Lecture 13: Global Illumination & Path Tracing (69)
TinnaLiu
Since there are so many possible n-bounce paths, in implementation, how do we choose which one (or ones?) to use? Is it by random sample?
yzliu567
I think basically it's by random sampling. For each pixel, we randomly sample some rays and their outgoing directions at reflection according to some distributions, e.g. BSDF, light source.
ncastaneda02
You can also do a markov chain monte carlo sampling, which starts of essentially uniformly random, but will ultimately converge to sampling only high-value paths/near high-value paths until you are satisfied with the result.
Since there are so many possible n-bounce paths, in implementation, how do we choose which one (or ones?) to use? Is it by random sample?
I think basically it's by random sampling. For each pixel, we randomly sample some rays and their outgoing directions at reflection according to some distributions, e.g. BSDF, light source.
You can also do a markov chain monte carlo sampling, which starts of essentially uniformly random, but will ultimately converge to sampling only high-value paths/near high-value paths until you are satisfied with the result.