Parallel Isosurface Extraction including Polygon Simplification via Self Adapting Vertex Clustering

S. Manten, I. Breuer, and S. Olbrich (Germany)


volume visualization, marching cubes, polygon simplification, vertex clustering


The compute power of modern supercomputers allows numerical simulations in high-resolution on very large data grids. The parallel, network-distributed process chain of the Distributed Simulation and Virtual Reality Environment (DSVR) can handle the visualization of the resulting huge amounts of raw data of those simulations by doing the extraction of 3D data (mapping) as a preprocessing step at the source (“in situ” visualization) and avoids the storage and communication bottlenecks of traditional post-processing approaches. A new parallel isosurface extraction approach has been developed as part of this framework for the purpose of volume visualization. The described method combines the marching cubes algorithm with a self adapting vertex clustering based polygon simplification, in which the vertex cluster size is automatically enlarged in flat, nearly planar regions of the extracted isosurface. The new method can generate isosurfaces of higher quality compared to the equivalent approach with uniform vertex clustering at the same level of reduction. An analysis of the speed up shows that the parallel isosurface extraction scales well up to 200 cores.

Important Links:

Go Back