PDA

View Full Version : Lighting + White/Yellow Surfaces = Blow out


phifehinds
07-27-2010, 05:12 AM
Hi, I work on a production where we light only using direct illumination techniques and ambient occlusion. I've noticed that no matter how hard I try it is very difficult to light surfaces that are pure white or a very bright yellow and not have those surfaces blow out.

If I lower the intensity of my lights then the rest of the scene is not bright enough.

Is there a way around this? Should it be that textures and shaders should never be a pure white color? Because if you think about it... nothing in nature is a perfectly white color unless it emits light perhaps...

Opinions?

azazel
07-27-2010, 09:00 AM
Try exponential color mapping (if you use Vray), or some kind of exposure control, if you use Mental Ray. That can really, really help.

playmesumch00ns
07-27-2010, 01:09 PM
Hi, I work on a production where we light only using direct illumination techniques and ambient occlusion. I've noticed that no matter how hard I try it is very difficult to light surfaces that are pure white or a very bright yellow and not have those surfaces blow out.

If I lower the intensity of my lights then the rest of the scene is not bright enough.

Is there a way around this? Should it be that textures and shaders should never be a pure white color? Because if you think about it... nothing in nature is a perfectly white color unless it emits light perhaps...

Opinions?

Yep nothing should ever be pure white. The brightest things in your scene should be around 80-85% reflective (white paper, snow). Middle grey is around 16-18% reflective.

sundialsvc4
07-27-2010, 02:24 PM
I presume that you are working in a linear color-space? That the "sudden blow-out" effect is not being caused by gamma?

You can control the light on a per-light and per-object basis in ways that do not exist in real life. But, if you find yourself routinely doing so, it implies that you are not now using a linear work flow (where the mapping of RGB values to apparent color is 1:1). Suddenly you would be faced with an exponential curve, such that "a slight change here (on the color curve) is 'slight,' where the same slight change there is 'drastic.'" Upshot: out of control.

If you did have such an issue, "yellow" would be particularly affected, because in the additive, RGB, color-system yellow is produced by the mixture of two colors. (Whereas in the subtractive, CMYK, color space of printing, it is produced by yellow ink ... and so the "troublesome" colors are Red, Green, and Blue.) "White" would be even more affected, because it's a mixture of all three. The more out-of-control electron guns are firing in a very small space (so to speak...), the worse it gets.

Careful sampling of pixel values, and study of the "histogram" of reference images, can be a big help. You really want it to get to something that you can objectively measure, graph and quantify. Although the famous railroad photographer, O. Winston Link, was working in an entirely different medium and with what would now be considered hopelessly primitive technology, he did have a lot of things to say about lighting control, and the use of math to do it.

ndeboar
07-28-2010, 12:40 AM
there's a siggraph paper on just this problem (ill dig it up when im not insanely busy).

What it suggests is desaturating your color the brighter it gets. If you look at real film, you'll notice that highly saturated colors (orange/yellow/red) become less saturated the brighter they are.

playmesumch00ns
07-28-2010, 10:29 AM
there's a siggraph paper on just this problem (ill dig it up when im not insanely busy).

What it suggests is desaturating your color the brighter it gets. If you look at real film, you'll notice that highly saturated colors (orange/yellow/red) become less saturated the brighter they are.

That's a property of the capture device so should be handled in compositing, not by modifying your 3d scene.

salki
07-28-2010, 03:03 PM
You can try changing your frame buffer "Data Type" to RGBA (Float) 4 x 32bit and your file render type to OpenEXR. The latter will allow you to have ridiculous control over your exposure in post production. After Effects has an exposure effect you just throw on the clip to adjust, photoshop has the exposure slider on the bottom left, etc.

Hope that helps out.

sundialsvc4
07-28-2010, 07:04 PM
That's a property of the capture device so should be handled in compositing, not by modifying your 3d scene.
I agree. This is a "difference" that would be manifest (and, manifested differently) if I chose to shoot the scene with Fujichrome Velvia vs. Sensia. But it would not actually be a difference in "the scene, itself."

So, you would want to replicate such effects "when you are re-applying Gamma, etc. to the final output, in the final processing stage." This is the point when you are seeking to replicate the behavior of physical film-stock. At every point prior to that one, it would be most undesirable. ("We would have to be dealing with that sort of thing constantly if this were 'real film,' but, by Gawd, this is not 'real film!' So, let us eat, drink, and be merry until then.")

(And, when "that time came," we would be taking full advantage of the fact that we had done everything that Salki just suggested ... thereby giving ourselves "perfect, un-biased, 'digital film'" to play with.)

CGTalk Moderation
07-28-2010, 07:04 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.