(L) [2014/03/28] [tby Chaos] [Re: Microfacet (RoughGlass) and Multiple Importance Sampling] Wayback!Hi,
This is indeed the paper I was using. There seem to be an error in equation (40), where they should use eta*eta inside the square root (since they are manipulating sine/cosine). I didn't find that mentioned anywhere :/
At the moment I'm stuck on 'why do they use sign(i.n) in this equation (40)'? It means that the equation (16) fails to retrieve the appropriate microfacet orientation, since it's not using (i.n) at all! I'm still unsure of why they are doing that...
(L) [2014/03/28] [tby ypoissant] [Re: Microfacet (RoughGlass) and Multiple Importance Sampling] Wayback!I had a discussion about Walter et al BTDF here on ompf a few months ago. You might find the thread interesting.
[LINK http://ompf2.com/viewtopic.php?f=3&t=1807 viewtopic.php?f=3&t=1807]
And I, too, had difficulties implementing it. In the end, I just abandoned.
(L) [2014/03/28] [tby Chaos] [Re: Microfacet (RoughGlass) and Multiple Importance Sampling] Wayback!I finally managed to get something working! The thing that slowed me down the most was naively inverting the eta based on the 'sidedness' of my view ray, while (as far as I understand) the paper actually relates eta to the normal orientation. This means that the paper's equation manage the special-case vector mirroring, but my custom-eta was nullifying this. And I learned my lesson: maintaining/fixing someone else's algorithmic code can take far more time than actually rewriting all of it [SMILEY :)]