Stochastic Sampling and Distributed Ray Tracing
1. Introduction
Rendering algorithms can be classified as analytic and discrete according to how they approach the aliasing problem. Discrete is better than analytic algorithms.
- super sampling
- adaptive sampling
- stochastic sampling (here we introduce)
2. Uniform point sampling
A review to uniform sampling as mentioned in paragraph 1.
3.Poisson disk sampling
Nonuniform distribution sample found in human eye called Poisson disk distribution, which replaces aliasing by noise.
4. Jittering a regular gird
This method produces good results and is well suited to image rendering algorithms, but it can remain small amount of aliasing as it is not as good as Poisson disk sampling.
5.Distributed ray tracing
Definition of distributed (probabilistic) ray tracing
If we regard the variables of the integration as additional dimensions, we can perform a Monte Carlo evaluation of the integrals by stochastically distributing the sample points(rays) in those additional dimensions.
5.1 Shading
How to calculate intensity I of reflected light at a point on the surface.
5.2 Depth of field
I don’t understand the formulas thoroughly.