CAREER: The Inverse Shade Tree Framework for Material Acquisition, Analysis, and Design

University Of Virginia Main Campus, Charlottesville VA

Investigators

Abstract

CAREER: The Inverse Shade Tree Framework for Material Acquisition Jason Lawrence, University of Virginia Digital models of the way materials scatter light are necessary to synthesize the visual appearance of real-world 3D objects and have applications ranging through digital arts to manufacturing, immersive virtual environments, remote sensing, and archaeology. However, capturing and authoring these models remains impractical due to the delicate and tedious measurement techniques currently available and the paucity of structured representations and editing tools. This research involves casting these problems in a new mathematical framework that will lead to the development of more useful representations and easy-to-use acquisition methods. The project is investigating hierarchical Bayesian models and inference techniques to provide compact and structured representations for measured appearance data in computer graphics. These models have the property that the latent structure inferred from high-dimensional and large (multi-gigabyte) datasets corresponds to intuitive and editable components, thus greatly extending the utility of measured data within production settings. Existing representations become special cases within this general probabilistic setting, leading to the development of entirely new representations for a broader class of materials than previously studied. Additionally, this research investigates new measurement techniques that combine probabilistic inference with a material database to provide fast and robust acquisition in noisy environments