Computer Science Colloquia
Wednesday, May 1, 2013
Advisor: Wes Weimer
Attending Faculty: Jason Lawrence, Worthy Martin, and Andrew Grimshaw
4:00 PM, Rice Hall, Rm. 242
Master's Project Presentation
Synthesizing Novel Reflectance Functions with Genetic Programming
Accurately modeling material reflectance plays a critical role in photo-realistic rendering. Reflectance functions consist of a compact, analytical expression that models light scattering off a material. Deriving reflectance functions is difficult, with only a handful of such models currently in use despite decades of computer graphics research. This project presents an automated technique for synthesizing reflectance functions that better model real world materials. The technique uses Genetic Programming (GP) search to explore the space of possible reflectance functions, favoring those that more accurately model a diverse set of example materials. The result is a new compact, accurate, adjustable, and generalizable expression that models a class of materials better than any existing model.