SurfLab, CISE’s computer graphics research group, received the best paper award (1st prize) of the 2014 Symposium on Solid and Physical Modeling (formerly ACM Solid Modelling) in Hong Kong for the paper entitled “Correct resolution rendering of trimmed spline surfaces”.
SurfLab has been at the cutting edge of geometric and graphics research for the last 20 years. Besides academia, SurfLab alumni work at all major graphics hardware and graphics software companies and some industrial and entertainment modeling companies.
SurfLab has been engaged in a number of interdisciplinary collaborations (dentistry, astronomy, mechanical engineering). Most recently, Surflab has received a major grant to develop the TIPS 3D graphics, simulation and force-feedback environment. (TIPS will allow surgeon-educators to author 3D interactive surgical training and learning units. This research combines aspects of Human-Centered Computing and Interactive Biophysical Simulation.)
“Correct resolution rendering of trimmed spline surfaces” addresses the hard problem of real-time, pixel-accurate display of curved surfaces, called spline or NURBS surfaces — with the additional challenge of complex cut-outs, called `trims’. Trimmed NURBS surfaces are the industry standard representation for designing free-form shape. Free-form designs cover effectively everything curved that is not a sphere, cone or cylinder. Free-form designs are ubiquitous in every-day life, ranging from cell-phones to air planes to modernistic building designs. NURBS are also the common language of high-end simulation of physical processes, and of compressing and visualizing large data sets.
The contribution of the paper is both theoretical and practical: theoretical in that it leverages a new explicit polynomial interval arithmetic to obtain sharp estimates (approximation theory); practical, in that it develops a new efficient data structure for parallel GPU computing (graphics and parallelism). Together this yields a fast algorithm with guaranteed sub-pixel-accuracy. The key breakthrough is to automatically and in real-time predict how fine an evaluation of the trim curves (in the NURBS domain) results in correct trim decisions for each sub-pixel on the screen.
Computer & Information Science & Engineering
Herbert Wertheim College of Engineering