Institute for Data Analysis and Visualization

We present an octree-based approach supporting multiresolution volume rendering of large data sets. Given a set of scattered points without connectivity information, we impose an actree data structure of low resolution in the preprocessing step. The construction of this initial octree structure of low resolution in the preprocessing step. The construction of this initial octree structure is controlled by the original data resolution and cell-specific error values. Using the octree nodes, rather than the data points, as elementary units for ray casting, we first generate a crude rendering of a given data set. Keeping the pre-processing step independent from the rendering step, we allow a user to interactively explore a large data set by speicifying a region of interest (ROI), where a higher level of rendering accuracy is desired. To refine an ROI, we are making use of the octree constructed in the pro-processing step. Our approach is aimed at minimizing the number of computations and can be applied to large-scale data exploration tasks.