[argyllcms] Re: FWA compensation.

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Tue, 11 Mar 2008 12:32:27 +1100

Roger Breton wrote:

Compare the spectra with and without the UV-cut filter mounted on the
instrument : the two superimposed spectra tell the story?


To apply your heuristic, I suppose you need two datasets with spectral data,
one measured with the UV-cut filter and one without.

No. The whole point of the heuristic is to avoid having to do this, since
few instruments provide a programmatic means of measuring with and
without UV (ie., it is generally a manual process).

My only interrogation, Graeme, is do you end up calculating some kind of
global correction or do you apply it 'selectively' according to the hue
angle or chroma?

It's global. Naturally it doesn't change the color much when the FWA
is having little effect.

I always believe the theory of Martin Open about matching the level of OB in
the press paper to the level of OB on proofing paper, but I can tell you,
for having experimented with this, that it does not work. It works better
for some images than proofing to the GMG media but it's not the panacea.

I would imagine it's hard to get a good match for both FWA/OB and
the white point of the paper at the same time. In my experience the latter
was extremely critical - a delta E of 0.5 is plainly visible in media color
difference. Trying to match paper color by laying down tints using absolute
colorimetric is quite difficult, both because it's hard to get right 
and profiles can't do much better than 1 delta E, and you need 0.5
or better), and because it's hard to find a proofing paper that is
sufficiently white that it can actually tinted to the target paper color.

I have this because, either way, it's a fudge. Unfortunately, in the
presence of OB, colorimetry cannot help me.

It can if FWA is accounted for, but the biggest drawback is that
colorimetry is only practical if the viewing conditions are
known and controlled. Too often in practice this is not the case.

Graeme Gill.

Other related posts: