Theoretical Cost Comparison of Remote Visualization Strategies

We compare three remote visualization strategies used for interactive exploration of large data sets in distributed environments: image-based rendering, parallel visualization servers, and subsampling. To determine the regimes for which each approach is most cost-effective, we develop performance models to study the computation and communication costs associ-ated with the common visualization task of isosurface generation. For one particular strategy, subsampling, we further investigate the tradeoffs between multiresolution and uniform grid methods in terms of performance and approximation errors