Recent searches
Search options
An unusual blog post today! I've long wanted to see some actual numerical estimates on how good importance sampling is at variance reduction in raytracing, and, failing to find anything like that in the literature, I've ended up deriving my own :)
https://lisyarus.github.io/blog/posts/estimating-variance-reduction-in-importance-sampling.html
@lisyarus that's super cool! That last one seems to explain why strong light at grazing angles tends to produce butterflies, right? A typical cos theta importance sampling would undershoot a high energy contribution from grazing angles, leading to higher noise, if I understand your reasoning correctly.
@artificialmind Ohhhh yeah that might be the case! Great observation
@artificialmind Upd: added this observation to the article :)
(with reference to your post, of course)