QTR-Quadtone RIP

--- In QuadtoneRIP@yahoogroups.com, MGochnauer <goch@u...> wrote:
>
> > PP is a scanner based color printer profiling system, and even the
> > best of those is not worth bothering with.
> 
> Can you explain *why* this is so? The basic approach of PP *appears*  
> to me to be sound: print a defined or known source file with no  
> printer corrections at all, scan it at the same time as a target of  
> known colours and densities, once again with no corrections at all,  
> and then create a profile based on comparing the printed target with  
> the standard target.

Because of something called "observer metamerism". You're probably
familiar with "illuminant metamerism", where a color appears different
under different lights. (more properly, two colors that appear the
same under one light appeaer different under another).

The opposite of that is "observer metamerism". Two colors that appear
to be identical to one observer appear to be quite different to
another observer. Now, imagine one of those observers is your eye, and
the other is your scanner. They're nowhere near close in color
perception: two colors that look identical to your eye may look very
different to the scanner, and two colors that look identical to the
scanner may look very different to your eye. So, if you create a
profile with the scanner, and it works out all the math to make the
colors on the patch you printed out look identical to the patches in
the reference card that came with Profile Prism, you can still have a
very wrong profile, because the colors that look identical to the
scanner look totally different to your eye.

The reason why spectrophotometers such as the eye-one break color up
into 16 or 31 different bands, then do a whole bunch of math on those
bands, is to get color numbers that match the characteristics of a
human eye, so there will be no observer metamerism.

Now, it's possible to get lucky with a scanner based solution, and
it's possible to tweek it until it works well under a relatively small
set of conditions. i.e. if the Profile Prism reference card uses inks
that are similar to a particular printer's ink, then if you calibrate
the printer to match those colors when observed by the scanner, they
will look similar when observed by the eye. But any ink that deviates
from that match will make a poor profile. Or any paper that differs
from the paper used in the reference, as far as fluorescence (OBAs,
optical brightening addatives) will also cause a poor profile.

> The resulting profiles are unique to the  
> equipment, of course, but why would they be inherently faulty?  [I  
> just bought PP!!]

Are you getting results that you like? That's the big issue. There's
two kinds of profiles, just like there's two kinds of color:
"accurate" and "pretty". A scanner based profiler can make a "pretty"
profile. It can fix problems with shadow details blocking up, or color
crossovers. It may leave you with a profile where all grays, whether
light or dark, are too magenta, but it shouldn't leave you with one
where light grays are green, while dark grays are magenta.

If your income depends on the client seeing that you've accuratly
duplicated the color of their corporate logo, or the model being
satisfied that you've got her makeup color right, or the bride being
happy about the color of her dress, "pretty" profiles don't cut it,
and you end up in the world of "accurate" profiles.

> > ... If I were not capable of writing a tool to measure
> > grayscale densities using a scanner, I would consider buying a lower
> > priced tool intended specifically for that purpose. But not something
> > carrying color profiling baggage.
> 
> Can you suggest such a "tool to measure grayscale densities using a  
> scanner" for those of us who have insufficient time or talent to do  
> it for ourselves. Is there anything out there you would recommend?  
> Most of us have scanners, and a decent "software densitometer" would  
> be helpful.

There is one, and I'm having the devil of a time remember ing the
name. I want to say VueScan, but that doesn't feel right. I do recall
that it did need periodic calibration with a reference chart that had
several known density squares. Not too hard to make.