Talk:CMYK color model

From Wikipedia, the free encyclopedia

This article is supported by the Color WikiProject, a project that provides a central approach to Color-related subjects on Wikipedia. Help us improve articles to good and 1.0 standards; visit the wikiproject page for more details.

Start

This article has been rated as start-Class on the quality scale.

High

This article has been rated as high-importance on the importance scale.

Article Grading:
The article has been rated for quality and/or importance but has no comments yet. If appropriate, please review the article and then leave comments here to identify the strengths and weaknesses of the article and what work it will need.

[edit] Old discussion

Adopted orphan redirects for searching: CMYK colour model

Can you convert the following RGB coordinates to CMYK??

Color    RGB         CMYK
Gray     128 128 128
Black    0   0   0
White    255 255 255
Red      255 0   128
Orange   255 83  0
Yellow   255 255 0
Lime     83  255 0
Green    0   255 128
Sea blue 0   172 255
Blue     0   0   255
Purple   172 0   255

66.32.71.96 02:21, 2 Apr 2004 (UTC)

Note: (or note to self :-P ) we need an RGB -> CMYK conversion formula, like on HSV color space , that'd be handy. Kim Bruning 12:38, 3 Apr 2004 (UTC)

Okay, let's snarf some formulas off of : http://www.easyrgb.com/math.php?MATH=M12#text12

(mathematics isn't copyrightable AFAIK, and I'll be rewriting anyway :-) ) and then rewrite them to 1 form:

// RGB -> CMY
//RGB values = From 0 to 255
// CMY values = from 0 to 1

C = 1 - ( R / 255 )
M = 1 - ( G / 255 )
Y = 1 - ( B / 255 )
//CMY values = From 0 to 1 

// and CMY-> CMYK
var_K = 1 

if ( C < var_K )   var_K = C
if ( M < var_K )   var_K = M
if ( Y < var_K )   var_K = Y

C = ( C - var_K ) / ( 1 - var_K )
M = ( M - var_K ) / ( 1 - var_K )
Y = ( Y - var_K ) / ( 1 - var_K )
K = var_K

Okay, so we'd better write that out tidily :-)

so let's see

$\begin{matrix} C_{CMY, float} = 1 - ( R / 255 ) \\ M_{CMY, float} = 1 - ( G / 255 ) \\ Y_{CMY, float} = 1 - ( B / 255 ) \end{matrix}$

$var_K=\left\{\begin{matrix} C_{CMY, float}, & \mbox{if } C_{CMY, float} < var_K \\ M_{CMY,float}, & \mbox{if } M_{CMY, float} < var_K \\ Y_{CMY,float}, & \mbox{if } Y_{CMY, float} < var_K \end{matrix} \right.$

$\begin{matrix} C_{CMYK, float} = ( C_{CMY, float} - var_K ) / ( 1 - var_K ) \\ M_{CMYK, float} = ( M_{CMY, float} - var_K ) / ( 1 - var_K ) \\ Y_{CMYK, float} = ( Y_{CMY, float} - var_K ) / ( 1 - var_K ) \\ K_{CMYK, float} = var_K \end{matrix}$

finally:

for: $C,M,Y,K \in ( \mathbb{Z}, [0,255] )$

we can simply do:

$\begin{matrix} C = 255 \cdot C_{CMYK, float} \\ M = 255 \cdot M_{CMYK, float} \\ Y = 255 \cdot Y_{CMYK, float} \\ K = 255 \cdot K_{CMYK, float} \end{matrix}$

~~Okay, as a sucky biologist, my math sucks. Can a real mathematician look this over before adding it to the main article? Else I'll add it in a week or so.~~ Note that I really don't like the float inbetween step, and it's not really nescesary :-)

Kim Bruning 13:21, 3 Apr 2004 (UTC)

Oh heck, let's be bold and add it, some mathematician will just come along and simplify the float step I'm sure :-) Kim Bruning 13:27, 3 Apr 2004 (UTC)

Ugh, I think I made some slight errors converting from algorithm to formula. Um I'll leave as is for now, some mathematician can pick up the pieces :-/ Sorry! Kim Bruning 13:32, 3 Apr 2004 (UTC)

Here's rewriting this so it's a little neater. This is going off that site before:

Converting CMYK -> RGB = CMYK -> CMY -> RGB

t_CMYK ∈ [0,1]⁴ is the CMYK quadruple, t_CMY ∈ [0,1]³ is the CMY triple, and t_RGB ∈ [0, 255]³ is the RGB triple. Then:

