I might be mistaken but the only non-symmetry thing is that you don't scale photon (particle) energy, tracing it, at all (besides dividing by rr survival prob. And such).
And Yea 5 minutes with cpu usage jumping like crazy due to the garbage collector abuse :).
Veach said in his dissertation that tracing from the light source might yield different results than tracing from the eye and that's for two reasons:
1- Refraction (Light gets compressed when going from a less dense to a more dense medium and "spread out" in the opposite direction) 2- Smoothed normals
For example, if you don't handle point 1) explicitly you should see brighter caustics compared to path tracing.
Thats why I said about not scalling the photon energy. Like I said before, I skimmed veach thesis on that issue yesterday and if I am not mistaken it comes just to that. That's also what has been said in Jensen's book.
The smoothing normals though, are yet not that clear issue to me, and I am not yet sure how does it affect reflection/refraction. Purely estimating by eye, looking on test renders, just using smoothed normals in place of geometric ones, while tracing particles in interaction with glass/perfect specular seem to work just fine.
5 mins? Damn this is fast :o now sppm is on my todo list for shure :)
ReplyDeleteBtw, in sppm do you have to handle the non-symmetry due to refraction?
I might be mistaken but the only non-symmetry thing is that you don't scale photon (particle) energy, tracing it, at all (besides dividing by rr survival prob. And such).
ReplyDeleteAnd Yea 5 minutes with cpu usage jumping like crazy due to the garbage collector abuse :).
Veach said in his dissertation that tracing from the light source might yield different results than tracing from the eye and that's for two reasons:
ReplyDelete1- Refraction (Light gets compressed when going from a less dense to a more dense medium and "spread out" in the opposite direction)
2- Smoothed normals
For example, if you don't handle point 1) explicitly you should see brighter caustics compared to path tracing.
Thats why I said about not scalling the photon energy. Like I said before, I skimmed veach thesis on that issue yesterday and if I am not mistaken it comes just to that. That's also what has been said in Jensen's book.
ReplyDeleteThe smoothing normals though, are yet not that clear issue to me, and I am not yet sure how does it affect reflection/refraction. Purely estimating by eye, looking on test renders, just using smoothed normals in place of geometric ones, while tracing particles in interaction with glass/perfect specular seem to work just fine.