Parallel Volume Rendering based on Isosurface Extraction Using Commodity Clusters

摘要

Volume rendering approaches based on isosurface extraction enable the extraction of topological information and thereby more sophisticated post processing steps than their counterparts based on projective methods. However, they are burdened by performance problems stemming from the huge triangle meshes which are created during this process. Mesh optimization algorithms help to decrease the number of triangles and thereby ease the handling of the final surfaces, but also further contribute to the overall application execution time. A suitable way to overcome these problems is the deployment of parallel processing. This work presents such a parallel surface extraction code implemented on top of commodity clusters, which provide an excellent price/performance ratio. This allows a drastic reduction of the overall execution time bringing it closer to a range suitable for interactive visualization.