[comp.sys.ibm.pc] Color questions

chris@mimsy.UUCP (Chris Torek) (05/31/87)

In article <725@bsu-cs.UUCP> dhesi@bsu-cs.UUCP (Rahul Dhesi) writes:
>... Your typical printer will tell you he or she uses only three
>colors, with a little black thrown in.

Actually, high quality color magazine printing usually uses four
colours plus black.
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690)
Domain:	chris@mimsy.umd.edu	Path:	seismo!mimsy!chris

jon@eps2.UUCP (06/02/87)

In article <6861@mimsy.UUCP>, chris@mimsy.UUCP (Chris Torek) writes:
> In article <725@bsu-cs.UUCP> dhesi@bsu-cs.UUCP (Rahul Dhesi) writes:
> >... Your typical printer will tell you he or she uses only three
> >colors, with a little black thrown in.
> Actually, high quality color magazine printing usually uses four
> colours plus black.

No, Rahul is correct.  What you see in magazines usually begins as
continuous tone art (photographs or transparencies) and are then
scanned with a Hell or Crosfield laser scanner to produce four
color separations (cyan, magenta, yellow and black) which look like
pieces of black and white transparency film with halftone dots on
them.  A printer can then etch printing plates from these color seps.
For the past year I have been sending images back and forth from our
workstations to Scitex and Hell pre-press equipment at various color
separation houses in the Bay Area and making Cromalins (proofs) so
I've got some idea of what's going on here.

In printing packages they often use a different technique.  They use
the color of ink that is actually getting printed, rather they simulate
continuous tone with CMYK halftone dots.  I'm looking at a "Bounce"
fabric softener box and it looks like they used blue, red, and yellow
ink to produce a white, yellow, orange and blue package.


Jonathan Hue	DuPont Design Technologies/Via Visuals		leadsv!eps2!jon

chris@mimsy.UUCP (Chris Torek) (06/03/87)

>In article <6861@mimsy.UUCP> I wrote:
>>Actually, high quality color magazine printing usually uses four
>>colours plus black.

What I neglected to mention was that this information was old; I
learnt it perhaps ten years ago.  The extra color (switching
spelling systems will drive me crazy yet!) was brown; apparently
whatever processes were used did not reproduce browns very well.

In article <93@eps2.UUCP> jon@eps2.UUCP (Jonathan Hue) describes
a laser scanning process that produces `four color separations
(cyan, magenta, yellow and black)'.  Clearly this is a newer
process that has taken over now that scanning lasers are cheap.
I stand corrected.

(We now return to your regularly scheduled news :-) .)
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690)
Domain:	chris@mimsy.umd.edu	Path:	seismo!mimsy!chris

falk@sun.UUCP (06/06/87)

In article <93@eps2.UUCP>, jon@eps2.UUCP (Jonathan Hue) writes:
> 
> No, Rahul is correct.  What you see in magazines usually begins as
> continuous tone art (photographs or transparencies) and are then
> scanned with a Hell or Crosfield laser scanner to produce four
> color separations (cyan, magenta, yellow and black) which look like
> pieces of black and white transparency film with halftone dots on
> them.


Here's a related question, does anybody know the conversion factors to
map colors in Red-Green-Blue space to Cyan-Yellow-Magenta-Black?  What
I would like to do is generate four images on a high-res laser printer
that can be used directly for printing; i.e. do the seperation in
software.

		-ed falk, falk@sun.com, sun!falk

jon@msunix.UUCP (06/08/87)

In article <20537@sun.uucp>, falk@sun.UUCP (Ed Falk) writes:
> Here's a related question, does anybody know the conversion factors to
> map colors in Red-Green-Blue space to Cyan-Yellow-Magenta-Black?  What
> I would like to do is generate four images on a high-res laser printer
> that can be used directly for printing; i.e. do the seperation in
> software.

Yeah, I do, but not in a way that would be considered acceptable by
the Graphic Arts community.  :-)  Soon, though...

If anyone does know how to do the RGB->CMYK conversion and has results
which the Graphic Arts community has judged acceptable, please send me
your resume because I have a job opening for you.  (no :-))

Makers of prepress equipment don't work in RGB, they use CMYK frame
buffers and some op-amp adders to generate (RS-170,-343) RGB video
to drive the monitor.  There is a good reason for this - their
requirements are such that they can't trust the colors on the monitor.
When they are picking a color to use, they look the color up in process
color book and read the CMYK numbers they need to use.  Besides, the color
gamuts of video monitors and printed paper are different.  So they just
think of the monitor as "in the ballpark", but use their Pantone guides to
get the *right* color.  Also, ambient room light has an effect on the eye's
perception of the colors on a video monitor.  You can't match the colors
on a monitor with those on printed paper without taking a zillion things
into account, so why bother?  Look the color up in the book and punch the
numbers in.

