[argyllcms] Re: Creating a camera profile with ColorChecker Passport Please Help
- From: Ben Goren <ben@xxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 12 Jun 2015 10:33:28 -0700
First, I think you could benefit from a bit of background theory. The following
link is excessively long and there're some bits in there (such as how to evenly
light a target) that should get edited out and I'm working right now on a much
superior method of generating profiles and and and...but the basic theory is
sound.
http://trumpetpower.com/photos/Exposure
To address your goals:
On Jun 12, 2015, at 9:51 AM, Brett Howard <brett@xxxxxxxxxxxxxxxx> wrote:
I'd really like to be able to create a profile from every shoot so that
things like light reflecting off a field of grass, or a slight cast to a
softbox front or something of that nature can be accounted for and calibrated
out on a shoot by shoot basis. I'm hoping to be able to use it much like
people do automatically with the light room plugins but in a slightly less
automated way (mainly because I'm a Linux user and that is the way we have to
do things). Am I off base here on expecting to be able to do this?
Yes...and no.
First, if you're doing, for example, fine art reproduction copy work, you're
not going to be doing it in a field of grass casting green tints onto the
artwork. And, if you <i>are</i> out in a field of grass with your camera, if
you edit out the green tint the grass casts onto its surroundings, things will
look weird and not at all like you're shooting in a field of grass. The grass
itself won't be green enough, too. If you don't want a green tint, don't shoot
on grass; if you're shooting on grass, embrace the green tint the same way a
painter would (and that's one of the first things painters learn). Or, of
course, do what photographers have done since forever: modify the light to your
tastes. An assistant with a mylar reflector (and automobile windshield
sunshades are excellent cheap options) can blast more than enough reflected
sunlight onto your subject to overwhelm the green tint from the grass, and also
even out and shape the light...the Sun behind the subject's shoulder, the
reflector roughly opposite the Sun, which also gives you a darker and more
dramatic and saturated background....
What I'm working on now is a spectral simulation of a camera's sensor. When I'm
done, I'll be able to predict the RGB values for any given spectral input. At
least at first, I'll be using that to create dummy .ti3 files to feed to Argyll
and then use Argyll as normal to create profiles. Perhaps at some point in the
future -- and certainly not before I've finished my own homework -- I might be
able to twist Graeme's arm into incorporating something similar directly into
Argyll, but that's looooooong down the road, if he's even interested in that
sort of thing.
Cameras are, naturally, extremely sensitive to the actual illuminant in the
scene. "It's all about the light," as photographers and painters before them
have been saying since the stone age at the least. If you photograph a chart
and build a profile from it, you'll "bake" the illuminant used for the
photograph into the chart. Subsequent photos in that same illuminant will have
their colors mapped to appear as if you had shot the scene not with the actual
illuminant but instead with D50 lamps. This is what you want for copy work.
But, if you use that profile to shoot under some other illuminant, you wind up
with an useless multiplication of the original illuminant and the new
illuminant such that the color rendition is just plain totally worng. Yet,
alas, this is how pretty much every camera profile ever created has been made.
(With a notable exception for Iliah Borg's profiles.)
If you have a model of the camera's spectral response (what I'm working on),
you can build a profile that renders the scene with the actual XYZ values as
they reached the camera. This means that, for example, if the illuminant is a
purple bulb, a styrofoam cup will get rendered as the same hue of purple as the
illuminant. I need to do some experimentation, but I suspect that such an
universal profile may well be desirable for any sort of "mood" or "atmospheric"
photography, especially landscapes in the golden hour. Also, starting with such
a profile but doing a simple white balance to map something spectrally flat in
the scene to R=G=B in the file is likely to be desirable for most any other
types of general-purpose photography. And, lastly, if you know the spectrum of
the scene's illuminant (typically easy to determine or guess; there's not that
much variation -- and, of course, it's something you can measure with a
spectrometer), you can use the camera's spectral response to build a profile
with the scene illuminant that maps it back to D50; that's what you'd do for
copy work -- and product and fashion photographers would also likely appreciate
such a workflow.
...but, again, this is all very much a work in progress at my end...and, in
fact, I need to head out right now to go get some more foamcore to build the
next iteration of the spectroscope....
b&
Attachment:
signature.asc
Description: Message signed with OpenPGP using GPGMail
Other related posts:
