Acquiring, Simplifying, and Rendering Geometry

My computer graphics research has involved the representation of geometry, specifically in three areas: (1) the creation of surface models from scanned data, (2) the multiresolution representation of surface geometry, and (3) the parametrization of surfaces for efficient rendering. In this talk, I'll present an overview of this research. When I begun studies in computer graphics, the main source of geometric models were the computer-aided-design (CAD) descriptions of manufactured parts. In the early 90's, 3D scanning systems became commonplace, allowing geometry to be captured for everyday objects. My PhD work addressed the reconstruction of surfaces from scanned data. Surface reconstruction has led to an abundant source of complex geometric models. In turn, model complexity posed to a number of problems. In particular, it is difficult to visualize detailed models interactively. Thus, my next area of research was the multiresolution representation of geometry. I developed progressive representations that allow efficient transition between different levels of detail, and permit detail to be locally adapted in real time as the viewer moves. Such representations are also well suited for compression and progressive transmission of geometry. In the last few years, my research emphasis has shifted to creating parametrizations over surfaces, to allow signals to be mapped onto geometry. While graphics hardware has long supported "texture mapping", recent programmable hardware offers many types of interesting effects. The key to realizing these effects is to develop automated tools to efficiently associate attribute signals with geometry.