[argyllcms] Re: Gamut of output profiles

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 14 Sep 2007 00:32:08 +1000

Lars Tore Gustavsen wrote:
I have played a lot with the different gamut tools from argyll recently.
I was little shocked when found out that the printer profile I had
with largest gamut for my Epson R2400, was one produced with eye-one
match "easy rgb 1.2" chart. That is a small 45 patch chart.

Output from iccgam -v is
Total volume of gamut is 764113.519232 cubic colorspace units

The one produced by Epson for the same Premium Luster paper is

A typical 1000 patch rgb profile that I have created myself is

And if I process the "easy rgb" target in Argyll I end up with

And a similar sized target (45 patches) created in Argyll gives me

Question is :
Can I trust these values?

The volume calculated depends on a number of factors, but they are
reasonable given the way they are derived. Factors include:

1) Which intent (table) you chose. The most literal gamut volume
   is the absolute colorimetric one (ie. iccgamut -ff -ia).

2) The colorspace. The default is D50 L*a*b*, but CIECAM02 Jab can
   also be chosen.

3) The accuracy of the profile. This depends on the instrument, the
   instrument setup (ie. illuminant, observer, UV filter etc.) and
   the way the profile was created. A profile created from few
   test points is likely to be less accurate, since there is lots
   of space between the point to interpolate.

4) The ink limit. This may not be applicable for an RGB driven printer
   (although it will appear indirectly as an effect of the drivers
   paper type setting), but has a very direct bearing for CMYK devices.

5) The exact printing conditions. Maybe the manufacturer had a different
   batch of paper, ink, or the printer they used was in a slightly
   different condition (ie. production batch, wear etc.).

6) Whether the profiling package applies some sort of gamut mapping
   in the A2B tables. Profilers are not meant to do this for the
   colorimetric tables, but who knows ?

7) The algorithm used to interpolate and smooth the test points.
   Argyll uses one that balances fit and smoothness. Others may
   fit exactly, without regard to smoothness, resulting in a different
   profile and gamut.

I have made a wrml file showing the easy target processed in eye-one
match against the same measurement file processed in Argyll here.
Argyll is the color one.

This looks like a maximum density issue. If the two profiles are derived
from the same source points, and it is the colorimetric A2B table we're
looking at, how do the respective fit errors compare (ie. profcheck output) ?

Graeme Gill.

Other related posts: