3
$\begingroup$

I am writing a pathtracer, and I want to somehow incorporate importance sampling of the environment map. The obvious way to do this seems to use MIS to either importance sample from my material bsdf or a ray to my environment map. I am having a hard time making it work though, and I am feeling uncertain as to whether I can really say that it is unbiased.

The main problem I have is that if I have a very bright environment light on the opposite side of the material, it will pretty much be sampled at least 50% of time time according to MIS. I should know that sampling that direction won't contribute to reducing variance in any way, as the brdf in that direction is 0. But how can I incorporate this into my MIS "flow" while keeping it completely unbiased???

Improve this question
$\endgroup$
2
  • $\begingroup$ You could add a third sampling strategy that samples in the general direction of that bright light (e.g., a spherical Gaussian) and compute the MIS weights accordingly. This will remain unbiased. $\endgroup$ Commented Nov 21, 2024 at 19:36
  • $\begingroup$ There can be more than two strategies. For example, use one-sample model with three different sampling methods: (1) BRDF sampling (2) precomputed inverse transform importance sampling of the env-map (in PBR-book) (3) cosine-weighted sampling: you know the normal of the hit position, then sample around the normal using cosine-weighted sampling. Method (1) (3) will ensure to have front-facing samples, while method (2) can yield high contribution samples. $\endgroup$ Commented Jan 1 at 4:52

1 Answer 1

Reset to default
0
$\begingroup$

Not sure I understand your question. You can't know ahead of time what direction to the env map you are sampling, so there's not really any way to know it's BRDF contribution.

Improve this answer
$\endgroup$

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.