A novel sampling technique called "toroidally progressive stratified sampling" combines the benefits of stratified sampling with golden ratio ordering to create a sampling pattern that maintains quality at any number of samples and can seamlessly restart without quality degradation. The method uses a low discrepancy shuffle iterator based on rank 1 lattices to visit stratified bins in golden ratio order, making it particularly useful for applications like temporal anti-aliasing where different processes may restart integration at different times while sharing the same sampling sequence.
36 Impressions