I was curious about the various workflows to get what you want on the paper.
Here's where I am: I'm on a Mac so the ICC approach is very convenient. So this is
what I use all the time these days. I mostly do photo paper these days but did the
same thing with matte for a long time.
With matte papers the dMax is relatively weak so its always been an issue of how to
map an image (on screen) with wider Dynamic Range (DR) onto a paper with smaller DR.
The matte paper loss of DR is almost entirely on the dark end so this is where the
mapping is most notable and potentially controversial. Anyway you slice it you
have to lose DR somewhere -- all over or more some places. The QTR linearization
from early days has always been a straightforward straight-line or basically DR loss is
spread out across the whole range. With this approach a DR of L=96 to L=16 has
a midpoint of L=56 which a fair bit lighter than a middle gray of L=50. (in actuality
it is even more complicated because editing spaces are all different so middle will
have lots of different values). But the bottom line is that most people noticed that
their prints were almost always lighter and weaker than they expected. (if you were
from the darkroom days you were used to just editing till you got what you wanted).
But the beauty of digital photography is reducing the trial and fix cycles. Screen to
print matching to me seems like the ultimate goal. Its never going to be perfect,
first a light emitting screen is never identical to a reflecting piece of paper, and
second a smaller DR paper will never be as contrasty as a high DR screen. You will
always need to use your experience to "see" what you will get and will need a minimum
trial/fix cycle. So lets do the best we can.
I think the workflows fall into two categories -- dumb down the screen somehow
so you edit in the reduced DR that matches the paper, or -- do a correction curve
at print time that gives the best you can do mapping (i.e. something you can
easily learn to pre-visualize). I prefer the later since the former ties the image
file to the specific paper -- what happens when you want a different paper?
The ICC folks have done lots of this with color so I figured it ought to apply to
grayscale too -- actually lots simpler because it's one dimension instead of three.
Learning how it all worked (for gray at least) was a bear because everyone just
treats it like a mysterious black box that no one needs to know how it works.
All-in-all for grayscale it's all just curves that maps input values to output values
according to math calculations. I got all this to work on Macs but its a bit more
cumbersome on PCs.
So given that the PC workflow w/ICC is a bit more awkward there have been a number
of workflows to accomplish something similar.
I've know Paul Roark for a long time and he's certainly someone who knows
what he is doing. He's got a Photoshop curves (.acv file) that he just puts on a
layer above everything else. Turn it on for printing and off for editing. In a sense
it works very much like a printing ICC profile -- just applied for printing.
So I decided to compare his .acv method to my ICC method. With Photoshop
this turns out to be very easy. Just take a 21step file assigned to GG 2.2. We're
going to see that actual data values would be sent to QTR driver with the 2 methods.
So for Paul's I just apply the .acv and flatten the image, for mine/ICC I just do a
Convert to Profile with my generic Gray-Matte-Paper. These give files that
would go to QTR without any more changes. I just used Photoshop to calculate
the difference function for these. Even to my surprise they are just about identical.
The maximum difference was just 2 (8-bit values) or less that 1%, the average
difference was less than 1/2 a bit-value or less that 0.2% -- amazing.
Paul -- how did you make this .acv ?? Just by eye?
I think this gives a lot more credence to the simple mathematic correction curves.
I also read Brian's paper quite a few times to get it all (maybe). I can't so far
think of an as easy test as Paul's but it does seem to depend on the standard
ICC methodology so I'm inclined to think it'll yield similar results. But I think
it's more of an edit-in-the reduced-DR workflow.
BTW, Brian had a link to an earlier discussion that had more info -- thanks Brian:
Another comment I have: both Paul's .acv and Brian's second graph show
fairly pronounced flattening/compression of shadow area. What I think is not
obvious is that there are two components to this. There's the compression due
to a weak dMax -- i.e. the basis of most of this whole discussion, but also there's
the Gamma 2.2 curve of the source file. Both of these corrections are trying
to mimic the L-values of the G2.2 curve NOT the K-values that are being
graphed. So the severe compression of shadows in G2.2 are also being
shown in these curves.
Roy
-- a little aside to those feeling like the regular ICC workflow with Gray-Matte-Paper
is a bit strong in its compression of shadows. There generic ICCs are super
simple, you can easily make your own with different parameters.
The input for Matte-Paper is simply two L-values-- 96 16
Those were chosen as typical matte dMin and dMax. If you'd like a weaker
correction just use 96 12 say. All you need in a .txt file is those 2 numbers
and Create-ICC will create a new generic ICC for you. Try pairs till you like it.
Roy