DLS in Graphics

The artform of line drawings is a dramatically efficient mechanism for conveying an idea of shape with minimal visual input. This talk describes a nonphotorealistic rendering system that generates line drawings automatically from 3D models. We introduce the suggestive contour, a new type of line that is drawn at certain types of view-dependent inflections of the surface. Suggestive contours have several provably equivalent definitions, including one that relates them to the local curvature of the surface. We present algorithms for computing suggestive contours efficiently, describe their smoothness and continuity properties, and analyze their stability under changes in geometry and viewpoint. We argue that suggestive contours complement other commonly-used types of lines, such as silhouettes and depth discontinuities, to increase the effectiveness with which shape is conveyed.
Szymon Rusinkiewicz has been an assistant professor in the Computer Science Department at Princeton since September 2001, after obtaining his Ph.D. from Stanford University. His main research interests are in computer graphics and vision, focusing on techniques for measuring shape, light, and reflectance. In addition to data acquisition, he has worked on real-time rendering of large models, shape analysis, and 3D data representation.