A data-driven BSDF framework

We present a data-driven Bidirectional Scattering Distribution Function (BSDF) representation and a model-free technique that preserves the integrity of the original data and interpolates reflection as well as transmission functions for arbitrary materials. Our interpolation technique employs Radial Basis Functions (RBFs), Radial Basis Systems (RBSs) and displacement techniques to track peaks in the distribution. The proposed data-driven BSDF representation can be used to render arbitrary BSDFs and includes an efficient Monte Carlo importance sampling scheme. We show that our data-driven BSDF framework can be used to represent measured BSDFs that are visually plausible and demonstrably accurate.