Users expect high and constant rendering frame rates when they interactively navigate in a Virtual Environment (VE). However, when the VE is too large to fit into the main memory, the frame rates can become unacceptable. In this paper, we combine walkthrough semantics and database techniques, such as indexing, caching and prefetching, to improve the performance of walkthrough of a very large VE. We implemented a prototype walkthrough system called REVIEW (REal-time Virtual Environment Walkthrough) and evaluated its performance on a 1 GB synthetic data-set generated to simulate a large cityscape. Our results show that the proposed techniques are effective for generating constant frame rate and improving the visual effects.
展开▼
