Surface Visibility for Large Model Visualization

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Many uses of computer visualization rely on displaying surfaces so that a person can easily understand the relationships between the different parts of the visualized object. This project will study new methods to identify the visible parts of an object, and will use this information to produce better images (that is, more accurate and complete images) of scientific data. Some possible applications of this work are medical visualization (e.g. showing the chambers of the heart), studying fracture mechanics (e.g. identifying interior cracks), and computational fluid dynamics (e.g. finding and studying vortices in turbulent flow). Technically, the project has two sub-goals. First, it will automatically identify interior structures of a complex surface, that is, those regions that are not visible from locations outside the object. Currently, this is done manually or by connected component analysis, both of which have limitations. This project will explore both analytical and sampling approaches to quickly computing a visibility function for all surface points. Secondly, the project will automatically produce a group of viewpoints and associated images (an image coverage) that collectively show every visible portion of a given surface. This can be used to both save the user from the error-prone task of manually rotating the object to see all visible portions, and may aid in creating simplified models for rapid navigation and display. This project will explore several approaches, including sampling-based ones, to this previously unsolved problem.