Stochastic Screening - Color Correction
In my last column I presented
an overview of Calamus SL. There is one ability of Calamus
which remains obscure. This aspect of the program is
color correction and the related topic of raster screening.
My next two columns will address these questions. First
a little digression about stochastic screening. I do
this because I love the Calamus implementation and discussing
its module (Star Screening) will lead into color correction.
Now down to some brass tacks.
One of my favorite Calamus
modules is the Star Screening module. Star Screening
Lite comes with the standard Calamus package, in other
words, free. This can be updated to Star Screening Pro.
Well, don't bother upgrading. Everything the Pro version
can do can also be done with the Lite version, although
image conversion takes a few more steps. Now after this
loaded introduction let me explain what all this is
about.
There are basically three
ways or forms which can be used for printing. Raster
screening, dithered printing and stochastic screening.
The first has been around for quite a while and is the
standard method used by printing presses. Dithered screening,
or as I call it, "Poor man's" stochastic screening,
was relatively rare. Its use is now basically limited
to the personal printers owned by computer users. The
same is true of stochastic screening. Both dithered
and stochastic screening produce better results on the
home printer than rastered screening.
Rastered screening is
the creation of "raster points" meaning the
individual dots a printer is capable of printing are
grouped together to form raster points. These points
are combinations of the various printing colors, which
fool the eye into seeing the full spectrum of color.
The pattern created is easily photographed by a pre-press
machine and transferred through a number of steps to
a press. This is done because there are limitations
to the transfer technology and the larger raster points
are much easier to photograph than the individual dots
themselves. So for example a line screen of 100 means
there is a capability of squeezing in 100 of these raster
points per horizontal inch. This should not be confused
with the term "DPI". DPI (Dots Per Inch) refers
to the capability of the printer. While a 2,400 DPI
Lino-type machine can print out many more dots than
a 720 DPI printer, both can print out a 100 LPI photograph.
But the Lino can cram many more dots into each raster
point than your handy home Epson. Dithered
printing or ordered dither is simply the conversion
of the printed target (photograph, work of art...) into
individual dots which match the printer's capabilities.
It is often the default method which programs like Photoshop
use. With some printers, because of the printer driver
software, it is the only method you can print with.
Dithered printing is a straight conversion. Your photograph
which consists of individual pixels of color (24-bit
printing means the capability of 16,777,216 colors)
is converted mechanically into the four (or six) printing
colors).
Stochastic screening
(also called FM (Frequency Modulation) screening) takes
dithering a step further. Instead of simply converting
the underlying colors to CYMK (Cyan, Yellow, Magenta
and blacK), stochastic screening uses an intelligent
algorithm to make this conversion. Your 24-bit photograph
consists of pixels, areas of colors, which roughly correspond
to the DPI of the image. Thus the software examines
the first pixel of your image and makes decisions about
which color dots to apply. It then makes the next decision
keeping its previous choice in mind. The resulting image
has certain advantages and disadvantages over conventional
screening (rastered) and dithered screening.
On low resolution printers
(anything under 2,400 DPI) the advantages are obvious
to the eye. Much sharper images result, much more detail
is evident in the image. Another advantage is the moiré
patterns sometimes created by raster screening is not
a problem. Moiré patterns result when an image
is not properly sized to create the raster pattern.
A dithered or stochastically screened image can be of
any size, even dramatically distorted, without any hint
of a moiré pattern. There are also a number of
disadvantages. One is that extremely light areas of
the image may appear grainy but the overall effect is
one of extreme clarity and detail. This is not a problem
with either raster or dithered screening. Another disadvantage
is that once an image is converted it can no longer
be resized or manipulated. It becomes fixed at the size
you've screened it.
Aside from the above
neither dithered or stochastically screened images can
be sent to a press. They are both good for individual
prints. The dots are too compact to be photographed
for a conventional press, though this might change as
technology advances. Finally, the process of creating
stochastically screened images is time-consuming, unlike
either of the other two methods. Still when one compares
print-outs where all else is the same, the stochastically
screened image appears sharper, sometimes dramatically
so compared to raster screening and better than dithered
screening as well. Take a look at the three photographs
below for examples of all three screening methods. Notice
the detail or lack of same in the crops of these images.
