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

  • From: Chad Johnson <chad.chadjohnson@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 27 Aug 2015 19:20:53 -0400

How are you converting from XYZ to RGB? That depends upon the illuminant, no? If I assume D65 (which is probably reasonable for these lights), I get RGB values that make sense (here I'm using the calculator from easyrgb): 100,95,92.

Just to be clear I am not setting the WB from that patch. Or any patch for that matter -- it's set from shooting a gray card and then I check it against one of the grey patches on the Color Checker.


On 8/27/15 7:00 PM, Ben Goren wrote:

On Aug 27, 2015, at 9:35 AM, Chad Johnson <chad.chadjohnson@xxxxxxxxx> wrote:

This file was created from the scanin command. See attached. So this one should
have the camera's WB data and all that.

Assuming no more crossed wires, that's the .ti3 file that Argyll would
presumably be using for building the profile, and the one that Graeme will want
to use for his investigations.

But all the other problems remain. In particular, the RGB values in this new
.ti3 are (on a scale from 0 - 100): 88.617 88.800 88.979

According to the XYZ values in the same .ti3, were the photo properly exposed
and white balanced, the right values would be something closer to: 86.025
92.920 111.800 (assuming CIE RGB primaries and a 1.0 gamma).

You can tell two things right off the bat.

First, your photo is slightly underexposed, though not by an awful lot. However, that
underexposure will get "baked into" the profile. If the photos you make with
the profile are similarly underexposed, that won't be such a problem.

But the bigger problem...is again with the chart and your use of that patch for white
balance. That fluorescence requires an RGB value with a blue component "10% more
blue than blue." None of the standard color spaces I checked could encode it without
saturating...and I'm not even sure if the geometry would work out for any ICC-permitted
primaries to encompass it.

Or, in other words: not only is it not white; not only is it very unsuited to
use as a white balance reference; it's not even possible to accurately
photograph in the first place (unless you go far away from standard
photographic workflows). And that's why I suspect a counterfeit...I can't
imagine something like this getting past the superlative color scientists they
have at X-Rite -- they know better than to design something like this.

I appreciate that you're happy with what you've done in the past and just want
to do the same thing with the newer version of Argyll...but it seems to me that
there're problems in your workflow, and it's entirely possible that the fixes
and enhancements Graeme has made to Argyll in the long time since version 1.5
are now exposing those flaws.

Regardless, I would very strongly recommend starting with a good workflow that
did not include the chart you're using (and, similarly, a less spiky light
source) and go from there.

There are always edge cases and bugs to uncover...but Graeme's polished the
hell out of Argyll to the point that the limiting factors these days are almost
always in workflow and equipment...and, I know you don't want to hear it, but
that's where I'd start given your case....


Other related posts: