Everything is Shiny

Every once in a while I’ll see a tutorial online or a paper that talks about specular. And it will have a line like “Some materials such as cloth and cardboard are pure diffuse surfaces with no specular”. This is a lie. Every object in the real world has specular, even good old cloth and cardboard. I took these with my homebrew polarized lighting setup. Since I was holding the polarizer manually, the separation isn’t perfect, but seems to work well enough.

First, we’ll start with a cotton shirt. This is the most diffuse one I could find. Most cotton shirts have more specular.




Here is some denim, a shinier cloth material.




Next, here is some cardboard. Yep, it’s shiny.




As a comparison, here is one of my plates. Specular highlights are really bright!




Here are a few other ones. Notice how much specular bricks have.




Finally, here is a piece of concrete from the sidewalk. Sidewalk concrete is much less specular than road pavement.




18 Responses to “Everything is Shiny”

  1. Looks nice, it seems everything is a bit shiny!

    I was wondering if you could comment on the ‘diffuse highlight’ on the left side of the diffuse plate image. Is that really some sort of diffuse ‘highlight’, or is that just a limitation of your home setup?

  2. So, what you think about color of specular?
    Is it “spec.light=light color * gloss” or it may be modified by material?
    Does it correct to use color specular textures on models?

  3. Brad: Yes, it’s a limitation of the setup. The light and camera are so close to the plate that the spot on the left side has a slightly different polarization, so some of the highlight bleeds through.

    Roman: In general, specular is very close to white (i.e. specular color=light color). Some surfaces do get a bit of a tint in their specular such as cloth, some metals, some metallic paints. In general though, I think artists overuse it and put too much color in specular. Good question.

  4. Is all ‘diffuse’ completely absorbed/depolarized, or can it be reflected in random directions by micro-facets without being depolarized?

    Could this mean there is two kinds of diffuse, and the capture is only able to separate one of them (diffuse on the plate), but not the other (‘reflected’-diffuse on the cloth)?

  5. As I understand it, if it hits a microfacet and bounces off without getting absorbed, then it is specular, even if it is a near-random direction. Diffuse only counts if it gets absorbed and retransmitted, with random polarization. Of course, that model is an approximation.

    Now, there are cases where the specular component doesn’t have perfect polarization. So it could be possible that some diffuse surfaces have a not entirely randomized polarization, but I’m not the expert on that.

  6. I guess at that level the distinction between specular/diffuse becomes blurred. One extreme example would be polarization preserving paint for stereo movie screens.

  7. Can you tell me a bit about your setup for taking these pictures? They’re really interesting.

  8. Ah just found your other post where you do just that. Thanks!

  9. Np. (-:

  10. [...] me for information on specular behaviour for materials, to which I sent him a link to the post Everything is Shiny over on the excellent Filmic Games blog. While browsing the site, he came across the post Where are [...]

  11. [...] me for information on specular behaviour for materials, to which I sent him a link to the post Everything is Shiny over on the excellent Filmic Games blog. While browsing the site, he came across the post Where are [...]

  12. I have a couple of things to say about this.

    First, the word “specular” has historically been used for at least two different phenomena:

    o a mirror reflection from a smooth surface, and

    o a smooth-shaped lobe correlated with the “specular direction”, the mirror reflection of the illumination direction.

    This leads to all sorts of confusion. I like to use the following terminology:

    o “Specular” means a mirror reflection.

    o “Diffuse” means an exact Lambertian component (BRDF is constant for all exitant direction).

    o “Directional diffuse” is everything else including, but not limited to, the smooth near-specular lobe.

    This is a phenomenological classification rather than describing the physical processes involved.

    John’s experiments apparently define “specular” as scattering that preserves polarization. This normally correlates to scattering from a single surface as opposed to subsurface scattering or multiple scattering from the surface. But it’s possible to get directional scattering without preserving polarization. And I have heard that for certain metallic surfaces, you wind up with circular polarization from a linearly polarized source! I don’t know how that’s possible, though a transparent dielectric layer could do it, as I recall.

    I’m not sure that everything shown in these images as “specular” fits into the normal connotation of “strongly directional”. The cardboard, in particular, seems a bit odd to me.

    As for color in “specular” scattering (single bounce from the surface), it’s white (actually, the source color) for all dielectrics that I’ve ever heard of. This is because reflectance comes from Fresnel, and the index of refraction is almost constant for all visible wavelengths. Metals are different, as there are several (copper, gold) whose refractive indices vary strongly within the range of visible wavelengths. For those of us who are extremely picky, most metals impart subtle coloration to first-surface scattering. And, of course, first-surface scattering is all you get from a metal.

    Anyway, if we amend the proposition to “everything is non-Lambertian”, I agree. Even surfaces like Spectralon, which is carefully engineered for Lambertian scattering, becomes non-Lambertian near grazing angles.

    It remains a question how perceptible any deviation from Lambertian really is. Obviously lots of people have gotten away with Lambertian for a long time, but of course there are cases (velvet comes to mind) where Lambertian just doesn’t cut it. As in so many areas of computer graphics, it depends.

  13. Stephen,

    I don’t really have a specific physical definition. I definitely do not define specular as scattering that preserves polarization. Rather, polarization is the best tool that I have to help artists visualize the specular-only portion of reflected light. Obviously, manually holding up a polarizer in a room with quite a bit of ambient lighting is imperfect, but if you have a better idea, I’m open to suggestions. (-:

    In general I think of it as given the equations that we have to work with, what values should we use to give us the best approximation of the real world. The point of the post is that artists tend to underestimate the strength of specular. Hence, the images. Also, I’m not quite sure what you’re saying about cardboard. Are you arguing that the light that we see in the polarized image is the result of diffuse? It looks like a typical specular reflection to me, albeit with a wide falloff.

    In terms of color, you should check out the MERL BRDF measurements from several years ago. Go to http://people.csail.mit.edu/wojciech/ and click on “Analytical Fits to Measured BRDFs”. The source data comes from very high quality measurements and he performs best fits of the most common models for Diffuse and Specular. As a practical matter, given the specular models that we actually have, real-world surfaces seem to be best approximated with a non-white specular.

  14. [...] physically based rendering”  http://graphics.cs.kuleuven.be/publications/Phong/ [9] Hable, http://filmicgames.com/archives/547 [10] S. Mikkelsen, Microfacet Based Bidirectional Reflectance Distribution Function [...]

  15. [...] http://www.robinwood.com/Catalog/Technical/Gen3DTuts/Gen3DPages/RefractionIndexList.html [3] Hable, http://filmicgames.com/archives/547 [4] Epic game, “UDK documentation” [...]

  16. [...] but it should be much more saturated than the real-world appearance. I may be useful to have photos taken with a polarizing filter as a [...]

  17. [...] how to split specular and diffuse in real images. He later wrote another article showing that specular often contributes more than we expect for matte materials. Both help giving an idea of the importance of the specular term as well as [...]

  18. [...] value that changes depending on your view angle, which ruins basically all of this. Unfortunately, everything is shiny (even things like cloth that are usually used as an example of what’s not shiny). The [...]