Multi-resolution visualization of time-dependent three-dimensional data*

Due to the improvements in performance in computer power and storage capacity achieved over the last decade, today's data-intensive scientific applications and simulations are capable of generating massive amounts of data. Sensor networks will soon consist of thousands of (possibly moving) sensors, distributed in a three- dimensional (3D) environment and recording multiple parameters. Standard visualization techniques are not capable to render the huge data sets at interactive frame rates. "Multi-resolution methods" provide a means for representing data at multiple levels of detail. In general, interactive data exploration and visualization can be performed better for "structured rectilinear grids," i.e., grids where space is represented by a collection of the same type of bricks. Different types of grids cannot be used straightforward for real-time data visualization purposes.
We have developed two multi-resolution methods for structured grids. The first approach is based on octree refinement and uses a special storage scheme for fast data loading from external storage media. A novel hierarchical 3D storage scheme ensures that data points that are close to each other in 3D space are also stored close to each other on disk. The second approach is based on a new "subdivision scheme." This scheme starts from a coarse representation of 3D space, using cubes, and then refines the representation. In each subdivision step, the total number of points is only doubled. We can take advantage of special filter schemes to avoid aliasing in our visualizations and obtain higher-quality visualizations at coarser levels of resolution. For dealing with data varying over time, we have generalized this approach to 4D data. Our approach provides scalability in spatial and temporal dimensions. We have applied our hierarchical data
representation scheme for visualization, and we have tested it for standard methods including iso surface visualization, volume rendering, and cutting planes.
*This project is not officially supported through CITRIS funds, but the faculty and topical affiliations are sufficiently strong that it is listed here for referral and convenience.