I thought it was pretty interesting how using the actual normal distribution function allowed for more detailed rendering of the materials. I wonder how much more computational power is needed to generate the actual one instead of the statistical distribution. If my understanding is correct, the actual one has a much more varied and precise directionality of the underlying surfaces. Given this, does it use 3 (in the isotropic case), 4 (in the anisotropic case), or some other number of parameters for the dimensionality?
I thought it was pretty interesting how using the actual normal distribution function allowed for more detailed rendering of the materials. I wonder how much more computational power is needed to generate the actual one instead of the statistical distribution. If my understanding is correct, the actual one has a much more varied and precise directionality of the underlying surfaces. Given this, does it use 3 (in the isotropic case), 4 (in the anisotropic case), or some other number of parameters for the dimensionality?