ABSTRACT
Fluid flow simulations play an important role in many scientific disciplines, for example, in the modeling of thermal hydraulics in a nuclear reactor. To visualize vector data generated from these simulations, one popular method is to display flow lines, such as streamlines or pathlines, computed from numerical integration. The primary challenge of displaying flow lines, however, is to place the particle seeds in appropriate positions such that the resulting flow lines can capture important flow features without cluttering the display. Setting parameters for flow visualization algorithms so that no important features are
(a) (b) (c) (d)
(e) (f) (g) (h)
missed is non-trivial. Furthermore, it is difficult to evaluate the quality of the visualization output. For large-scale datasets, scientists will not be able to set visualization parameters through trial-and-error because the cost of recomputing visualizations will be too expensive. To generate a visualization that can capture a maximum amount of information from the data with a minimal amount of visual clutter and user intervention, it is crucial to have automatic visual analysis algorithms combined with quantitative quality measures. In this chapter, we introduce several information-theoretic measures and their related algorithms to tackle this problem.