$\begin{matrix} t_{CMYK} = (c, m, y, k), \mathrm{then}\\ t_{CMY} = (c', m', y') = (c (1-k)+k, m (1-k)+k, y (1-k)+k )\ \mathrm{and}\\ t_{RGB} = (r, g, b) = (255(1 - c'), 255(1-m'), 255(1-y'))\\ \end{matrix}$

Converting RGB -> CMY, with the same color vectors as before

$\begin{matrix} t_{RGB} = (r, g, b)\\ t_{CMY} = (c', m', y') = (1-r/255, 1-g/255, 1-b/255)\\ t_{CMYK} = (c, m, y, k) = (c-\mathrm{min}(c, m, y), m-\mathrm{min}(c, m, y), y-\mathrm{min}(c, m, y), \mathrm{min}(c, m, y))\\ \end{matrix}$

Dysprosia 14:05, 3 Apr 2004 (UTC)

Thanks, looks tidier at least! On computers, CYMK is usually represented in 32 bits (4 bytes: C, M , Y, K), so each C, M, Y, K value is an integer from 0 to 255, and not a floating point value from 0 to 1 as was stated on the referenced webpage. Cutting out the floating point values would improve the formula further. (something to sort out in my copious free time :-) Kim Bruning 14:26, 3 Apr 2004 (UTC)

Can you put the coordinates in the above table that already shows RGB?? 66.32.123.29 15:36, 3 Apr 2004 (UTC)
- Instead of giving you a fish, we gave you a fishing rod. :-) Calculate the cmyk values using the formulas, or the webpage or what have you, and multiply by 255 (and round off) to get the compuer representation. Kim Bruning 16:00, 3 Apr 2004 (UTC)
  - What do fish and fishing rods have to do with RGB and CMYK?? Is this just an expression?? What does it mean?? Does Wikipedia have an article for expressions like this??66.32.113.34 16:55, 3 Apr 2004 (UTC)
    - "give a man to fish, and he will eat for a day, teach a man to fish, and he will eat for the rest of his life". The idea is that we just explained how you can figure out cmyk values for yourself, into perpetuity :-) Unfortunately the formula given by dysprosia gives values from 0 to 1, and we want to scale that to 0-255. (so just multiply by 255, easy enough). Well I'll do some examples to see if the formula is actually correct...

[edit] some tests

$\begin{matrix} (128, 128, 128)_{RGB} = (0, 0, 0, 0.5)_{CMYK} = (0,0,0,128)_{CMYK, scaled to bytes}\\ (255, 0, 0)_{RGB} = (0, 1, 1, 0)_{CMYK} = (0,255,255,0)_{CMYK, scaled to bytes}\\ (255, 255, 0)_{RGB} = (0, 0, 1, 0)_{CMYK} = (0,0,255,0)_{CMYK, scaled to bytes} \end{matrix}$

Wow! This formula is really easy to use too! :-) I'm sure there's a way to skip the float phase , but this is nice :-)

But 255 0 0 was not among the colors in the above table. 66.245.99.11 18:20, 3 Apr 2004 (UTC)
- But I'm not going to calculate your table for you, I was figuring some edge cases to see if the formula actually works. Kim Bruning 18:33, 3 Apr 2004 (UTC)

Hmm, note I was doing mental arithmetic, so values are rounded off. Kim Bruning 21:48, 3 Apr 2004 (UTC)

okay, one example for 66.32.123.29, though really wikipedia isn't for this. Note that your value for red is really weird?

We're using "red, green blue" so:

$\begin{matrix} red = (255, 0, 0)_{RGB}\\ green = (0, 255, 0)_{RGB}\\ blue = (0, 0, 255)_{RGB} \end{matrix}$

tell you what, let's find blue in CMYK

disprosia said:

Okay, well, $b l u e = (0,0,255) R G B$

so we do:

$t R G B = (0,0,255)$
$t C M Y = (c', m', y') = (1 - 0 / 255,1 - 0 / 255,1 - 255 / 255)$
$t C M Y = (c', m', y') = (1,1,0)$
$t C M Y K = (c, m, y, k) = (c - min(1,1,0), m - min(1,1,0), y - min(1,1,0),min(1,1,0))$

Well ... $min(1,1,0) = 0$ , so:

$t C M Y K = (c, m, y, k) = (1 - 0,1 - 0,0 - 0,0)$
$t C M Y K = (c, m, y, k) = (1,1,0,0)$

scaling:

$t_{CMYK, scaled}=255 \cdot t_{CMYK}$

So in the end we get:

$t C M Y K, s c a l e d = (255,255,0,0)$

Which I happen to know is the right value for blue. :-)

