This was coded as a part of my diploma thesis on physically-based shading (or rendering). Importance sampling (IS) is essentially a Monte Carlo integration method used in rendering diffuse and glossy surface reflections (BRDFs evaluated in semi-random directions across a hemisphere domain). But plain importance sampling requires a large number of samples to produce accurate results in many cases. Filtered IS adds filtering (doh) of environment maps (mipmaps) to reduce the number of needed samples. See resources for more theory. Online here.
- Importance sampling Trowbridge-Reitz normal distribution fuction (see e.g. http://graphicrants.blogspot.com/2013/08/specular-brdf-reference.html)
- Halton quasi-random sequences
- HDR (RGBM encoded) environment map in panoramic format
SIGGRAPH 2010 Course: IS for Production Rendering
Importance Sampling for Production Rendering