Hi,
Say I want to extract base color (for PBR texturing, using metallic approach) from a photo, where materials are clearly affected by lighting. How to do so more accurately?
I am not certain how this problem is solved commonly. Through experience? Through better reference photos? Is there a methodology I can use to achieve better result?
Replies
that would be my approach at least if I was being pedantic.
otherwise I would just say trust your artistic eye, keep the levels and exposure consistent and do comparisons as you adjust the color.
I was told a few times to not "eyeball estimate things" on this forum, glad to hear at least in some instances I can trust my eyes
I think if you can not research it, build the scene, match the lighting and then start working on the materials to match the reference.
The color checker balances the photo to a more "true" colour hue. The spheres help to determine light direction, intensity, GI and HDR image.
They then recreated the lighting conditions in a 3D program and baked maps that cancelled out the natural lighting conditions of the environment, then an artist would do touch-up. So even a procedure like this requires some artist eye.
The only way to actually get the base color is to use a material scanner (extremely controlled environment) which.. are expensive and rare.
So most of the time - you are guessing. Which is why it's "3D artist".
No object in life is naturally brown, blue, purple etc. What we're seeing are rays of light hit an object and then it bounces back into our eyes. It's why if you took out a purple lightbulb and shine it on a white piece of paper, the paper's texture didn't become purple. It simply shined back those [purple] rays.
What the PBR model did was standardize this process so the textures should look scientifically accurate in all lighting scenarios. Like how only Coal material has a really dark albedo, because black absorbs a lot of light, whereas White Paint is going to be more lighter, because white reflects light.
This process is called white balancing.
Edit: this is useful too:
https://en.wikipedia.org/wiki/White_point#White_point_conversion