Multiple Importance Sampling
First of all, I need to familiar with Multiple Importance Sampling (Learned by veach-chapter9)
Different sample strategies have different advantages.
Monte-Carol method need to divide the probability depend on what kind of sample method you use
For combination of different sample strategies, we use following formula and assumptions :
p(x) is probability for area, a cos from the projection and another from projected solid angle
The things can be improved is the kind of weighting function as the weighting function can be different between different sampling strategies and different sample points. The examples of weighting functions are shown below:
Bidirectional path tracing
Now different sample method become different s and t. They represent the length of light subpath and eye subpath,
We can divide the formula into two parts which can reduce the calculation cost as the sample strategy is sample two kind of subpath independently and then combine them.
Wait continue…