Marching Correctors – Fast and Precise Polygonal Isosurfaces of SPH Data

This paper presents the first method for isosurface extraction from smoothed particle hydrodynamics (SPH) data that is exact with respect to the functional representation provided by SPH. The Marching Correctors algorithm is an extension of the Marching Cubes algorithm which is adapted to the SPH representation and avoids resampling to a full grid. The algorithm operates on a virtual grid of sufficiently high resolution to faithfully reconstruct the fields represented by the SPH data. The virtual grid is efficient in terms of both memory usage and computing time, because cells are only materialized and processed if they are either seed cells or intersected by the isosurface. Besides the virtual grid, a key idea of our algorithm is to add a correction step to the isosurface vertices. An evaluation of the algorithm in terms of accuracy and performance is given based on three SPH datasets. By comparing with [1] on similarly sized data a performance gain of almost two orders of magnitude was achieved. Moreover, it is demonstrated how the correction step effectively reduces the typical artifacts produced by the Marching Cubes method.