[argyllcms] Re: Targen & chart gamma correction

  • From: Nikolay Pokhilchenko <nikolay_po@xxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 30 Nov 2009 13:52:22 +0300

Graeme Gill wrote:

> Nikolay Pokhilchenko wrote:
> > Okay, it's will be useful. But as I understood, the device values for 
> > spacers and
> > preliminary device<->XYZ dependency in *.ti1 will be the same. In this case 
> > the problem
> > with spacers can still persist. I ask gamma for whole the device space, 
> > even for -f
> > patch generator (at least at first step before preliminary profile).
> I'll see what I can do, although it will be of limited use for anything other
> than calibration (ie. such a highly non-linear device will not profile well).
> Graeme Gill.

Thank You, Graeme! I'm agree that gamma will be useful mainly for calibration. 
But it will be usefully for profiling without calibration, at least for 
preliminary profile.
By my experience, after 4 or more iterations (~3200-3500 patches in summary) 
with perceptual grid spreading (-I key in targen) even highly non-linear 
devices get profiled well with Argyll v1.0.4 with max dE between 3.2 and 5.2. I 
say again, that (IMHO) perceptual and curvature patch spreading optimization is 
needed, but the device optimization for main chart - not. If the device is 
described well in perceptual space and at the curves bends, the device 
spreading isn't needed. If there is any issues with black generation and device 
patches spreading? IMHO for preliminary the best optimization will be:

for preliminary - 100%-device (with user-predicted gamma), 0%-perceptual, 

for main profile - 10%(or less?)-device, 45%-perceptual, 45%-curvature.

My experience with non-linear printers and V1.0.4 says that in the cases the 
device space spreading is useless. With non-linear printers I'm selecting the 
needed black generation by preliminary profile, then rebuilding the preliminary 
profile with desired black generation, then generating by "targen -I" the main 
profile charts. At the 3-th..4-th step with perceptual spreading the black 
generation curve becomes smooth and the profile becomes quite exact.

Other related posts: