:idea: Big Idea, maybe it helps :-)
Posted: Mon Sep 17, 2012 3:39 pm
OK Guys, i have an idea, maybe you could give it a try...
Maybe im talking BS
but better to say it, maybe this would be THE solution for the fireflies as well as for way faster GI (if it works).
Is it possible to modify the kernel so that it makes an interpolated MonteCarlo pathtracing for GI and caustics (say1:10 samples direct light) and simultanous mixes it with the direct light/reflections/refractions ?
It would take 2 Buffers, i know, but this would give a smooth (no fireflies!) GI suited for animations, also altough interpolated, still good for many pics and nearly noise free, but FAST (fewer rays).
So to summarize again. 2 Buffers. The first (with z-channel!) like Mray's final gathering, GI-diffuse/caustics, interpolated, and mixed with the second one direct lightning/raytracing.
It is clear that it would need to be implemented so that the user doesn't see the 2 buffers, and that the premultiplied light from the 2 is physically correct (although little GI resolution biased)
Of course it is a (little) biased, BUT
it is fast and eliminates fireflies for good. .
How is my idea, is it possible, or tough to implement ?
Maybe im talking BS
but better to say it, maybe this would be THE solution for the fireflies as well as for way faster GI (if it works).Is it possible to modify the kernel so that it makes an interpolated MonteCarlo pathtracing for GI and caustics (say1:10 samples direct light) and simultanous mixes it with the direct light/reflections/refractions ?
It would take 2 Buffers, i know, but this would give a smooth (no fireflies!) GI suited for animations, also altough interpolated, still good for many pics and nearly noise free, but FAST (fewer rays).
So to summarize again. 2 Buffers. The first (with z-channel!) like Mray's final gathering, GI-diffuse/caustics, interpolated, and mixed with the second one direct lightning/raytracing.
It is clear that it would need to be implemented so that the user doesn't see the 2 buffers, and that the premultiplied light from the 2 is physically correct (although little GI resolution biased)
Of course it is a (little) biased, BUT
it is fast and eliminates fireflies for good. .How is my idea, is it possible, or tough to implement ?