InVis guarantees interactivity by using a combination of various speedup technologies. It invests much in the pipeline preprocessing steps, contains a time control unit and progressive refinement, and works with hybrid data.
Volumedata and geometric objects are prepared in their own structure and rendered at the same time. The raytracing algorithm casts a ray into both data structurs and scans discret volume data and geometric objects at the same time.
Every pipeline step is scalable by a single parameter. A time-control unit changes this parameter interactively and ensures the selected frame rate.
First, a coarse visualization is created. Later, if no user action happens, a refined image is calculated in hierarchical time steps. Therefore an object can be manipulated interactively and reaches high quality as soon as possible.
dagstuhl.ps: InVis-paper in postscript-format (1.4MB, 98-10-29)
dagstuhl.ps.gz: link to InVis-paper (gzip-format) from Joerg Meyer