A DATA-DRIVEN BSDF REPRESENTATION
Project no. 542 · Status: Kapanmış · 2015
Summary
In this project, we review measurement techniques used for capturing light reflection/transmission effects in optically thin translucent materials and BSDF (Bidirectional Scattering Distribution Function) models used for representing these measurements. In addition, we propose a novel interpolation technique for sparse BSDF measurements and data-driven BSDF representations are constructed by using the proposed interpolation technique. Light reflection/transmission effects in optically thin translucent materials are represented by four dimensional (4D) Bidirectional Scattering Distribution Function (BSDF). Since BSDF measurements in anisotropic materials are especially needed to be done at 4D space, these BSDF measurements are generally sparse and noisy. Measured sparse 4D BSDF data are interpolated by using a Lagrangian mass-transport solution based technique. In this technique, a group of radial basis functions are fitted to each measured distribution. Then, an interpolation scheme is applied between sparse incident directions by using fitted radial basis functions. In this project, interpolated 4D BSDF data are represented by matrix based and tensor tree based representations compactly and accurately. It’s shown that the proposed BSDF representations provide high accuracy at high compression ratios. To validate the proposed interpolation technique and data-driven BSDF representations, BSDF measurements are done in optically thin translucent materials. Additionally, a comprehensive comparison and analysis is done by using our own BSDF measurements and BSDF measurements in the computer graphics literature.