Hierarchical splatting of scattered 4d data

Visualizing large uncorrelated point data sets with regular volume rendering produces unsatisfying re-sults due to low spatial resolution. We present a system that can deal with time-varying uncorrelated data at high resolution and interactive rates. Figure 1: Molecular dynamics simulation with one million particles per time step. Quite a number of physical simulations create large point-based data sets, for example molecular dy-namics (Fig. 1), n-body simulations in astrophysics (Fig. 2), and smoothed particle hydrodynamics (SPH) (Fig. 17). Other sources of scattered point data are astronomical observations where new tech-niques for measuring three dimensional positions of stars as in the GAIA project will create huge real-world data sets in the near future as well. These data sets contain up to hundreds of millions of points, each with information about position xi, diameter si, and intensity ci at various wavelengths. We cannot visualize these data sets interactively by simply ren-dering millions of colored points for each frame. Therefore, in many visualization applications a scalar density corresponding to the point distribu-tion is resampled on a regular grid for direct volume rendering. However, this technique imposes a low-pass filter on the data set, and many fine details ar