35
$\begingroup$

Every time I think I understand the relationship between the two terms, I get more information that confuses me. I thought they were synonymous, but now I'm not sure.

What is the difference between "diffuse" and "albedo"? Are they interchangeable terms or are they used to mean different things in practice?

$\endgroup$
28
$\begingroup$

The short answer: They are not interchangeable, but their meaning can sometimes appear to overlap in computer graphics literature, giving the potential for confusion.


Albedo is the proportion of incident light that is reflected away from a surface.

Diffuse reflection is the reflection of light in many directions, rather than in just one direction like a mirror (specular reflection).

In the case of ideal diffuse reflection (Lambertian reflectance), incident light is reflected in all directions independently of the angle at which it arrived. Since in computer graphics rendering literature there is sometimes a "diffuse coefficient" when calculating the colour of a pixel, which indicates the proportion of light reflected diffusely, there is an opportunity for confusion with the term albedo, which also means the proportion of light reflected.

If you are rendering a material which only has ideal diffuse reflectance, then the albedo will be equal to the diffuse coefficient. However, in general a surface may reflect some light diffusely and other light specularly or in other direction-dependent ways, so that the diffuse coefficient is only a fraction of the albedo.


Note that albedo is a term from observation of planets, moons and other large scale bodies, and is an average over the surface, and often an average over time. The albedo is thus not a useful value in itself for rendering a surface, where you need the specific, current surface property at any given location on the surface. Also note that in astronomy the term albedo can refer to different parts of the spectrum in different contexts - it will not always be refering to human visible light.


Another difference, as Nathan Reed points out in a comment, is that albedo is a single average value, which gives you no colour information. For basic rendering the diffuse coefficient gives proportions for red, green and blue components separately, so albedo would only allow you to render greyscale images. For more realistic images, spectral rendering requires the reflectance of a surface as a function of the whole visible spectrum - far more than a single average value.

$\endgroup$
  • 4
    $\begingroup$ Also, I've gotten the impression that "albedo" in astronomy is averaged over the whole visible spectrum (and sometimes a wider spectrum, including UV and infrared), while diffuse/specular coefficients are RGB or ideally spectral quantities. $\endgroup$ – Nathan Reed Aug 26 '15 at 21:00
  • 3
    $\begingroup$ Interesting. In that case, it doesn't make sense to call something Albedo Map does it? $\endgroup$ – Felipe Lira Aug 26 '15 at 22:31
  • $\begingroup$ @PhilLira that does seem like an unhelpful use of the term... Hopefully it won't catch on and cause more confusion... $\endgroup$ – trichoplax Aug 26 '15 at 22:34
  • $\begingroup$ I don't think I get this. Isn't it basically just a minor difference in this case? E.g. what you said implies that the only difference would be that with albedo, the diffuse reflection is (albedo * (1 - specular)) and specular is albedo * specular, instead of flat diffuse and specular numbers? I really don't get it :( $\endgroup$ – Llamageddon Nov 22 '15 at 4:11
  • $\begingroup$ @Llamageddon there are a number of differences covered in the answer but as a simple example: the albedo of a surface could be 0.8, but the RGB value of it's diffuse component could be (0.6, 0.5, 0.9). The albedo is generally just a single scalar value, whereas the diffuse component may have multiple values to give colour rather than just brightness. $\endgroup$ – trichoplax Nov 22 '15 at 12:44
2
$\begingroup$

Briefly:

  • Low albedo --> darker object
  • High albedo --> brighter object

  • Low diffuse reflection --> mirror-like reflection (aka Specular)

  • high diffuse reflection --> cotton-like reflection
$\endgroup$
  • 2
    $\begingroup$ Not sure that is strictly true for diffuse, although I may be wrong. Low diffuse does not imply high specular, what about strongly absorbing materials like carbon or charcoal ? $\endgroup$ – PaulHK Aug 19 '16 at 9:31
  • $\begingroup$ @PaulHK. charcoal have low albedo because low proportion of its incident light is reflected away from a surface. $\endgroup$ – Kamran Bigdely Aug 20 '16 at 0:36
  • $\begingroup$ Ah yes I get what you mean now $\endgroup$ – PaulHK Aug 22 '16 at 2:11
0
$\begingroup$

The Diffuse, specular and Reflection terms have lead to a lot of confusion because they have been commonly used to describe different lighting processes along the CG history and sometimes diverge from their scientific usage.

to clarify this I is use my own vocabulary made up of different terms i picked up here and there:


1- Surface Reflectance:

  • could correspond to specular map in the old system
  • Correspond to the fresnel reflection part of the BRDF model for the dielectric materials and to the global reflection for the metallic ones.

The Surface-Relfectance process description: the light is "bouncing" off the surface without any transmission inside the material or micro subsurface scattering involved in the process (no refraction, no absorbtion). Light color information remain inchanged during surface reflection process except for some precise cases (colored metallic reflection, shimmering)

1.1 - Rough surface reflectance: is light "bouncing" off a rough material (micro facets) in a more or less evenly distributed direction.

