Just How Important Are Specular Maps For Photo-Real Faces? Cameras, filters etc
Hi Guys!
Basically i'm currently analysing the 'latest technologies that are available for creating human faces for computer games' which invlolves looking at facial scanners, procedural face programs and of course traditional methods using max/maya/zbrush etc.
At present I am trying to determine just how important specular maps are in the creation of a photo-real face. So far I have managed to get a bit of information on this subject but still lacking a lot! I was wondering if any of you guys could recommend any Threads/Websites/Books etc that would be of assistance to me?
Cheers
Ryan
Basically i'm currently analysing the 'latest technologies that are available for creating human faces for computer games' which invlolves looking at facial scanners, procedural face programs and of course traditional methods using max/maya/zbrush etc.
At present I am trying to determine just how important specular maps are in the creation of a photo-real face. So far I have managed to get a bit of information on this subject but still lacking a lot! I was wondering if any of you guys could recommend any Threads/Websites/Books etc that would be of assistance to me?
Cheers
Ryan
Replies
OK, seriously EVERY model needs some sort of spec/gloss map setup, how else are you supposed to define a material? You can't define them with a Normal or Diffuse since those only represent static information, Specular and Gloss are what determines the 'look' of a model, and they're reactive.
Im definitiely more interested in the content creation of specular maps and what they should look like.
I'm also trying to find documents that i can use as evidence as to why specular maps are so important when creating photo-real faces. (hope that makes sense)
cheers
You've got a face(I presume) and likely a lamp, mirror and/or camera. Pay attention to how your skin reacts to various lighting conditions, which parts of your face are more/less reflective(specular), and which parts are dull/glossy(glossiness).
I'm not entirely sure why you need documents to support this, it should be fairly obvious to anyone who has... seen other people before. =P
You can look at tutorials or other peoples work, but referencing art isn't ever a great idea, you're referencing something abstract that is referencing reality. Skip that and just reference reality directly.
Once you understand how a material reacts to light, the whys and hows of asset creation should be straight forward.
If you've got a DSLR, you can use a polorizing filter to seperate specular from diffuse in the real world, which can be very helpful to understand what is going on. Even holding a CPL filter over the lens of your point and shoot camera should work. http://filmicgames.com/archives/547
You can get a lot more advanced with sub-surface scattering, fresnel, etc, but spec/gloss should be fairly easy to observe.
Note that the lighting is the same for each left/right shot, only the CPL has been adjusted. Each set has different lighting though.
It doesn't really completely remove the specular from the "diffuse" shots, but you should get the idea.
Basically what's above my post, but perhaps showing a bit more clear example.
http://udn.epicgames.com/Three/TakingBetterPhotosForTextures.html
You can see its not perfect, but pretty darn good.
It's like you're asking to change the camera position on a photo after it has been taken...
oh well
EDIT: Is a linear or circular filter needed? I keep seeing info on both...
1) Use the shadow/highlight feature in Photoshop to normalize the values.
2) Bring the oils out... duplicate the layer, desaturate it, then open up the levels and adjust the settings so you bring out highs of the image.
3) Take that, make an alpha of it, and then use it to select the pixels on your original diffuse layer, then add a curves adjustment layer to that and bring it into range.
4) If you start to see a loss of detail from this or a loss of saturation, mess with the contrast, curves, and hue/saturation.
It does a pretty bang up job for a software solution. Results may vary, and it still requires a good clean shot. You just have to ask yourself if doing all that polarization work is worth it for the content you're creating.
PS: These images and steps were taken from "Character Texturing for Production" with Ben Neall on Gnomon Workshop. Link to product page found here: http://www.thegnomonworkshop.com/store/product/425/Character-Texturing-for-Production
gir: You can get a $10 CPL filter and hold it in front of a cheap camera, its not a major investment. You could also buy something like this: http://www.ebay.com/itm/filter-mount-lens-adapter-SAMSUNG-EX1-TL-500-58MM-/330687469189?pt=LH_DefaultDomain_0&hash=item4cfe825685 duct tape it to your camera and easily use filters.
