Happy with how well it works already, but will be a lot more stable when combined with some low res but more accurate radiance/visibility information.
Happy with how well it works already, but will be a lot more stable when combined with some low res but more accurate radiance/visibility information.
I've done two things so far:
- Sampling a distant light instead of computing AO
- Added a heuristic for estimating occluder thickness
Details and links below
I've done two things so far:
- Sampling a distant light instead of computing AO
- Added a heuristic for estimating occluder thickness
Details and links below