It seems like allowing refractory bounces to die less often than normal bounces would help make the scene look closer to the higher-bounce versions in the places where the major differences occur---namely the glass. Is this a common-ish rendering tactic?
keenan
@clam That's a nice idea. For instance, you could use Russian roulette to kill diffuse paths with higher probability than specular paths. It's of course a heuristic, and may not work out well for all scenes: for instance, you can imagine a scene with a lot of very bright diffuse surfaces (e.g., white painted walls in direct sunlight) and a lot of dark refractive surfaces (e.g., nearly black gemstones sitting in the shadows). In this case you might still be putting a lot of effort into something that doesn't matter. Coming up with strategies that work well for scenes you didn't anticipate is kind of the "name of the game" in rendering. A classic paper in this spirit is the one on Multiple Importance Sampling by Veach and Guibas.
It seems like allowing refractory bounces to die less often than normal bounces would help make the scene look closer to the higher-bounce versions in the places where the major differences occur---namely the glass. Is this a common-ish rendering tactic?
@clam That's a nice idea. For instance, you could use Russian roulette to kill diffuse paths with higher probability than specular paths. It's of course a heuristic, and may not work out well for all scenes: for instance, you can imagine a scene with a lot of very bright diffuse surfaces (e.g., white painted walls in direct sunlight) and a lot of dark refractive surfaces (e.g., nearly black gemstones sitting in the shadows). In this case you might still be putting a lot of effort into something that doesn't matter. Coming up with strategies that work well for scenes you didn't anticipate is kind of the "name of the game" in rendering. A classic paper in this spirit is the one on Multiple Importance Sampling by Veach and Guibas.