Digital BW, The Print

Like Tyler, I can't get my mind around the math stuff, but I've got a 
good logical mind and a good understanding of ICC color management.

I think I'll throw some thoughts out between the quote below.

-bruce


On Wednesday, Dec 8, 2004, at 20:36 US/Pacific, 
DigitalBlackandWhiteThePrint@yahoogroups.com wrote:

> Message: 4
>    Date: Wed, 08 Dec 2004 23:49:16 +0000
>    From: Steve Kale <stevekale@...>
> Subject: Re: Re: Tonal range and linearization
>
>
> To keep a sense of perspective in this conversation, I ask some basic
> questions:
>
> 1.  Do we agree that the need to "colour manage" (broadly defined) 
> results
> because two devices don't have the exact same colour space?  (In this 
> case,
> our image file on the computer and our printer)

Kind of. The most important aspect for both color and B&W is that the 
the display image and underlying data be standardized, or at least 
known and tagged as such. Photoshop takes care of that by having us 
work in standard (or at least defined) color and greyscale workspaces. 
The big trick being that a measured and profiled monitor (tagged) is 
used to display the image after conversion from the work space. 
Ensuring that what we see is accurate
>
> 2.  Do we agree that if our printer and our computer did share the same
> colour space (or at least all the parts we care about) this whole
> conversation would be redundant?

Yes.
>
> 3. Do we agree that we have abandoned an automated method for managing 
> the
> difference between the two colour spaces (ie something like colorsync)?

Not quite.

The issue with b&w is simply that the paper is darker than monitor 
white and the black (especially on matte paper) is much lighter than 
monitor black.

We don't need to "automate" this, as we already have the qtr curve set 
for a particular output. The question is what rules were used to create 
the curve.

In color, icc conversions have rules. I think that "relative 
colormetric" has a specific standardized rule to map out of gamut 
colors to their nearest neighbor. Simple enough, but this method often 
results in posterization when the printer space is way different (and 
smaller) than the working space.

So they came up with "perceptual intent" which uses some "voo doo" to 
make the print "feel" more like the original, just using the colors 
available in the printer.  This print will be much less accurate, but 
more pleasing. I don't think there is a standard "rule" for "perceptual 
intent". These rules are designed by the programer of the profiling 
software. Each profiling package will produce a different "perceptual 
intent" where the rules have been made by different programers. What 
these rules are, are compromises to make a pleasing print (in someone's 
opinion) when the printer can't reproduce what's in the original file. 
What I've found is that this works well for many images, and lousy for 
others. Sometimes the compromise is so severe where the printer just 
can't make a good print (with a particular paper and ink) that I've 
converted color photos to black and white just to get good tonality in 
the print. An example being a dark red brick building that my 6 ink 
epson can not reproduce in color without loosing all red in the darker 
bricks. In b&w I can see all the bricks in the same color!

I guess my point is that "automated" is in the eye of the beholder. 
Either the programmer of the profiling package, or the creator of the 
qtr curve, or Roark curve has a say. It's not just an automated box 
with scientific rules.

And many profiles are "edited" by eye after creation. Should the rules 
be sufficient, this would be uncommon, but it's not at all.
>
> 4.  Do we agree that if our printer and our computer have overlapping 
> colour
> spaces that we would just have to decide what to do with the points 
> that
> didn't overlap and that such decision may or may not mean we decide to 
> alter
> the entire image to maintain a sense of relativity?

One cannot simply leave all in gamut shades as is, and change only 
those out of gamut shades. In b&w this would just result in clipping 
the blackest and whitest tones because bringing only out of gamut 
shades into gamut would mean making them the same tone as tones already 
used by other tonal values. Making black the same tone as dark grey is 
the same as clipping as the change in density on the print would be 
invisible. It does make sense to change the black and out of gamut 
tones to in gamut tones as long as one changes ALL the tones in the 
print.
>
> 5.  Do we agree that more overlap is better than less?
>
>
> I believe there is no reason for our workspace and printer not to 
> share the
> exact same "density" (which I will define as colour less hue) space 
> for much
> of the workspace range.  Specifically, for example, 16<=L=>96 for 
> printing
> to EEM and around 6<=L=>96 for printing on EPSG.

#5 makes no sense to me. They are not close to the same space. A 
partial overlap is irrelevant.
>
> At the moment they share no common ground.  A simple image that is well
> within the capability of the printer will not print properly out of 
> the box
> - instead it requires a _visual_ soft proof aid to get it to look 
> right and
> even then it may not print right.

I think the visual soft proof aid is essential here. It is essential 
for color work where one uses a colorsync (ICC) workflow as well. I 
think the simple answer is that most images are not within the 
capability of the printer, and must be altered for printing acceptably. 
I think that your frustration Steve is that you are trying to make a 
matte watercolor paper print look like a glossy print. No conversion of 
data can make this so.
>
> If they did share the good proportion of common ground they are 
> capable,
> deciding what to do about of gamut values would be easy and very 
> intuitive.
> Because they don't it is complex and not obvious, requiring the use of 
> a
> soft proof and "fiddling to match".

By compressing all the working space gamut into the printer gamut using 
a QTR curve, or other method, we have essentially created a b&w 
conversion using our own "perceptual intent". The soft proof is the 
best, and only way, I can think of to make a visual match from screen 
to print. In fact it solves the whole "print matching" issue provided 
that the ink separation curves are working to separate all the tones 
(within reason).
>
> This is all I am pressing on.
>
Of course if anyone out there makes a great discovery, I'm all for it!