I will present recent work in modeling the visual appearance of materials. These models are a critical component of computer graphics systems which aim to synthesize realistic images of virtual scenes and computer vision systems which aim to infer properties of a 3D scene from natural images. In particular, I will focus on “data-driven” strategies for modeling complex spatially-varying opaque and translucent materials such as brushed metal, plastic, cloth, wood, candle wax and human skin. This includes our Inverse Shade Trees framework that describes how to decompose these measured high-dimensional functions into compact and understandable pieces that are useful within a production setting. I will also discuss a new design interface that allows a user to specify sparse edits that are intelligently propagated across a large collection of measurements and a new empirical model of translucent materials based on a large simulation. I will conclude with a discussion of interesting open problems in this area.
Jason Lawrence received his Ph.D. from Princeton and is currently an assistant professor in the Computer Science Department at the University of Virginia. His research focuses on efficient representations and acquisition strategies for material appearance and 3D geometry, real-time rendering algorithms, and global illumination rendering algorithms. He was the recipient of a NSF CAREER award titled, “The Inverse Shade Tree Framework for Material Acquisition, Analysis, and Design.”