An implementation of the Normalized BSSRDF via Python port of LuisaCompute.
Tricks applied in the implementation includes:
- Importance Sampling of Normalized BSSDRF via the analytical form of the inverse CDF(yes, analytical).
from math import *
dmfp = 1.0 # diffuse mean free path
def sample_radius(u):
q = 4. * (u - 1.)
x = pow(-0.5 * q + sqrt(0.25 * q * q + 1.), 1 / 3) - pow(0.5 * q + sqrt(0.25 * q * q + 1.), 1 / 3)
return -3. * log(x) * dmfp- Multiple Importance Sampling(light && cosine)
The result looks like below:
