Eurographics is the top computer graphics conference in Europe, and is highly selective and competitive, with only 17 percent of papers being accepted. Furthermore, Eurographics is the top venue in computer graphics that presents best paper awards, and thus this award is a great credit to the authors' research.
The prize-winning paper, "Wavelet Rasterization," presents a new method of drawing (rasterizing) shapes such as fonts, 2-dimensional polygons and 3-dimensional objects.
"When you draw objects on a screen, the computer decides what pixels are inside the object and which pixels are outside of the object," Schaefer said. "There are only so many pixels in a screen and therefore precise rasterization of the object as we see it in the real world is impossible. The reconstruction process creates jagged edges, which is known as aliasing. There are many anti-aliasing methods out there that attempt to trick the eye into thinking that these jagged edges are smooth.
"Our method is much better at reducing aliasing. We use wavelets to more accurately reproduce a 2D or 3D image on-screen. Wavelets provide a basis for representing the functions that make up an image on a computer monitor through a hierarchy of highly localized refinements. Wavelets have a wide variety of applications, from solving differential equations, to digital image processing, signal processing, and surface reconstruction. The main advantage of wavelets over other representations of a function is that wavelets are localized in both the spatial and frequency domains."
The processes described by Schaefer and Manson are also hardware friendly.
"Another interesting aspect of wavelet rasterization is that it is extremely easy to parallelize," Schaefer said. "Given that there are many processing units on graphics cards, it is important that this solution is able to be processed in parallel. Our algorithm meets this demand by being robust yet modular in order to be executed by many processors at once."
Scott Schaefer is an assistant professor in the Department of Computer Science and Engineering at Texas A&M University. He received his Ph.D. in computer science from Rice University in 2006. He joined Texas A&M Engineering shortly thereafter. His research interests include computer graphics, geometric modeling, and scientific visualization. His student, Josiah Manson, is a Ph.D. candidate and recipient of a National Science Foundation (NSF) Graduate Fellowship.
Picture: The left model shows the result of a binary rasterization technique; the right shows the result of the wavelet rasterization method described in this article.