Lecture 13: Global Illumination & Path Tracing (63)
alvin-xu-5745
It can be seen that reflective elements are clearer with more bounces, as expected (most visible in the background table contents and the overhanging lights). Past a certain point though, I imagine the computational cost of this (which probably does not scale leniently) is no longer worth the quality gains they get. What sort of criteria are used to optimize between these two extremes?
ld184
If we have quite a bit of memory, wouldn't this scale linearly with the operation cost with dynamic programming? Can we not do a second pass to get the 2nd bounce by acting on the results of the first bounce?
joeyzhao123
I'm curious about similar applications in games. I remember times when stepping into a dark area or out into the light, there is a short pause between the calculation of light to illuminate everywhere.
It can be seen that reflective elements are clearer with more bounces, as expected (most visible in the background table contents and the overhanging lights). Past a certain point though, I imagine the computational cost of this (which probably does not scale leniently) is no longer worth the quality gains they get. What sort of criteria are used to optimize between these two extremes?
If we have quite a bit of memory, wouldn't this scale linearly with the operation cost with dynamic programming? Can we not do a second pass to get the 2nd bounce by acting on the results of the first bounce?
I'm curious about similar applications in games. I remember times when stepping into a dark area or out into the light, there is a short pause between the calculation of light to illuminate everywhere.