1.2 - smooth surface reflectance: is light "bouncing" off a glossy or smooth material in an more or less oriented direction.


2 - Body reflectance

  • correspond to the diffuse in the old system

  • correspond to the basecolor or albedo map in the new system.

The Body-Reflectance process description: The light hitting the surface that is not surface-reflected is first being transmitted in the interior of the object and then may be absorbed, further scattered and reflected, and in some cases exit the material again. It involve micro sub-subsurface scattering from internal irregularities. Light color information is changed during the absorbtion steps of the body relfectance process. And if the light manage to get out of the material again, it will transmit its color information. The body reflectance process is not applyable to metallic material as they only totally absorb or surface-reflect light depending on its wave lenghts.

Body-reflectance will not be influenced by material surface smoothness as there is scattering involved inside the material no matter the surface, except maybe for transparent materials where there is mostly absorbtion process involved (no light deviation) and very few scattering. Then when going out again, the surface roughness could really influence if these light rays are going out parallels or scattered.

Micro subsurface scattering is different than global subsurface scattering as, for a matter of simplification through approximation, the light is considered as going out of the material at the same exact point it went in. This is a just what makes that regular dielectric objects have color; there must be transmission, then absorbtion and micro scattering, then re-transmission outside of the material in order to get dielectric's color


Ok now, what i've understand from this naming confusion:

1 - concerning the diffuse reflection

What we call usually call diffuse reflection is the mechanism that include rough surface-reflectance and body-reflectance for a rough dielectric surface. But in some case the term diffuse reflection can be used to describe only the surface reflectance part when opposed to the transmission process.

Concerning Metallic materials, diffuse reflection concern, as a matter of fact, only rough surface reflectance. In case of smooth metallic material, the term diffuse reflection is replaced by specular reflection or direct reflection (which add to the confusion as specular is used here to mean "sharp").

When talking about smooth dielectric material, there are still diffusing processes, in the sense that the light transmitted into the material is still scattered when goign out of it (body-reflectance), but the surface-reflectance part of it could be called specular or direct reflection.

2 - Concerning the albedo

In the physicist field, Albedo seems to be the ratio between reflected light intensity (surface-reflectance + body-reflectance) and incident light. So it is a one dimentionnal value. In CG, in the other hand, we view albedo as a three dimentional value in RGB which correspond to the traditionnal "diffuse" of the old system and to the "baseColor" of the metall/roughness workflow. In this case albedo would be, for a metal/roughness workflow the body reflectance for dielectrics and and the Surface-reflectance for metals but without the fresnel composant of the fresnel surface reflectance.

But in the physicist way of the term, albedo also cover the surface-reflectance part of the light re-emmission (fresnel reflection).

In the metall/roughness workflow though, BaseColor doesn't have any incidence onto the fresnel reflection wich is directly embedded into the shaders. So BaseColor is basically the body-reflectance RGB value for the Dielectric material and the surface relfectance RGB value being surface-reflected by the metallic material (being "Surface-reflected", but in a colored way because of the conductive property of the metals and their cristalline organisation combined).

It is all really confusing indeed... and i'm not even sure that i entirely get it

One of the doc I'm refering to along with substance PBR guidelines: http://creativecoding.evl.uic.edu/courses/cs488/reportsA/brdf.pdf

$\endgroup$
-1
$\begingroup$

Think of diffuse as the amount of light scattered by the roughness of the surface.

Surfaces such as wood has a high diffuse coefficient due to it's material texture, but if it was sanded, varnished and polished, the surface would be far smoother, thus decreasing the diffuse coefficient, but increasing the "mirror shine", or specularity.

Albedo would be the amount of light, on average, that a surface reflects. The moon, despite being cratered and pitted, reflect a lot of light due to the distance between the observer and the moon, and so, the diffuse of the individual surface regions can be ignored, in favour of an average value. Going back to the lunar example, we know that the moon has many regions, with varying amounts of diffuse and specular values, but at a far enough distance, those values become less important, as we are too far away to make out such detail with the naked eye.

At far distances, you could theoretically, in computer graphics terms, use albedo to compute the average colour of a surface (using sampling), and use the specular coefficient to compute the brightness of that colour. But in practice, LOD texture sampling does the job just fine.

$\endgroup$
  • $\begingroup$ This sounds incorrect. Roughness doesn't cause diffuse, it only affects it due to geometric occlusion. A fine polished marble slate still has the same a strong diffuse, just with slightly different repartition of brightness depending on the view angle. $\endgroup$ – Julien Guertault Jan 8 '17 at 3:18
  • $\begingroup$ Roughness might not be the right word for it, but the principle is there. $\endgroup$ – Ian Young Jan 9 '17 at 4:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.