[Openicc] Drop size calibration

[Robert Krawitz]

   Date: Tue, 29 Jan 2008 12:45:23 +1100
   From: Graeme Gill <graeme at argyllcms.com>

   Robert Krawitz wrote:

   > I think you're right here.  We'll need to figure out error
   > diffusion/EvenTone type algorithms that use three drop sizes also.

   My experience was that it was more difficult to get dot size
   overlap with error diffusion. I managed it, but it wasn't as smooth
   and controllable as a stochastic threshold array.  At the end of
   the day there was little to choose between these two approaches,
   with the stochastic approach being faster and slightly smoother,
   and the error diffusion being better for dot proofing where you are
   trying to reproduce microscopic detail.

Interesting.  I agree that EvenTone does better for fine detail (such
as the 1 degree circle in the CUPS test page), but it also tends to
produce slightly smoother output than the threshold array that we use.

Thomas Tonino, who created our arrays, isn't active any longer on
Gutenprint, but the arrays he came up with generally produce quite
good output.

I need to find some time to investigate Raph Levien's EvenBetter
Screening.  I'm not certain, but this may be multi-level as well as
well as handling multiple inks.  The code's a bit convoluted because
it's vector-optimized (and because it's Ghostscript-based -- I always
find Ghostscript code harder to figure out).

   >    Sufficient overlap should smooth some of the worst of this
   >    out, since it's the near 100% fill that has the highest dot
   >    gain, and calibration should be capable of taking care of the
   >    rest.
   > 
   > Yup, 50% overlap seems to work well.

   Maybe this depends on how randomized the screen is. I don't recall
   using such a great overlap, and still got a perfectly acceptable
   raw transfer curve to use with calibration.

It probably would work to use less, but the particular combination I
was testing is one of the ones that may be badly tuned.  If people
would like, I could make it a parameter, but it doesn't sound like
that's really necessary.

50% seemed like a good number to me.  At 50%, the drops are all at
sqrt(2) distance from each other.  Above that, they get closer
together, so overlap gets significantly more common.

-- 
Robert Krawitz                                     <rlk at alum.mit.edu>

Tall Clubs International  --  http://www.tall.org/ or 1-888-IM-TALL-2
Member of the League for Programming Freedom -- mail lpf at uunet.uu.net
Project lead for Gutenprint   --    http://gimp-print.sourceforge.net

"Linux doesn't dictate how I work, I dictate how Linux works."
--Eric Crampton