A Simpler and Exact Sampling Routine for the GGX Distribution of Visible Normals

Eric Heitz (1)

1 Unity Technologies [San Francisco]

Heitz and d'Eon introduced importance sampling based on the distribution of visible normals (VNDF) and provided analytic solutions for Beckmann and GGX distributions. In this document, we propose an alternative sampling routine for the GGX VNDF that is simpler and exact. The solution provided by Heitz and d'Eon was approximate for GGX because it required a fitted curve: the derivations were made in slope space and the conditional inverse CDF of GGX does not have a closed form with this parameterization. This improved sampling routine uses the property that for unit roughness alpha = 1 the GGX distribution is a uniform hemisphere. Sampling the GGX VNDF is thus equivalent to sampling the projected area of a hemisphere, which can be done exactly.

Domaines: Synthèse d'image et réalité virtuelle [cs.GR]

Eric Heitz. A Simpler and Exact Sampling Routine for the GGX Distribution of Visible Normals. [Research Report] Unity Technologies. 2017. ⟨hal-01509746⟩