[argyllcms] Re: Characterizing a scanner using Argyll

  • From: Gerhard Fuernkranz <nospam456@xxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 02 Feb 2009 22:07:52 +0100

Asman, Andrew J wrote:
> ... but because of the fact that I want to analyze printed output, as opposed 
> to photos, I found that using an Argyll-generated target (using targen and 
> printtarg) produces better results because the target matches the type of 
> media that I want to  measure.

Of course, for optimal results, the prints from different printer
models, or with different inks, or different kind of halftoning, etc.
would need to be treated as different media types as well, i.e. you
should profile your scanner individually for each
printer/paper/ink/print-mode/etc. combination you want to measure with
the scanner. Then you'd also no longer have a gamut issue, since the
calibration target printed on each printer could always cover the full
gamut of this printer.

Also keep in mind that CMYK is a 4-dimensional space, while a scanner
can only capture 3-dimensional readings, and typically there is
significant observer metamerism between scanners and the human vision.
This implies that there will exist pairs of CMYK tuples resulting in
identical scanner RGB numbers, although they don't have the same XYZ
color, or vice versa there will exist  pairs of CMYK tuples which do
have the same XYZ color, albeit the scanner will nevertheless see
different RGB numbers. Only for a subset of the full 4-dimensional CMYK
space it may be possible to establish a unique, bijective RGB <-> XYZ
mapping, and the colors of all other CMYK tuples (not belonging to this
subset) can only be measured by the scanner with more or less systematic
error (the same applies to scanning media types different from the media
type used to create the profile).

> Additionally, I usually modify the color values on the patches to include 
> significantly more pure colors because I am predominantly measuring a 
> printers ability to uniformly print pure colors (CMYK) on a page.

Due to the reason mentioned above, if the use case is the scanning of
only pure C, M, Y or K ramps, then IMO the target should possibly better
not cover the full 4-dimensional CMYK space, but only a selected subset
the full 4D CMYK space for which a bijective RGB <-> XYZ mapping can be
established  (the desired C, M, Y and K ramps you intend to scan should
of course be part of this subset).


Other related posts: