rouncer at August 24th, 2012 13:51 — #1
I tried this once, I did 100 samples a frame, I averaged the samples recieved by the amount of them, but the screen slowly reduced to black with all the bounces that never recieved light i guess, im not sure really, but I didnt get the "gradual converge" I actually boosted it by a value and at least got an image but it slowly got brighter and brighter the more samples added to the result.
Is there something special your meant to do with the light recieved at a pixel?
If there is any cool material someone has a link to, that would be cool too.
alienizer at August 24th, 2012 14:58 — #2
Not suppose to do that. If you get 100 sample and use outColor = avgColor / 100.0f; then it should stay at the same intensity no matter what! But then you have to do the same per frame, if you have 8 frames of 100 samples, then outColor = avgColor / (Samples*Frames);
vilem_otte at August 24th, 2012 18:20 — #3
Actually naive path tracing will converge very slowly for small lights (you will need tens of thousands, or maybe millions of samples to achieve good looking quality). So using naive path tracing is not as useful as it seems to be. What you probably want to look into is explicit path tracing (first) - this one estimates direct light visibility for each path tracing iteration - speeding up small lights a lot, although as you think this isn't a miracle and caustics will converge almost as slow as using naive ray tracing ... of course there are other extensions to this (bidirectional path tracing, and when you research into it, try looking on importance sampling also).
rouncer at August 25th, 2012 02:13 — #4
Thanks for info, so it is just average, I think something was going wrong with my height map intersection. I was lighting up a depth map... incorrectly
naive would work best for sunlight? thats all my test was.