Then you get to worry about different types of ink and paper, along with
dot gain (inks absorb light better when they land on paper, worse when
they land on other ink).  Also, you can have square, elliptical, or round
dots, and about four different sets of screen angles.  Because of all these
variables, a proof (Cromalin, Matchprint) is made before the seps are
taken to the printer.  The proof is supposed to show you what the art is
going to look like on paper, given the materials, etc. which you are going
to use.  Oh yeah, another thing to consider is the amount of UCR (under
color removal), replacement of CMY ink with black, not to get a blacker
black, but just to reduce the amount of ink used, and sometimes increase
contrast and detail.

So the people with extreme requirements don't even work in RGB space.
Where does that leave the rest of us with RGB frame buffers?  Well, I
don't know what Eikonix does.  I think they're still using the uvL
color space.  Whether they have a uvL frame buffer and do the RGB
conversion on the fly, or use an RGB frame buffer is unknown to me.
For people with pure RGB systems, if you can get hold of a uvL color
TV analyzer, and a uvL transmission-reflection densitometer, you can
come up with tables for the conversion.  Make sure you can map the colors
from the process book exactly (have an RGB triple in your system that
maps to a CMYK quad which is a process color) to keep the Graphic Arts
types happy.  Or you can use the process color book, and match those
to the monitor by eye.  Don't laugh, I know some people who have done
just that.  Another way would be to model the absorption of light by the
inks to come up with tables.  I'm too much of a feeb at Math and Physics
to figure that one out.  Anyone ever come up with a model?

One important thing that I forgot to mention is that the inks used in
printing do not absorb light like an ideal ink.  As I recall, yellow ink
is the closest to being pure, and magenta is the dirtiest (I think it
absorbs a lot of blue in addition to green).  So there is no simple
formula, despite what a lot of uninformed people think.

As far as resolution goes, typical output scanning for good quality
magazines (continuous tone art) is 12dots/mm (304dpi), with 255 different
dot sizes.  Also, each layer of dots is at a different angle, and the dots
can be round, square, or elliptical.  If there is a laser printer out there
capable of printing 192dots/mm (that's > 4800dpi!) on a transparency, there
are a lot of Hell flatbed laser scanner owners that would be interested
($500K a pop!).  I find it hard to believe that a "high resolution" laser
printer could produce anything except rough comps.  Final separations,
no way.  If these numbers are making you laugh, you've never seen
a Graphics Arts department manager whip out his loupe and start looking at
screen angles, dot shapes, and dot sizes.

dave@onfcanim.UUCP and ghn@munsell.UUCP know a lot about this field and
could correct any errors in this posting.  I'd be interested in hearing
anything they had to say about this subject.


Jonathan Hue	   DuPont Design Technologies/Via Visuals    leadsv!msunix!jon

amamaral@elrond.CalComp.COM (Alan Amaral) (06/11/87)

In article <20537@sun.uucp>, falk@sun.uucp (Ed Falk) writes:
> Here's a related question, does anybody know the conversion factors to
> map colors in Red-Green-Blue space to Cyan-Yellow-Magenta-Black?  What
> I would like to do is generate four images on a high-res laser printer
> that can be used directly for printing; i.e. do the seperation in
> software.
> 
> 		-ed falk, falk@sun.com, sun!falk

It's trivial.

where R=G=B,
	then Black = 1-R (or G or B)
	C=M=Y=0
else

	black = 0
	+- -+	+- -+   +- -+
	| R |   | 1 |   | C |
	| G | = | 1 | - | M |
	| B |   | 1 |   | Y |
	+- -+   +- -+   +- -+
-- 
uucp:	 ...decvax!elrond!amamaral		I would rather be a
phone:	 (603) 885-8075				fool than a king...
us mail: Calcomp/Sanders DPD (PTP2-2D01)
	 Hudson NH 	03051-0908

jon@eps2.UUCP (06/12/87)

In article <943@elrond.CalComp.COM>, amamaral@elrond.CalComp.COM (Alan Amaral) writes:
> It's trivial.

Oh come on.  Everyone who reads this newsgroup knows how to translate between
the RGB and CMY color spaces.  And I'm sure everyone could figure out how to
do the crude undercolor removal...

The question Ed asks is not the simple one that you think he is asking.  From
his posting it is obvious that he knows about how real inks used by printers
work, and that they do not absorb light like ideal cyan, magenta, and yellow
inks.  None of the simple solutions I've posted to the net are adequate.

