.

Running multiple models with individual resolutions in parallel is feasible. The synchronization approach has to be chosen carefully, so that an interactive modification of the segmentation remains possible. The proposed technique is aimed at making medical image segmentation more usable while delivering high performance.