Interactive Direct Volume Rendering of Curvilinear and Unstructured Data

Abstract: "Methods for interactive volume rendering of nonrectilinear 3D scientific data sets, such as those generated by the finite elementmethod, are investigated. We focus on the use of projection methods, in particular splatting algorithms, for volume rendering curvilinear and irregular data. The data is rendered without interpolating it to a rectilinear mesh. The goal is interactive performance even when the data sets are very large. To achieve this, we investigate parallelization, graphics hardware support, a suite of splatting approximations, and mesh filtration. Using these techniques, we have generated images of nonrectilinear data sets with over 1,000,000 cells interactively(in less than 15-30 seconds). Using filtering methods this performance is possible for even larger data sets.