Lecture 13: Global Illumination & Path Tracing (104)
litony396
It is interesting to see the diminishing returns of adding more samples per pixel. Going to 32 samples from 1 sample made a huge difference while further increasing to 1024 makes the image look nicer but with much more work for way less improvement.
Boomaa23
@litony396 I agree that it's interesting to see the diminishing returns, but that maybe it would be better if we could see how 64 or some other smaller intermediate value between 32 and 1024 behaves. That way you can tell if that diminishing returns actually occurs at a certain point or is just due to the finite limitations of screen resolution.
Edge7481
are there methods similar to anti-aliasing that allow lower samples to look more palatable without taking more samples? the 32 samples example appears pretty terrible by itself but I'd imagine there are ways to average the values and get it looking like the 1024 samples example without doing more path tracing
It is interesting to see the diminishing returns of adding more samples per pixel. Going to 32 samples from 1 sample made a huge difference while further increasing to 1024 makes the image look nicer but with much more work for way less improvement.
@litony396 I agree that it's interesting to see the diminishing returns, but that maybe it would be better if we could see how 64 or some other smaller intermediate value between 32 and 1024 behaves. That way you can tell if that diminishing returns actually occurs at a certain point or is just due to the finite limitations of screen resolution.
are there methods similar to anti-aliasing that allow lower samples to look more palatable without taking more samples? the 32 samples example appears pretty terrible by itself but I'd imagine there are ways to average the values and get it looking like the 1024 samples example without doing more path tracing