Theoretical material discussion
polycounter lvl 11
I would like to discuss and gain a better understanding of the logic behind materials, no programming, just theory.
One of the key reasons for me bringing this up is because, from my understanding, there are just about no real materials that have Lambert shading. Also real materials obey the conservation of energy law, which often seems to be overlooked.
So from my understanding of when an object is lit, some of the light bounces off of the surface as specular and some bounces around in the surface and is partially absorbed by the material, giving it its color. So when there is a linear diffuse texture/color, multiplying that by the light describes the light being absorbed and the resulting diffuse. So what the diffuse texture really describes here is the percent of light that the material absorbs/does not absorb.
The diffuse itself has a material specific falloff based on the molecular structure and roughness of the surface.
The diffuse and the specular are added together as percentages of the light. So if I multiply my specular by .3 I should multiply my diffuse by .7 before adding them both together.
The ratio of the specular (reflected light) and diffuse (scattered light) changes with the angle between the light source and the camera, and that ratio depends on the material and its molecular structure.
As far as some ideas, I think it would make sense to base all lighting for an object (its diffuse lighting, its fresnel and specular reflections) on the "image" of the environment, so something like an ambient cube map for the diffuse and an environment cube map for all specular and reflections.
If I have anything wrong here or if there is anything that could be expanded more, please say so. If there are any properties that a material should have that aren't considered here please add them and how they would factor into things.
One of the key reasons for me bringing this up is because, from my understanding, there are just about no real materials that have Lambert shading. Also real materials obey the conservation of energy law, which often seems to be overlooked.
So from my understanding of when an object is lit, some of the light bounces off of the surface as specular and some bounces around in the surface and is partially absorbed by the material, giving it its color. So when there is a linear diffuse texture/color, multiplying that by the light describes the light being absorbed and the resulting diffuse. So what the diffuse texture really describes here is the percent of light that the material absorbs/does not absorb.
The diffuse itself has a material specific falloff based on the molecular structure and roughness of the surface.
The diffuse and the specular are added together as percentages of the light. So if I multiply my specular by .3 I should multiply my diffuse by .7 before adding them both together.
The ratio of the specular (reflected light) and diffuse (scattered light) changes with the angle between the light source and the camera, and that ratio depends on the material and its molecular structure.
As far as some ideas, I think it would make sense to base all lighting for an object (its diffuse lighting, its fresnel and specular reflections) on the "image" of the environment, so something like an ambient cube map for the diffuse and an environment cube map for all specular and reflections.
If I have anything wrong here or if there is anything that could be expanded more, please say so. If there are any properties that a material should have that aren't considered here please add them and how they would factor into things.
Replies
The lambertian diffuse model is no where near physically correct.
Most often this is because our output environment, the screen, has a limted luminance output, far less than your average artificially lit environment, and many orders of magnitude lower than direct sunlight.
The virtual lighting within the 3d environment is often limited in a similar way, from off (0) to full bright (1). We're getting better lighting tech all the time. now with floating point intensity lights and screenbuffers, and tonemapping to keep the scene's output within the display device's gamut, but we're still approximating things for aesthetic concerns, and computational speed. Physical accuracy hasn't been a concern, visual accuracy, or rather, visual acceptance, has been, and continues to be.
In real life, the reflected light from an object is never 100% of the incoming energy, regardless of how much is reflected or re-emitted. Much is either absorbed or converted to other wavelengths, such as heat, which do not figure in cisually.
Using cube maps for lighting works fine. It can give very nice results as well as its a sort of "radiance capture" of the scene. But again this comes to a technological bottleneck. Ideally each object would need its own radiance map, updated every frame, and stored in some format that preserved overall luminance. This gets very expensive very quickly.
You could probably apply the same conserving thinking when using your Cube Map as well. If you use your Cube Map for both ambient lighting and reflections you could use similar ratio thinking on them as well, since they represent similar things as the diffuse/specular. If your object is highly reflective you won't apply as much ambient factor, and vice versa.
http://renderwonk.com/publications/s2010-shading-course/
http://advances.realtimerendering.com/
http://www.huevaluechroma.com/021.php
http://www.itchy-animation.co.uk/light.htm
Don't get obsessed with making one master material that does everything. In my experience it's just too much hassle most of the time. It's good to instance materials but you can usually get better results just by making a few different master materials that do different things.