[Openicc] linearisation targets, Argyll, G7

[Graeme Gill]

Jan-Peter Homann wrote:


> 2) linerization targets for a shared visual appearance
> --------------
> This step optimizes the 1D LUTS in case of CMY-gradation and CMY gray 
> balance and also for K-gradation
> This can be either done by spectral measurements but also by comparing 
> printed samples with reference samples.

Sounds tricky on a print device. Easy enough to do with a display,
as the device can be interactively solved for the RGB that
returns grey. Harder with a print device, since it's high
overhead to be feeing charts into it an measuring them
over and over. If you're going to do it in one pass by
modelling the behaviour, then I'm not quite clear on
what the motivation for such calibration is. The normal
profiling used on top of calibration is a very similar
process, and can then produce grey in the normal fashion
from the PCS.
[ie. this type of calibration seems to be very narrowly
  useful for setting up a CMYK printing press to print
  separated jobs, rather than something needed for
  an electronic printing device.]

> 3) Inklimiting for the dark areas
> ----------------------
> This is applied after 1D LUT linearization and individual channel 
> limits. It can be either done by some generic GCR algorithm or by the 
> standard printer profile.

Generic GCR algorithms give very poor gamuts in the dark saturated
regions (ie. K = min(CMY) etc.)

> for papers. For some paper ink combinations, it is necessary to limit 
> the total amount of Color (TAC) to 200% or even lower. In this case, the 
> usual method is some generic GCR BEFORE the profiling target is printed. 
> The printer profile itself will than be calculated with a higher TAC.

Unnecessary with Argyll, since both the test chart and profile
can have ink limits down to 0 imposed on them (although naturally
the gamut suffers badly at TAC's below 200).

> 4) Re-Linearization or calibration
> -------------------------
> This steps brings defined ink /paper combination in a state to mimic a 
> given reference.
> This re-linearization could be done based on 1D LUTs for every channel 
> and also via a Device-Link as 4D LUT.
> High-end proofing solutions are using often a DeviceLink-based 
> calibration based on charts between 300 and 600 patches.

I think this is outside the scope of print drivers for
electronic printers as well, and is very much in the
realm of CMYK proofing, or press re-targeting.

> 5) linerization files in CGATS / ISO 12640 format
> --------------------------
> The CGATS-format  or ISO 12640 is an OS independent ASCII based format 
> for color-data. It could easily adapted also to transport 1D LUT data.
> This would allow e.g. to write export and import filter to manipulate 
> printer linerization-data with  imaging applications, which can handle 
> CMYK images and have a curve tool.  including a softproof of the 
> manipularions.

Argyll currently uses the CGATS format for the display calibration
curves. The main attraction about using (say) an ICC device link
as a format for calibration curves is that they can then be linked
into a color transformation link path using current tools.

> 6) Open Source Creation of reference print samples for visual 
> (re-)linerization
> -----------------------
> The Epson Photo 2400 with Ultrachrome K3 Inks delivers very stable 
> output results and is supported by GutentPrint. On a linearized and 
> profiled device (maximum GCR), it is possible to produce print sample 
> with defined L*a*b* values. This could be e.g. neutral gray samples with 
> L* 80, L*50. L*30 and L*20. printed as square with 7x7 and a cut out 
> hole in the middle. The reference samples are compared with 
> (re-)linerization prints for K- and CMY-scales. The user reads the 
> number of the patch, which matches best the reference sample and types 
> it into the "(re-)linerization wizard". This application calculates the 
> new linerization curve for the current device/paper/ink combination.

I gather this is a means to enable calibration without an instrument ?
Interesting idea.

Graeme Gill.