With Calamus, stochastic
screening goes under the name of Star Screening. Other
companies use different algorithms and use different
trade names, "Crystal Screening" et al. The
design used by Calamus is excellent in its results and
its big advantage over other software is that with Calamus
you can Star Screen one image without affecting any
other aspect of the page. The Star Screening module
(once again not to be confused with dithering) is far
cheaper than any other source of this technology. That
goes for even the Pro version. In some cases stochastic
screening on other platforms is a hardware solution.
Some companies sell dithering as stochastic screening
which is really a pointless exercise, since most high-end
programs have this as an option. The basic Calamus package
cannot convert images to dithered format. This lack
is meaningless since Calamus can set dithering as the
screening method and your image will print out as a
dithered photograph. Or the optional filters module
contains a filter to convert images to dithered format.
These last questions are unimportant.
Two last points in this
introduction. Star Screening is memory-intensive. To
Star Screen a 16x20, 24-bit color image requires something
in the order of 200 MB of RAM! And it's slow. For example, on a Hades 060 you can expect
this task to take 25 to 30 minutes. This assumes you
have that kind of memory. It will take longer if using
the virtual memory option. On a G3 Mac at 400 MHz it
takes 12 to 15 minutes, once again assuming you have
the RAM (for a black and white image divide by four).
I like the results so much that this is my preferred
method of screening for individual prints, but as the
above indicates it takes time. Naturally, smaller images
can be screened quite a bit faster. I should also add
that laser copiers actually do a pretty good job of
copying these print-outs so already there are some commercial
possibilities.
How are images prepared?
RGB to CYMK conversion
Now to tackle the most difficult aspect of desktop publishing
and the only area of the Calamus manual which is not
up to par. I'm going to be tossing around the expression,
"color correction." The normal assumption,
when people hear this term, is going to your color look-up
table (CLUT) and increasing the contrast, or increasing
blue until your picture looks right. This is not what
I am talking about. Color correction is the conversion
of RGB to CYMK for the actual printing process. It does
not mean correcting an image so it looks good on your
screen, but rather what your photograph looks like when
it prints to hard copy.
Calamus has no automatic
way to create corrected color prints. Photoshop, for
example, has a number of methods for color correcting
your prints. On the other hand Calamus has extensive
methods of manually correcting your prints while programs
like Photoshop have almost none. In Photoshop you select
between a small number of options and you are happy.
You are happy because color correction is a very difficult
thing to accomplish and here someone is reaching out
of the computer monitor to hold your hand. Quite a few
printer drivers available on Mac and Windows machines
also have a number of choices for creating good output.
There are even a number of slider bars for correcting
color tone and depth. These methods work fairly well
and eventually you will get prints that roughly correspond
to what you want. Do you care that your deep blue sky
has too much cyan? Who's going to know the difference
as long as it looks good? Calamus provides only manual
controls but these in turn can lead to an exact reproduction
of what you want or for that matter the ability to add
a color cast to the exact degree you want. This month's
article is going to concentrate on color correction
so consider this half of what you need to know and it's
the important half. My next article will deal with raster
screening.
Color correction On-screen images are presented
to the viewer as RGB (Red, Green, Blue). This is the
way a monitor creates the spectrum of color. This is
done by light passing through the image. No RGB color
and your monitor looks black. Your common ink-jet printer
(or press) uses CYMK. We actually see color on paper
as light being reflected from the image. No CYMK and
we see white. When you do the physical action of hitting
the print command you are also giving a command to make
this conversion. At this time in technological history
there is no absolutely fool-proof method for doing this.
This is because while it would seem that a formula could
be concocted, either by engineers in a lab or by witches
screaming incantations over a cauldron, the fact is
no method works with all images. Is the yellow you see
on the screen a patina or a concentrated glop? Cyan
and magenta make red. What is the intensity of the red?
And so on. There's another problem. Green is the product
of mixing yellow and cyan. But dark green has black
added to the spectrum. While theoretical mixing CYM
would produce black, actually it produces a very deep
brown.
The printing trade basically
uses two forms of conversion. There are more, but these
two are the basics you really should know. If you ever
wander down to a service bureau, filled with ignorant
people anxious and willing to totally screw up your
job it pays to know how to get them to do what you want.
We don't need their help, we can screw it up on our
own. UCR (Under Color Removal) also called Chromatic
Separation and GCR (Gray Component Replacement) also
called Achromatic Separation. These are the two basic
methods. Both have unlimited adjustment capability.
Before I go any further
let me say this is all not
that difficult!
It only sounds that way because we are learning a new
language, but hey there are only a limited number of
words to learn!
Essentially when you
give that print command you are creating four color
plates which your printer then methodically prints,
one on top of the other. With UCR we are going to make
some of that black plate with a combination of CYM.
With GCR we are going to be using black for this black
plate. Now the weaker the under color removal, meaning
the weaker the use of CYM to make black UCR fades into
GCR. Although these methods (UCR and GCR) are referred
to separately in the literature, this is a semantic
convenience although the differences are real. If you
are using a strong UCR method you will also be using
a lot of ink. This is because black in such a print-out
is a combination of CYM. Everywhere in the print-out
three colors are being printed on top of each other
to create black. So such a method would be inadvisable
for a newspaper. This paper cannot handle this amount
of ink. Also, the more ink you use the greater the problems
of "trapping." Trapping is the effect of ink
being printed one color on top of another. With a printing
press this can sometimes be a real problem. If you are
doing color prints of paintings in a high quality book,
with high quality paper, UCR will work just fine. It
would be inadvisable to remove all the black, but in
a strong UCR print true black is used only to create
contrast. Flesh tones and areas of light color might
very well look better with a UCR print. At the same
time areas of light gray, like stone or concrete, will
look better with a GCR method. This is because you are
using real black to form these areas of gray. How is
all this implemented in Calamus.
The UCR dialog The dialog for creating
these CYMK separations is referred to as the UCR dialog.
It could just as well be called the GCR dialog, but
never mind, here is where the separations are done.
Let us pause for the rest of this article. Let us dispense
with these cruel and unnatural terms. Let us say that
inside Calamus is a dialog in which we can make smooth
transformations of our photographs on the monitor into
smooth and beautiful prints on our paper. To start with
this relatively easy (once you know it) process is made
more difficult by one big screw-up. The dialog, which
resembles the CLUT controls, works in the opposite manner.
By this I mean raising the line decreases the intensity
instead of increasing it. Keep this in mind or you will
eventually wind up homeless and insane, picking up used
soda bottles for the five cent deposits.
Tests We are faced with seven
control lines, conveniently of different colors. They
represent the four linearity (intensity) colors of CYMK.
With these four lines you can increase or decrease the
amount of the respective color, just as you do in the
CLUT controls (but backwards, remember backwards). The
next three lines are the black from cyan, yellow and
magenta respectively. By this I mean the amount of these
three colors that are being used to create black. These
three lines determine how much of the ink from CYM is
going to be used, not to produce the color itself, but
to produce black. Green, as an example, is produced
by mixing cyan and yellow. What about dark green? With
dark green we have the addition of black. It can be
true black or it can be CYM black. Either way you need
black.
Now load in a color photograph.
Go to the raster screening module while in the printer
dialog and choose ordered dither. Now print it out.
Looks terrible. While I don't have your photograph in
hand my prediction is a good one. From the printer dialog
go to the color separation dialog (you can access these
dialogs from the page module but this is more convenient).
You will notice all the lines are in a neutral position.
There is no CYM black being printed.
|
A
grayscale fountain made from RGB not black. |
This leads us to
our first step in creating a control line for your printer.
Create or steal a greyscale fountain. You can steal
it from the example image in this article. Have this
ready as a file. Load it into Calamus, go into the color
separation dialog and delete the black linearity line.
Yes, you move it all the way to the top. When you try
to print, nothing will print. In the neutral position
these lines eliminate all CYM black and you just deleted
true black - nothing will print! To create color separation
lines to match our printer all we have to do is print
out a grayscale fountain using just CYM! As you near
perfection, and like trying to get up to light speed,
it never quite gels, add true black. You are done. Your
photographs will print out astoundingly beautiful!
|
The UCR dialog with the control
lines set to optimize for the Epson 3000
printer. The above control lines will print
out a grayscale fountain without any black
ink being used.
|
A few tricks Your monitor is not calibrated.
There are too many variables to create a perfectly calibrated
monitor. Invers Software sells a module to do just that.
