Is it equivalent to use the dot product of the reflection direction and view direction, instead of this dot product of h and n?
jacklishufan
The precise value differs because the angle between n and h, and that of v and r are not necessarily equal. In particular,in a 2 d case (assume all vector are co-planer), we can have n=0,l=-45,r=+45,v=+60,h=+7.5, so hn is cos(7.5) but vr is cos(15)
LeslieTrue
Also curious about the angle alpha. Is there any proof in Physics? From my intuition, projecting the reflection light on the view direction would directly lead to a factor of cos 2*alpha.
I'm wondering why Blinn-Phong is, somehow, better than other reflection models? Any proof?
chetan-khanna
(The same question applies to Lambertian shading too.) Let's say we're using this to perform Specular Shading in our color image, so we have one equation L_s = k_s(I/r^2)max(0, cos \alpha)^p for each "channel" RGB in our image. Is there anything wrong with having different k_s or I values for each channel? I ask this because I am wondering if this would be helpful in a situation where we were trying to color the reflected light in a different way to the light source.
justin-shao
It seems the difference between the two methods (nh angle versus vr angle) differs more as the angle increases. Consider the scenario where the vector n coincides with vector l, and vector v coincides with the surface (i.e. orthogonal to vector n). This would produce an L_s that is still >0 - would this be a case where the realism falls apart a bit? Or is it compensated by utilizing a high p-value?
Is it equivalent to use the dot product of the reflection direction and view direction, instead of this dot product of h and n?
The precise value differs because the angle between n and h, and that of v and r are not necessarily equal. In particular,in a 2 d case (assume all vector are co-planer), we can have n=0,l=-45,r=+45,v=+60,h=+7.5, so hn is cos(7.5) but vr is cos(15)
Also curious about the angle alpha. Is there any proof in Physics? From my intuition, projecting the reflection light on the view direction would directly lead to a factor of cos 2*alpha.
I'm wondering why Blinn-Phong is, somehow, better than other reflection models? Any proof?
(The same question applies to Lambertian shading too.) Let's say we're using this to perform Specular Shading in our color image, so we have one equation L_s = k_s(I/r^2)max(0, cos \alpha)^p for each "channel" RGB in our image. Is there anything wrong with having different k_s or I values for each channel? I ask this because I am wondering if this would be helpful in a situation where we were trying to color the reflected light in a different way to the light source.
It seems the difference between the two methods (nh angle versus vr angle) differs more as the angle increases. Consider the scenario where the vector n coincides with vector l, and vector v coincides with the surface (i.e. orthogonal to vector n). This would produce an L_s that is still >0 - would this be a case where the realism falls apart a bit? Or is it compensated by utilizing a high p-value?