Very Large Scale Visualization Methods for Astrophysical Data

Abstract

We address the problem of interacting with scenes that contain a very large range of scales. Computer graphics environments normally deal with only a limited range of orders of magnitude before numerical error and other anomalies begin to be apparent, and the effects vary widely from environment to environment. Applications such as astrophysics, where a single scene could in principle contain visible objects from the subatomic scale to the intergalactic scale, provide a good proving ground for the multiple scale problem. In this context, we examine methods for interacting continuously with simultaneously active astronomical data sets ranging over 40 or more orders of magnitude. Our approach relies on utilizing a single scale of order 1.0 for the definition of all data sets. Where a single object, like a planet or a galaxy, may require moving in neighborhoods of vastly different scales, we employ multiple scale representations for the single object; normally, these are sparse in all but a few neighborhoods. By keying the changes of scale to the pixel size, we can restrict all data set scaling to roughly four orders of magnitude. Navigation problems are solved by designing constraint spaces that adjust properly to the large scale changes, keeping navigation sensitivity at a relatively constant speed in the user s screen space.