We accelerate spherical filtering using Isotropic Spherical Decomposition, decomposing spherical filters into a linear combination of isotropic kernels. Our formulation is flexible to the choice of isotropic kernels, and we demonstrate practical applications: RT shading with measured BRDFs, anti-aliasing environment mapped mirror reflections, and filtering of noisy reflectance data.

Cyril Soler, INRIA
Mahdi Bagher, Microsoft Corporation
Derek Nowrouzezahrai, University of Montreal