Hi!

I'm trying to use pre-integrated sss with SH as mention in the RAD 2013 course note, and I have some question.

1. One need to convolute D(theta,r) with Y(l,0), why is theta in the range of [0, Pi/2]? For clamped cos, only the upper hemisphere matters, but for skin scattering function, shouldn't it be [0, Pi]?

2. I have compute three 1D texture, parameterized by curvature, using R8G8B8 format, and the color of the l=3 texture looks too blue.

I have the same problem with directional light texture as well, especially in the bright side of the small radius(big curvature) region. But after some curve adjustment in photoshop, the blue vanish and look exactly like the one in the Penner and Borshukov PPT.

Any suggestion? Thank you.

[attachment=23250:sh2.jpg][attachment=23251:scatter.jpg]

I think the blue part has something to do with the Gaussian.

According to http://http.developer.nvidia.com/GPUGems3/gpugems3_ch14.html , the weight of the blue light at variance 0.0064 is greater then the red light, so on the left side of the curve, blue should be "taller" then red. This is confirm by the plot given in the Penner PPT. But in GPU Gem3, the red curve is always on top of the blue and green one(GPU Gem3 plot is r x R(r), while Penner PPT plot is R(r)). I have try to plot the r x R(r) curve, the blue one is still the tallest.

This Gamasutra post also has the "blue problem" : http://www.gamasutra.com/view/news/128934/Indepth_Skin_shading_in_Unity3D.php