| Volume visualization is characterized by the manipulation of large quantities of data; algorithms for volume visualization can be improved by careful data structure design that economizes on space and access time by exploiting surface and volume coherency.;New designs for compact quantization of normal vectors are derived and evaluated. These normal vector codes allow fast table based shading, and can be adapted to arbitrary density distributions. This adaptability allows codes that can adapt either to a restricted class of volume data for increased compression, or to perceptually optimized distributions.;The mathematical properties and metaphorical qualities of volume data and volume visualization rendering techniques are examined. Compact data structure designs for surface, binary volume, and scalar volumes are reviewed, and new designs based on image compression techniques are proposed and evaluated. In particular, the Laplacian octree shows great promise as a structure that can both highly compress a volume while improving the speed and robustness of ray cast volume rendering. |
