I’m no asset creation specialist, but from what I’ve heard it’s not that difficult to get textures, although making a decent normal map that actually corresponds to a texture’s source is tricky. Some boffins at Dolby thought there had to be an easier way than what was available, so they got to thinking.

What they came up with is essentially taking a surface’s picture twice: once with ambient lighting that highlights bumps and cracks, and once with a direct flash that allows the camera to see all the color. A system then compares the two, determines whether a light or dark area signifies a bump, divot, crack, or just color gradient. It’s pretty smart, and hopefully it works.