You don't really need a high end DSLR to do this sort of stuff, a CPL, some bright lighting, shooting at low ISO(to avoid noise/loss of detail) and zoom in to a longer focal length on your camera to avoid the wide angle perspective distortion. That should do it.
Anyway, my example above wasn't really meant as a guide on how to make the content itself, more so to show a method to observe the differences, and see what the specular component is contributing to the end result.
Pretty much the same stuff that Stradigos mentioned, only much more easier for the lazy in us.
Also, don't forget to take pictures in RAW format, you will have extra access to certain tools if your application allows it.
I just bought nDo2 last night. I take it you're talking about the "Removing Shading" and "Remove Highlights" sliders on the diffuse tab in CrazyBump? Yeah, seems to do a decent job for only two controls. With nDo2 though, I would probably start with the process outlined above. I haven't really looked at it enough though to see if they have sliders to help automatically deal with this stuff. That would be pretty neat. I'll check it out later and report back if I find anything.
I understand that is what he is doing, my question is how you would do the opposite.
taken from http://gl.ict.usc.edu/Research/DigitalEmily/.
THAT LOOKS AMAZING.
EarthQuake, a massive thank you for your detailed explanation of how to understand the power of specular and how to seperate the specular from a diffuse with the use of a CPL. I'll definitely have to get my hands on one for my 550D.
Also Equil's photo looks amazing!
Thanks again for the assistance guys, it's much appreciated
How are you supposed to get the steps 3 and 4 to work? I don't how how to get the levels right to remove the highlights. I'm using 3d.sk images so I don't have access to fancy poloarization photography. I also use the desaturated image as an curve adjustemnet layer but it not taking out all of the specularity.
I guess we do, when we are under what looks like hundreds of lights (Reflections in her eyes).
I was wondering if anyone knew the names of filters one would use to filter out your standard Spec/Diffuse/Cavity setup in the images one takes with a camera?
I know that Polariztion Filters (Linear) is what you need to get a flat diffuse, and negate the Specular, but what would I need to do for the opposite, where I want to only 'see' and pull the Specular?
I found this 'formula' for the Specular, which nice and all, but I don't recall windows or photoshop having the ability if create and manage HLSL code:
http://filmicgames.com/archives/233
So yeah, any pretty help here peeps? Would be grand!
I don't think there are any filters that will do what a polarizer will do to get just spec or "cavity". You'll have to do something more complex.
On the other hand, a simple subtract from layers will totally make the object look inverted in color, as opposed to the site math, which gives it a more 'relaxed' hue. Mine currently look's completely 100% dialectic, which doesn't look the same in the site.
@EQ: Haha, thanks, but I kinda did already post and look up in that thread, problem is, the only 'reference' image is that of "Emily", and even then the math behind how to get Spec only is not explained.
The link I posted up has a HLSL method of 'retrieving' the Spec and the 'correct' color, but I don't know how it would be compiled or be made into an Action or PS.
Here is the text I'm referencing to:
I would like to know the difference too, in the article Zac linked too I think the linear Polarizer was referring to the film the author placed over the flash?
Some stuff which might help shed some light:
I have an instruction manual here next to me which is pretty big, but here is what is written in it:
So basically, same stuff, only used really for certain camera's that lack an automatic/or manual controllable exposure and which fall under a certain qualifier.
Another one that seems to be popular in the manual is this:
So yeah, been scouring the manual for over several hours and still no dice on anything only captures reflections, all of them seem to negate it.
If you wanted a home brew version, you need your camera with CPL, and polarized light sources. To get two images w/o your talent moving is the hard part. My thought was to get four flashes, and set them up in tandem... one w/o polarization and one with on each side. Then depending on how fast your camera syncs w/ flashes... maybe 1/250-1/400s try to pop off one shot with and without. I've seen custom flash rigs for sports photography, but I have no idea how to rig something like that up.
As far as the CPL/Linear polarizer goes... you need a CPL if you are using modern cameras (and you all will be) so your camera meters properly. You should be shooting everything on manual anyway, so everything is as consistent as possible. There is no filter that is going to give you straight cavity/specular/reflection... that's not how the world works. It's just like how polarized sunglasses work. This is all fairly new in production for film.
http://www.ted.com/talks/paul_debevec_animates_a_photo_real_digital_face.html