I have a Pantone color simulator in front of me.  Just looking at the front
page, I can see that there is no way you can print a bright orange color on
paper without using ink that is actually bright orange.

I asked this same question about a year ago.  One useful piece of information
I received was from someone at the University of Utah.  He told me that
Maureen Stone at Xerox PARC has done work in this area.  I found out she
has published papers and she recently gave a talk out here.

Ed, one thing you may want to do is buy a Pantone color simulator.  You can
get a feel for what the different inks look like when mixed together.  The
book consists of about 500 of the Pantone process colors, and then a color
chip next to them showing a close approximation using CMYK ink.  There are
three more CMYK color chips which can be torn out next to the big chip.  You
can find this at most any art store.  The price is a bit steep, $200.  The
other four Pantone books which make up the rest of the set are only $250 total.
Of course the colors printed are dependent on the ink, paper, screen angles,
etc, but you can get an idea of how things work from this book.

I know that Eikonix has a lot of people who are experts in this area, but
they're not going to divulge a lot of the information because it is
proprietary.  My personal opinion is that if Eikonix ever gets around to
announcing a Sun-3 prepress machine (if they haven't already), one of their
big features will be their expertise in color fidelity.  Our systems are
targeted at design, not production, so our "big features" are our tools for
graphic designers.

I made an error in a previous posting about halftone dot sizes.  Although
prepress machines store their pixels as 8-bit quantities, typically dot
size is controllable on an output scanner only to 2% increments, so there
are only 50 different dot sizes.  The top of the line Hell scanner generates
dots electronically at a maximum of 146 dots/mm.  If you are doing 12 halftone
dots/mm, that gives you about a 12x12 grid of dots to simulate your halftone
dot, so I guess you could assume 144 dot sizes.


Jonathan Hue	DuPont Design Technologies/Via Visuals		leadsv!eps2!jon

tkxyv@mergvax.UUCP (William Klein) (06/15/87)

In article <307@msunix.UUCP>, jon@msunix.UUCP (Jonathan Hue) writes:
> ...
> come up with tables for the conversion.  Make sure you can map the colors
> from the process book exactly (have an RGB triple in your system that
> maps to a CMYK quad which is a process color) to keep the Graphic Arts
> types happy.  Or you can use the process color book, and match those
> to the monitor by eye.  Don't laugh, I know some people who have done
> just that.  Another way would be to model the absorption of light by the
> inks to come up with tables.  I'm too much of a feeb at Math and Physics
> ...
> One important thing that I forgot to mention is that the inks used in
> printing do not absorb light like an ideal ink.  As I recall, yellow ink
> ...there is no simple formula, despite what a lot of uninformed people think.
> As far as resolution goes, typical output scanning for good quality
> magazines (continuous tone art) is 12dots/mm (304dpi), with 255 different
> dot sizes.  Also, each layer of dots is at a different angle, and the dots
> can be round, square, or elliptical.  If there is a laser printer out there
> capable of printing 192dots/mm (that's > 4800dpi!) on a transparency, there
> are a lot of Hell flatbed laser scanner owners that would be interested
> ($500K a pop!).  I find it hard to believe that a "high resolution" laser
> printer could produce anything except rough comps.  Final separations,
> no way.  If these numbers are making you laugh, you've never seen
> a Graphics Arts department manager whip out his loupe and start looking at
> screen angles, dot shapes, and dot sizes.

At Linotype we make equipment for the graphics arts industry. The odd thing
about this industry is that the money (in color, at least) is where the
quality is. We have just evaluated our high resolution laser output printer
for its suitability for color work, and it is just barely good enough to
do kind of lousy color proofs (in the words of a graphics arts dept mgr).
This device outputs 2540 dpi, with an error across the whole raster (for
newspaper or wider widths) of <60 microns. No slouch, but nowhere near good
enough for color work. Indeed, you do need ~5K dpi to get good color output.
And the error must be extremely low for screen registration: the eye is
amazingly good at picking up moire (beat frequencies between the different
colored dot screens in the space domain) which can be caused by only a few
microns of misregistration between the screens.

The kind of color photographs that you see on the covers of Vogue, Life,
National Geographic, etc. go through an incredible number of iterative
proofs which involve one test print after the other being passed by the
customer for color checks, overall impression evaluations, and all kinds
of subjective judgements. Not easily automatable (yet, where are you AI and
CE people!?!).

Forget electronic (screen) proofing: you can't get color anywhere
close (to a graphics arts person) on the screen to what the ink will look
like. You can't even print proofs that look close enough for many customers
without actually printing the proofs on the same press, using the same ink
and paper as will be used on final output. Some customers even have a
'standard' light which will illuminate the proof with a known set of 
frequencies for the color evaluation. (A printed color photo will look
quite different under tungsten light, flourescent, 'warm' flourescent,
and daylight conditions).  With all of these variables, most of what the
video pre-proofing is used for is cropping, some editing, perhaps some
drawing (esp in ad layout) and the like. No color adjustment.

