A rendering technique that attempts to gain 10-100x speed using the same methods that image compression tools use to gain 10-100x space--spatial and temporal coherence. The basic idea is to check geometric and inter-frame predictions using a sparse set of raytraced samples. The method should scale to either CPU or GPU renderers, offline or realtime.
To check out the code and examples:
git clone http://projects.cs.uaf.edu/directcompression
- Permanent Project: Yes