Unfortunately this is one of the few modules I don't
own and the demo version just demonstrates how the module
works. One of the things I did was after embedding my
UCR control lines in an image I used the built-in conversion
options in Calamus to convert RGB to CYMK. This will
not show an exact reproduction of the print-out but
it helps. Make a copy of your fountain and after embedding
your control lines - convert to CYMK. You might then
want to print out the result and try to move your monitor
adjustments to closer match the print-out.
Franklin
The full
version of the curve editing module, Franklin, is now
standard in Calamus. Franklin can be used in any aspect
of Calamus where control lines are being edited or adjusted.
Furthermore all
the created
lines can be saved in Franklin's own internal format,
CKD. So when you are editing in the UCR dialog, the
UCR format is CK7. Your test files should also be saved
in the CKD format so that changes can be made without
losing the ability to modify these lines. So, when you've
finished editing save as a "test" CKD file
and export as a CK7 file. The CK7 file is the color
separation format and CKD is Franklin's internal format
where Bézier control points are stored.
A few
warnings
If you are moving cyan out of black you are creating
extra cyan. Remember this cyan was being used to make
black, now it's just hanging around mucking up the works.
So as you move cyan out of black you must also adjust
the cyan linearity line to decrease the amount of cyan.
The same goes for the other colors. Make sure to save
all these lines you are creating! Create temporary folders
and keep track of the results. Once you are getting
near the results you want make notes of whether the
image is perfect but with a slight color cast. You might
want to keep these less than perfect lines for when
you want a color cast. For example I
keep about four of them around. One is my "perfect"
line the rest tint my subject very, very slightly in
the direction of cyan, magenta and yellow respectively.
This is in case a print-out needs more yellow despite
using my perfect line or for that matter, if I want
to provide a yellow cast to the image. The other bad
news is you'll need different versions of these lines
for the three methods of printing. The good news is
they don't differ by too much. You can even avoid a
separate line for stochastic screening. You can simply
reduce the intensity of the target photograph and use
the lines you created for raster prints. This little
trick will not work with ordered dither but the raster
control lines form a good starting point.
Finally we come back
to stochastic screening. There is so much ink being
laid onto the page by using this method that one further
step is necessary. In this case you can make separate
lines from the UCR dialog for stochastically screened
images or you can use the real CLUT controls to lighten
the image. I prefer making my UCR lines, but either
method will work.
This is a two-part article.
As I said the next part will deal with rastered screening.
If an image is not color corrected the result of printing
will look awful. Much of the ink being used will not
be set down in the right place. Users of Calamus who
always got good results with black and white and grayscale
images suddenly find they can't print out the simplest
photograph. I will discuss this question in much greater
depth but you will find that once an image is color
corrected the default screen provided by Calamus and
quite a few others will work flawlessly and your results
will thrill you.
|
The Hudson River at Bear Mountain
digitally photographed. |
|
Print-out from Photoshop using
the automatic settings. A very smooth print
but the colors of this test print are off
on the green side. Note that this was printed
at 1,440x720. The other images, from Calamus,
were output at 720 DPI. |
|
A 100 LPI raster screened print-out
from Calamus SL. |
|
An example of a Star Screened
image from Calamus SL. Colors have a bit
too much contrast and the water has a blue
cast, but note the detail. |
|
Detail
from original photograph. |
|
Detail from Photoshop print-out.
Once again this image was output at 1,440x720.
|
|
Detail from Calamus rastered
output. |
|
Detail from Star Screened print-out.
Note the detail as well as the increased
grain. Also keep in mind that despite the
dramatic increase in detail this image was
output at 720 DPI. |
The above images
represent an experiment. I tried scanning my print-outs
but the scanner picked up too much of the relatively
large dots produced by my printer. I've just purchased
a new digital camera, the Nikon D1X, and this machine
did an incredible job. This camera has a resolution
of 3,000x2,000 and by photographing the images as opposed
to scanning them, the results were pretty smooth. All
the images were first prints - test prints if you will.
Both Calamus and Photoshop will allow for further correction.
But I do believe that once mastered (that's the tricky
part, mastering the process) Calamus is actually easier
to use in terms of correcting images.
chatobarkin@myatari.net
|