[PAST EVENT] Colloquium: Appearance Modeling for Realistic Computer Graphics

September 9, 2011
3pm - 4:15pm
Location
McGlothlin-Street Hall, Room 020
251 Jamestown Rd
Williamsburg, VA 23185Map this location
Throughout its history, the field of computer graphics has been striving towards increased realism. Photo realistic image synthesis has traditionally involved the development of algorithms for the simulation of physically accurate light transport in a scene. However, the quality of rendering produced by these algorithms is limited by the quality of the input scene descriptions such as materials and illumination models. With the advances within the field of digital photography over the last decade, there has been significant interest in acquiring such models from photographs. This has given rise to the modern image synthesis pipeline for realistic computer graphics consisting of the following stages: acquisition, modeling and
rendering.

In this talk, I will present a set of new techniques for efficient acquisition and modeling of reflectance properties of real world materials and human faces, as well as high quality rendering with
acquired data. In particular, I will describe a novel approach for the
acquisition of bidirectional reflectance distribution functions (BRDFs) of materials by projecting basis illumination, and a technique for practical modeling and rapid measurement of layered facial
reflectance using a few controlled lighting conditions. Here, I will discuss some approaches for separation of individual reflectance components and fitting measured data to appropriate reflectance and scattering models. I will also introduce two novel computational
illumination approaches to reflectometry - one based on the second order statistics of measured reflectance and another based on analysis of the Stokes parameters of incident circularly polarized illumination
- that provide direct estimates of spatially varying material
reflectance and can drive analytic BRDF models for realistic rendering.
Contact

Department of Computer Science