Modeling of and navigation in complex 3D documents

Abstract During the past years, increasing computing power has opened the door to 3D scene complexities that were unimaginable a decade ago. But this quantitative leap reveals a new class of problems: software architectures for small- or mid-scale models typically do not scale and consequently fail to be appropriate for large-scale modeling and visualization. While there is significant progress in visualizing large models locally using libraries like OpenGL Optimizer or the Fahrenheit project, these libraries are still relatively low-level in terms of their geometric primitives. Our belief is that a true leap in scene complexity can only be achieved with much more compact and abstract model descriptions. Visualizing models at a detail resulting in triangle numbers that are—when accurately tesselated—four or five orders of magnitude greater than those that can be visualised interactively with current high-end graphics hardware, demands completely new approaches for the handling of complex scenes, thereby, opening the research fields of language-based modeling and efficient model evaluation.