What probability distribution function is used here? Will it make a big difference if we choose a different pdf and how should we know which pdf is better?
cornrow-kenny
I think the pdf is uniform over the area of the light. Different pdfs can have an effect on the variance of our estimator, with the intuition being that we want to sample more from points that contribute more to the integral.
hilary217
what does sample random point on light means exactly ?
go-lauren
I'm a little confused about what function we are integrating and what variable we are randomly sampling.
hershg
The image on the left shows the shadow cast when sampling directly from the center of the light. with the image on the left, how is that shadow being computed? for each pixel on the 'ground' surface, are we choosing a random value on the lighting surface and then seeing if light hits, which is what gives the spotty pattern? this would make sense, and as the number of samples from a 'ground' pixel to the light source increases and we average over these, we'd get a more accurate representation of the 'smooth' edges of the shadow towards the edges.
What probability distribution function is used here? Will it make a big difference if we choose a different pdf and how should we know which pdf is better?
I think the pdf is uniform over the area of the light. Different pdfs can have an effect on the variance of our estimator, with the intuition being that we want to sample more from points that contribute more to the integral.
what does sample random point on light means exactly ?
I'm a little confused about what function we are integrating and what variable we are randomly sampling.
The image on the left shows the shadow cast when sampling directly from the center of the light. with the image on the left, how is that shadow being computed? for each pixel on the 'ground' surface, are we choosing a random value on the lighting surface and then seeing if light hits, which is what gives the spotty pattern? this would make sense, and as the number of samples from a 'ground' pixel to the light source increases and we average over these, we'd get a more accurate representation of the 'smooth' edges of the shadow towards the edges.