COMPUTATIONAL STEERING OF MOLECULAR DYNAMICS SIMULATION USING PIT

点击翻译按钮获取中文摘要

Over the last few years, we have developed algorithms for automatic simplification of models represented using polygons. Objects in a model may be replaced with automatically-generated "levels of detail" of various complexities, allowing a trade-off of fidelity for speed. During the past year, graduate student Jonathan Cohen continued work on the polyhedral minimizer started by Amitabh Varshney as part of his Ph.D. work. A major paper by Cohen, Varshney, et al. was published in Computer Graphics, the premier publication of the field. Cohen has continued the work to make the simplified surfaces look better by preserving surface normals and other properties. Our main contributions include algorithms for generating and measuring the quality of these levels of detail. This quality is measured in terms of three appearance attributes: surface position, surface curvature, and surface color. Ultimately, the user can control the fidelity/speed trade-off using a single, intuitive, screen-space error tolerance. Our algorithms use local and global error metrics for surface deviation. To preserve the overall appearance, we convert the input surface model to a representation that decouples the sampling of appearance attributes and stores the color and normals in texture and normal maps. The simplification algorithms also employ a texture deviation metric, which guarantees that these maps shift by no more than a user-specified number of pixels on the screen. These algorithms have been successfully applied to a number of models. Significance: The techniques also apply to isocontour surfaces in electron density maps. Plans: Cohen is completing his dissertation in the coming year. He has had another major paper accepted for publication in the July, 1998, issue of Computer Graphics.