Fast, Realistic Image Synthesis

快速、逼真的图像合成

• 批准号: 9619853
• 负责人: Paul Heckbert
• 金额: $ 5万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1997
• 资助国家: 美国
• 起止时间: 1997-04-15 至 1998-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9619853&HistoricalAwards=false
• 关键词: Fast; Realistic; Image; Synthesis

The goal of this project is to accelerate the rendering of three dimensional surface models using multiresolution modeling techniques. Multiresolution modeling is the description of shape at multiple levels of detail. In conventional image synthesis, an object is rendered from the same geometric model independent of its screen size, but this is unnecessarily slow for small and distant objects. With multiresolution models, nearby objects are rendered using a detailed model, while distant objects are rendered using a coarse model employing many fewer polygons. To date, the creation of multiresolution models has been largely manual. New algorithms will be developed to adaptively simplify polygonal surface models, attempting to find an accurate approximation to the surface using as few polygons as possible, and to do this as quickly as possible. Methods for smoothing the transitions between levels of detail will be developed. The resulting image should be indistinguishable from a scene rendered from a fully detailed model. Multiresolution models can speed up a number of applications, including flight and driving simulation, computer aided design, visualization, video games, virtual reality, and computer animation. They can allow scenes of great complexity to be rendered in real time.

该项目的目标是使用多分辨率建模技术加速三维表面模型的渲染。多分辨率建模是对形状在多个细节层次上的描述。在传统的图像合成中，对象是从相同的几何模型渲染的，与其屏幕大小无关，但对于小的和远处的对象来说，这是不必要的缓慢。对于多分辨率模型，使用详细模型渲染附近的对象，而使用使用更少的多边形的粗略模型渲染远处的对象。到目前为止，多分辨率模型的创建在很大程度上是手工创建的。将开发新的算法来自适应地简化多边形曲面模型，尝试使用尽可能少的多边形来找到与曲面的精确近似，并尽可能快地做到这一点。将开发用于平滑细节级别之间的过渡的方法。生成的图像应该与从完全详细的模型渲染的场景难以区分。多分辨率模型可以加快许多应用程序的速度，包括飞行和驾驶模拟、计算机辅助设计、可视化、视频游戏、虚拟现实和计算机动画。它们可以实时渲染非常复杂的场景。