Would we do this once per ray, or cast several rays like this to essentially Monte Carlo our participating media? I would think that a Monte Carlo approach would lead to less noise in this instance, but also that it would look bad in video due to the randomization of each ray -- I'd imagine each frame would have slight color changes if randomized.
keeratsingh2002
@MillerHollinger I think for participating media rendering, a Monte Carlo approach with multiple rays would be used to average out the light interactions and reduce noise.
noah-ku
I think it's really interesting that randomness can generate realistic cloud effects. On the left, we see a light source that bounces around and hits the camera, whereas on the right, we only see the shadows since the light doesn't directly hit the camera. This is really useful for making hazy effects or soft shadows in scenes.
grafour
I think it's helpful to know (Ren also mentioned this), that this method is uses more computational power since we have to calculate the bouncing inside the media
CLV-Iclucia
It's just like normal path tracing with surfaces. The difference is that in volume rendering the "ray space" contains the whole scene, rather than just the surfaces. And you can also define measurement for these paths. In that way, the path tracing method and volumetric path tracing method are actually unified and you can do all the things you can imagine on monte carlo path integral for volume rendering just like how you do them for path tracing.
Would we do this once per ray, or cast several rays like this to essentially Monte Carlo our participating media? I would think that a Monte Carlo approach would lead to less noise in this instance, but also that it would look bad in video due to the randomization of each ray -- I'd imagine each frame would have slight color changes if randomized.
@MillerHollinger I think for participating media rendering, a Monte Carlo approach with multiple rays would be used to average out the light interactions and reduce noise.
I think it's really interesting that randomness can generate realistic cloud effects. On the left, we see a light source that bounces around and hits the camera, whereas on the right, we only see the shadows since the light doesn't directly hit the camera. This is really useful for making hazy effects or soft shadows in scenes.
I think it's helpful to know (Ren also mentioned this), that this method is uses more computational power since we have to calculate the bouncing inside the media
It's just like normal path tracing with surfaces. The difference is that in volume rendering the "ray space" contains the whole scene, rather than just the surfaces. And you can also define measurement for these paths. In that way, the path tracing method and volumetric path tracing method are actually unified and you can do all the things you can imagine on monte carlo path integral for volume rendering just like how you do them for path tracing.