Is this a new idea? (cool PBR workflow trick).
interpolator
While gathering texture references for my environment, I was reminded of something related to PBR.
When taking photos for your albedo, you want no light information like shadows or specular in it. So this means the best photographs are taken in a diffuse or overcast like situation.
But I thought "what about doing the reverse"? Using a camera to take flash photographs of different materials in a dark setting.
So I turned off all the lights, and got really close and started taking pictures.
So why is this important? Roughness. Under normal lighting conditions, it's actually harder to see all the micro surface details that you normally take for granted in texturing. Also, because there's only one lightsource and it's directly in front of you, you have greater control over how and what kind of specular reflections show up in your material.
When taking photos for your albedo, you want no light information like shadows or specular in it. So this means the best photographs are taken in a diffuse or overcast like situation.
But I thought "what about doing the reverse"? Using a camera to take flash photographs of different materials in a dark setting.
So I turned off all the lights, and got really close and started taking pictures.
So why is this important? Roughness. Under normal lighting conditions, it's actually harder to see all the micro surface details that you normally take for granted in texturing. Also, because there's only one lightsource and it's directly in front of you, you have greater control over how and what kind of specular reflections show up in your material.
Replies
But this could be used when google images can't always find the same result. For example, if I had a unique toy that you can't find on the internet, and it's in a condition that clearly been used instead of brand new (i.e fingerprints, smudge marks, dents etc).
I think for most searches, you wouldn't get high detail photographs of everday items with excessive wear, unless it's something high profile like a Car.
A lot of times people make textures based on photo reference taken with flash, and they totally mess up the specular intensity. This sort of head on flash always makes objects appear to be more reflective than they actually are, because the light from a flash is sooo much brighter than the ambient light.
But it can be a quick way to get a rough idea of the surface variation/roughness.
I'm not sure if the latter is a more realistic.
I could also try playing with the actual exposure settings to see if that makes a difference as well.
Edit: I went back to reshoot some more pics.
I found out the farther away the camera is to the object, the less intense the highlights/reflections are.
http://udn.epicgames.com/Three/TakingBetterPhotosForTextures.html
Edit: The flash pic was taken before I learned about the specular thing. It's also taken head-on whereas it would be better to show the roughness/reflections at an angle or something.
i honestly think you could generate more comprehensively useful reference if you put the object you want on a table, point a desk light at it, walk like 10 feet away and shoot it with a bit longer lens so it still fills the frame but doesn't have one tiny superbright spot on it
Just for reference, thats untrue.
Best diffuse photographs for non pbr diffuse are best in overcast because you don't have directional light mucking it up.
But you still get lighting information. The actual best way to get photos is using cross polarization, like oglu said.
Your pictures are useful for studying how light affects a surface..but that doesn't necessarily mean its roughness. The top image, not sure what I'm looking at, but some of the variation on that surface is from a buildup of something on that metal...so its not neccsarily roughness info...and the flat wall is more just showing what would be normal detail....and is pretty uninformative for roughness specifically. The bottle is a cool example though.
Point is...don't mistake lighting response for being just a single thing. You are merely observing how light affects a surface. Its an interesting way to analyze a surface, either way.