Eugene d’Eon | Jaroslav Křivánek
In ACM SIGGRAPH 2020 Courses: Advances in Monte Carlo rendering: the legacy of Jaroslav Křivánek
Summary
In these course notes we present an intuitive approach to zero-variance estimator derivation to complement the theoretical literature. Included are several novel, previously unpublished results that expand on the previous work of the authors from 2014.
Highlights include:
- Two new perfectly-zero-variance estimators for escaping half spaces with isotropic scattering
- The first application of zero-variance theory to non-exponential random/correlated media (which also generalizes the exponential transform)
- An exit-resampling modification of the Dwivedi scheme that reduces variance by 10-40x
Talk Video/h3>
Course notes (corrected)
Example Code
Mathematica implementations of exactly-zero-variance walks in half spaces with isotropic scattering:
mathematica_examples.zip
Official Publication
ACM SIGGRAPH 2020 Courses
Article No.: 3
Pages 1–366
https://doi.org/10.1145/3388769.3407458
