The real power of MLT comes out when you've got a scene with very small lights, mostly indirect lighting, or lots of participating media. I suspect that my scene will have all three :D
To that end, here's a little test of brute force vs. MLT for a scene with a very small, very bright light. Each algorithm was given 2 minutes. Here's brute force:
Ouch. Very bad. Here's MLT:
Wow...quite a world of difference!
You might say, why not just do explicit light sampling? Well, it's not that easy when non-homogeneous participating media is involved. I'd rather stick to "dumb sampling, smart searching" (MLT) rather than "dumb searching, smart sampling" (explicit lighting).


No comments:
Post a Comment