[argyllcms] Re: camera profiling with newest version of Argyll

  • From: Ben Goren <ben@xxxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 27 Aug 2015 17:27:22 -0700

On Aug 27, 2015, at 5:01 PM, Chad Johnson <chad@xxxxxxxxxxxxx> wrote:

Naw, D50 just isn't a good approximation for kinoflo lights.

Erm...methinks you're working under some misunderstandings.

D50 doesn't exist in the real world. It's a theoretical construct used in the
math of representing color and transforming from one representation to another.

For ideal results, you'd use a perfect D50 simulator when photographing
something -- but, also, your camera would have RGB filters that perfectly
matched the spectral response of your preferred standard observer. If you could
do that, profiling would be moot; the RGB file that came out of the camera
would already be in CIE RGB space. And you would then look at the picture on a
display whose primaries were again a perfect match for the standard observer,
and the RGB values would simply get directly and linearly translated into
voltages, and the match would be perfect. But, obviously, such perfection
doesn't actually exist.

The purpose of profiling is to take whatever the actual conditions are and map
them, as best as is possible, to what an ideal observer in ideal D50 light
would see.

And, as it turns out...the math doesn't even care all that much about the
illuminant you're using, so long as it's not excessively spiky and you've got
enough of it to avoid noise. With a correctly-built profile, you can do
dead-accurate colorimetric photography under blackbody incandescent lighting
just as well as under a D50 simulator -- but only with a correctly-built
profile. The profile will do the math to convert the RGB values coming out of
the camera (with its decidedly non-standard observer) to the RGB values you
would have gotten with perfect light and a perfect observer. (With, of course,
all the standard caveats about accuracy, noise, metameric mismatches, and all
the rest.)

The munki has no UV sensitivity

The file you attached shows readings of 86% reflectivity for 380nm -
420nm...not UV, but far enough into the violet and bright enough to identify
the effects of FWA on paper stock. Now, it could certainly be that those
readings are not valid...but that that just brings me to my final point:

, but then fluorescent lights don't emit in the UV range, so I don't think
that should be an issue either.

Regardless, the XYZ and spectral values in the .ti3 file you posted are not
what is expected that one should obtain from the chart. Something's amiss, and
that something will have to be fixed before you can expect good results from a
profile built with that reference file. Whether your chart is bad, or your
ColorMunki is bad, or you're not using the ColorMunki correctly...something's
broken and needs to be fixed.


Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

Other related posts: