Physically-based specular contribution

[airstyle]

Hi I've just got a question out of pure curiosity regarding physically based specular contribution and V-Ray.

This is a question for anyone at chaosgroup or anyone who knows the theoretical ins and outs of V-Ray.

Specular contribution has always been a 'fake' effect in incorrect renderers like mental-ray; From my understanding in v-ray when modifying the BRDF properties of a surface, when you leave the 'highlight-glossiness' and 'reflection-glossiness' linked it promotes energy preservation (being physically correct).

However in the latest V-ray for Maya version with a new "specular" and "diffuse" contribution slider, the ability to turn the specular component OFF but preserve reflection has lead me to some confusing queries about v-ray in general:

If specular 'highlights' are really a result of reflection glossiness (that is, the actual highlight is a reflection of an area light), how physically would you turn just the highlight off without turning the reflection off?

I guess the real questions is, what is the specular contribution we see when we render a specular pass in V-Ray?

Would I be way off-course by inferring that this has to do with manipulating BRDF properties differently for light-based and BRDF-based sampling?

Cheers,
Greg.

[Morbid Angel]

how physically would you turn just the highlight off without turning the reflection off?

If you split highlite and reflection glossiness, then set the highlite to 1.0 and reflection glossiness to something lower then 1, you will have reflection, but no specular highlite.

[super gnu]

hm.

i dont know if i understood fully the specific question being asked here, but its something ive always wondered.

we are advised to leave the specular linked to the reflection glossiness to get a correct result.

however, surely, as mentioned above, the specular highlight is just a fake effect, and having it turned off, we would still get correct "speculars" due to bright objects and lightsources being reflected on our reflective surfaces. it then follows that having it on, would we not get the fake specular highlight -and- the reflection of the area lightsource.. i.e an incorrect result? or doesn having the specular effect enabled somehow attenuate the highlights in the raytraced reflection..?

i think maybe my understanding is a bit off, but i dont see the difference between a specular highlight and any other reflection component. its just the reflection of a very bright object no? and if this is the case, why seperate it in this way, apart from allowing you to cheat it without doing the full blown raytracing job.

[vlado]

however, surely, as mentioned above, the specular highlight is just a fake effect, and having it turned off, we would still get correct "speculars" due to bright objects and lightsources being reflected on our reflective surfaces.

Nononono. This is a misconception when applied to V-Ray. "Speculars" and "glossy reflections" are two ways to calculate the exact same thing - the effect of a light reflecting in a glossy surface. One method relies on directly tracing shadow rays towards the light (the "specular" way), and the other - on tracing reflection rays according to the surface BRDF. In some cases, one method is better (i.e. less noisy), and in others it's the reverse. There is a technique called "multiple importance sampling" which allows a renderer to combine both strategies optimally in order to get a less noisy result. This is exactly what V-Ray does. So the complete final result is split between the "specular" way of calculating this effect, and the "reflections" way of calculating it. When both are added together, you get the correct result. More details can be found in the original paper that introduced MIS along with some useful images: http://www-graphics.stanford.edu/papers/combine/

but i dont see the difference between a specular highlight and any other reflection component.

The difference is not in what is calculated, but in how it is calculated.

why seperate it in this way, apart from allowing you to cheat it without doing the full blown raytracing job.

Because, you get a less noisy result with fewer samples. Which is a worthy goal

Best regards,
Vlado

[vlado]

If specular 'highlights' are really a result of reflection glossiness (that is, the actual highlight is a reflection of an area light), how physically would you turn just the highlight off without turning the reflection off?

There is no physically correct way to do that, of course. If you set the shader to use different reflection and hilight glossiness, it is no longer a physically accurate shader. The ability to do that was added simply as a convenience for "artistic" purposes by request of users. If it was up to me, I would not have allowed it.

Best regards,
Vlado

[joconnell]

Maybe that's a question then - why the two components of the highlights? Fair enough about the method, but why for example when you use a glossy material, are you getting an accurate raytraced style highlight component in the reflection pass, but also an old fashioned blinn type highlight also in the specular pass?

[vlado]

Maybe that's a question then - why the two components of the highlights? Fair enough about the method, but why for example when you use a glossy material, are you getting an accurate raytraced style highlight component in the reflection pass, but also an old fashioned blinn type highlight also in the specular pass?

They both represent the respective contribution of each calculation method to the final result. Neither of them is the full completely accurate result - you have to add them together for the final result. Again, if it was up to me, I would have included everything in just one reflection element, but separating them has been useful for "artistic" purposes.

Best regards,
Vlado

[joconnell]

Okay good to know. And yep having a black and white spec pass for all the compers to glow and highlight is definitely a useful one. They love their glows do our compers

[super gnu]

ahh.. so finally i understand.. thanks for the solid explanation Vlado.

one last question. if i turn off the specular component (i.e. unlink and set the specular to 1.0) does vray understand this and give me the fully accurate raytraced reflection including specular components, or is it now missing something...

im guessing since we are advise to leave specular enabled, i already know the answer.

its just im generally in the habit of removing the specular entirely, since i have been known in the past to get nasty surprises - a camera looking at a scene with some nice glossy tarmac turns towards the sun, BAM the whole road network dissapears in a big specular highlight..

now i expect this is my fault, shoddy materials work, not an issue with vray at all.. however i do find things more predictable without the specular element in there. i still get highlights from the sun etc, but they just seem to be less exaggerated (or more likely, they are now attenuated unrealisically)

[vlado]

one last question. if i turn off the specular component (i.e. unlink and set the specular to 1.0) does vray understand this and give me the fully accurate raytraced reflection including specular components, or is it now missing something...

V-Ray will still give you the same reflection result as if the reflection and hilight glossiness were locked together. So yes, it is missing something - reflections of [area] lights will be dimmer or missing. Depending on the case, this may not be a big issue.

now i expect this is my fault, shoddy materials work, not an issue with vray at all.. however i do find things more predictable without the specular element in there. i still get highlights from the sun etc, but they just seem to be less exaggerated (or more likely, they are now attenuated unrealisically)

It's good that we have the option to split them, then In any case, for many materials, using proper Fresnel falloff usually helps to reduce unruly hilights.

Best regards,
Vlado

[airstyle]

Haha Vlado you answered my question indirectly, MIS is what I was heading towards so as I thought, you manipulate the light-based sampling differently to the BRDF-sampling to achieve an artistic effect

This is something I first noticed a few years ago in V-Ray when I was trying to smooth out the 'specular' component of a render by pumping the glossy samples up - the reflection smoothed out but the specular remained noisy due to it sampling based on light, not the material

FWIW I love the fact you can split the specular and reflection components up, I'm just curious as to why how, and what makes V-Ray better than other renderers in this instance