Perhaps when we get video monitors with 60K by 40K pixels, and a RFB and
blit engine fast enough to draw them all in a short time, and some kind
of standard colors... (:-))

-- 
Life is what happens to you when you are busy making other plans.
Real Life: 		W. Ted Klein
UUCP:			ihnp4!motown!mergvax!ted
VOICE:			516-434-2687

klm@munsell.UUCP (06/18/87)

In article <153@mergvax.UUCP> tkxyv@mergvax.UUCP (William Klein) writes:
>
>Forget electronic (screen) proofing: you can't get color anywhere
>close (to a graphics arts person) on the screen to what the ink will look
>like.

I beg your pardon?  Our customers seem to be very pleased with what we've
done.  It took a lot of color science, brain smoke, and patience, but we're
very proud of what we've accomplished.

>You can't even print proofs that look close enough for many customers
>without actually printing the proofs on the same press, using the same ink
>and paper as will be used on final output.

Sure you can.  I think we're the only ones who have figured out HOW, or 
who have bothered to do it, but it can be done.

>Some customers even have a
>'standard' light which will illuminate the proof with a known set of 
>frequencies for the color evaluation. (A printed color photo will look
>quite different under tungsten light, flourescent, 'warm' flourescent,
>and daylight conditions).  With all of these variables, most of what the
>video pre-proofing is used for is cropping, some editing, perhaps some
>drawing (esp in ad layout) and the like. No color adjustment.

The main selling point of our CEPS is that you *can* do color correction
on the screen, and that while on many analog scanners (Hell, Crosfield,
et al.) you have to adjust, proof, adjust, proof, adjust... we put a
little more effort into the color science aspect of it instead of just
the mechanics and we came up with a system that, when properly calibrated,
produces a video proof that is nearly indistinguishable from a hardcopy
proof.  This is assuming of course that both are viewed under proper
lighting conditions.  Plus, we allow the customer the freedom of modifying
the calibration tables to suit his own taste, so HE/SHE can get the results
that look good to HIM/HER.  Unfortunately we still have the big problem
that while color can be described very accurately by science, human color
perception is still a very much subjective thing, and we all know from
experience that sometimes there's just no accounting for a person's taste.

Stay tuned for more.  The fun is just beginning.

-- 
Kevin McBride         |Disclaimer:  These   | harvard -\
Eikonix - A Kodak Co. |  opinions are mine, | ll-xn ---adelie-----> munsell!klm
23 Crosby Dr.         |  not my employer's, | decvax -v  talcott -v  |
Bedford, MA 01730     |  So There!          | allegra ------------encore

klm@munsell.UUCP (06/18/87)

Our ears are burning...

A few of us here at Eikonix are wading through this glut of postings and
we will shortly attempt to answer as many of the questions as we can.
As one poster already pointed out, some of what we know to be fact is
proprietary, so we'll have to gloss around it.  We're also going to give
references to some papers we've published on the subject, plus some other
hopefully useful sources.

Hold your breath until you turn a nice shade of cyan :-)

BTW, I really liked the 'trivial' answer to the RGB -> CMY transform,
although it really should have been posted to rec.humor.

The transform from a 3-D color space (RGB, uvL, etc.) to a 4-D color space
(CMYK) is not trivial and is not solvable in closed form.  You gotta go
backwards first and build a table that allows you to weed out the multiple
choices when you go forward again.  I know this explanation is rather
sophomoric, but it's kind of late, I'm getting tired, and I want to be
careful that I don't give out any secrets that would cause my employers to
get pissed at me.  I'm not quite ready to take up the job offer from the
guy at DuPont (?)

Stay tuned to sci.physics.  That's where we'll be posting.  No cross postings
unless I get sufficient demand.

-- 
Kevin McBride         |Disclaimer:  These   | harvard -\
Eikonix - A Kodak Co. |  opinions are mine, | ll-xn ---adelie-----> munsell!klm
23 Crosby Dr.         |  not my employer's, | decvax -v  talcott -v  |
Bedford, MA 01730     |  So There!          | allegra ------------encore

cetron@utah-cs.UUCP (Edward J Cetron) (06/19/87)

In article <1080@knopfler.munsell.UUCP> klm@munsell.UUCP (Kevin McBride) writes:
>Stay tuned to sci.physics.  That's where we'll be posting.  No cross postings
>unless I get sufficient demand.

	demand, demand, demand.... please cross post them here, I for one
would like to see them.

-ed