UCSC-SOE-08-20: Direct Volume Rendering of Multi-Valued Data Sets

Newton Der and Alex Pang
06/30/2006 09:00 AM
Computer Science
This paper deals with methods for direct volume rendering of 3D multi-valued scalar data sets. A multi-valued scalar data is one where there are multiple scalar values at each location. Hence, a 3D multi-valued scalar data contains multiple scalar values at each voxel. While there are a few techniques that deal with direct volume rendering of vector data (velocity components at each voxel), diffusion tensor data (eigenvalues at each voxel), multi-field data (multivariate vector at each voxel), or scalar uncertainty data (data plus scalar uncertainty at each voxel), the multi-values are about a single variable. As such, the emphasis of the visualization is not so much about the relationships among the different instances of the same variable, but rather about their collective pattern and behavior. We discuss two approaches for direct volume rendering of multi-valued scalars: (a) convert to scalar then volume render, or (b) volume render then convert to scalar. The latter offers a wide array of interesting possibilities which are described in the paper.