Completely wrong

[davidmanzanares]

I'm sorry but the code has multiple bugs, it doesn't match the Oren Nayar formula

• Variable A , is completely wrong calculated, your code even uses variables that are not needed to calculate A.
• Variables s and t are probably wrong too. I'm not sure about this one, since your formulas can be equivalent... But I don't think so... A prove would be appreciated if I'm wrong.

I guess that this is un-maintained code, but the fact that this is the first Google result of "oren nayar glsl" makes me want to warn everyone from using this.

The original Oren-Nayar paper can be found here: http://www1.cs.columbia.edu/CAVE/publications/pdfs/Oren_SIGGRAPH94.pdf

[mikolalysenko]

If you want to send a PR, go for it.

[davidmanzanares]

I'm currently writing my own Oren-Nayar code. I'll probably upload it to another repository since I don't like/need all the JS around this shader. However, I searched around this topic (Oren-Nayar) and I only found:

• your code (legible but wrong)
• http://shaderjvo.blogspot.com.es/2011/08/van-ouwerkerks-rewrite-of-oren-nayar.html (I think it is right, but not legible)
• http://ruh.li/GraphicsOrenNayar.html (Probably right but it uses acos which is not a good idea: hurts performance)

I can comment here my current version, which works. However, I'm not sure about its correctness.

[mikolalysenko]

There isn't any JS code in the repo, just GLSL. If you want to to fix it up just go for it.

[ranjak]

@dv343 Apparently the shader was implemented according to the instructions found on this post.

The A and B first mentioned in this post are not the same as the ones used in the "qualitative" model, as the author has apparently calculated them from the full model instead. Unfortunately, the author doesn't give the detail of his calculations.
As for the rest of the formula, I'm quite confident it is correct, as I did the math myself on paper starting from the original qualitative formula, and reached the same result.
If you're interested, I implemented it here using the original A and B from the qualitative formula (so my implementation should be 100% correct wrt the original qualitative model).

The only incorrect thing I can see here is in the documentation: sigma (the roughness parameter) is an angle in radians, AFAICT between 0 and PI/2, not a ratio.
However, it would be a ratio if using the A and B constants created by Yasuhiro Fujii in the aforementioned post (the one he uses to fix the energy conservation law violation). In this case, it is called sigma' instead of just sigma. For reference, this method is implemented in Blender's Cycles renderer (here).