[edit] Easily understood

Doesn't anyone think this page should be made more simpler so people can understand or add more explanations to the algorithm-ishs given in the end? My thought since I still do not understand how. -- WB 02:30, Mar 18, 2005 (UTC)

I think the algorithms given should be removed, as they are of little practical use. They do not reflect the real-world conversions actually needed, and by putting them into a Wikipedia article, it adds to the confusion and errors which can produce very expensive printing mistakes. Views? Notinasnaid 09:34, 11 Apr 2005 (UTC)

It is less productive to remove inaccurate algorithms altogether as opposed to improving the algorithms in the first place. Dysprosia 12:19, 11 Apr 2005 (UTC)

The crucial point is that there is no algorithm. There is no single conversion from RGB to CMYK, so - much as people yearn for there to be a simple algorithm - it is doing a disservice to put one there. Notinasnaid 16:31, 11 Apr 2005 (UTC)

I have tried to put some background in the new article absolute color space. Neither RGB nor CMYK are, by themselves, absolute, so there is no single conversion between them. Notinasnaid 20:27, 11 Apr 2005 (UTC)

It seems to me that the lack of reversibility in RGB to CMYK to RGB conversions isn't an issue which should prevent presentation of an algorithm. Such an algorithm is a necessity in many applications; for example, rendering an RGB-based image into CMYK inks on paper. CoyneT 00:54, 12 Apr 2005 (UTC)

To my mind, that's exactly the reason to not have it, because, frankly, it's rubbish. People have written software using this algorithm, and the results of the conversion don't really satisfy anyone, because the colors are different. Notinasnaid 07:43, 12 Apr 2005 (UTC)

The discretionary note added is sufficient enough. Those who want to make use of such an algorithm are warned for its use, and those who need something professional will be warned of the consequences for using the algorithm. Dysprosia 08:00, 12 Apr 2005 (UTC)

That makes sense. If I ever manage to find a way to express how it should be done, I'll add that too. (It won't be an algorithm). Notinasnaid 08:28, 12 Apr 2005 (UTC)

[edit] Still doesn't make sense

The sentence still doesn't make sense. Suppose that the conversion from RGB to CMYK is reversible. I take this to mean that there exist a conversion from RGB to CMYK, and from CMYK to RGB. But if we suppose now, seperately, that the conversion from CMYK to RGB is not reversible. Then I take this to mean that there is a conversion from CMYK to RGB, but not one from RGB to CMYK. This clearly contradicts the first supposition. One of the suppositions is wrong. Which is it? Dysprosia 10:20, 20 July 2005 (UTC)

Are we talking about colour-for-colour matching? Please clarify the sentence if this is the case. Dysprosia 10:27, 20 July 2005 (UTC)

This is nothing to do with colour matching, just simple mathematics.

Clearly it either (a) isn't expressed clearly or (b) is expressed clearly but is so unexpected it seems to be wrong. Yes, you can start with any RGB value, use the formula to convert to "CMYK" and then back to RGB. You will have the same value, in every case. But if you start with a CMYK value, convert to "RGB" and go back again, you will (except in particular lucky cases) not have the same value. If you want a worked example, start with the two CMYK values 0,0,0,1 and 1,1,1,1 and see what you get when converted to RGB then back to CMYK. Notinasnaid 10:31, 20 July 2005 (UTC)

Let me try and clarify the wording based on this, so it doesn't sound so paradoxical. Dysprosia 10:36, 20 July 2005 (UTC)

That's pretty good, thanks! We might also want to speak of gamut here, as one of the reasons that RGB and CMYK aren't completely inteconvertable is that certain colors are "out of gamut" for one color space or the other and simply can't be represented (accurately) in the other color space.

Atlant 11:33, 20 July 2005 (UTC)

No, gamut doesn't apply here. These are artificial conversions (which I personally would prefer to see removed) rather than anything that does accurate colour. But people do seem to want to see formulae, even if they are rubbish. There is no gamut mapping implied. I'd like to see a separate subsection about the principles of an accurate conversion. Notinasnaid 11:48, 20 July 2005 (UTC)

I'm sorry, but you're wrong. "Out of gamut" colors are one of the principal reasons the conversions aren't reversible. While a theoretical CMY/RGB system might use identical primaries (if you think of C as -R, M as -G, and Y as -B), but we live in the real world and real-world CMY(K) and RGB use different primaries that enclose different gamuts.

Atlant 12:31, 20 July 2005 (UTC)

One more point: there cannot exist any RGB <-> CMYK mapping that is invertible in the CMYK -> RGB -> CMYK case. I don't know how to express it mathematically, but it's something to do with mapping four unknowns into three, and back to four; it's impossible to preserve all information. Or is it? Now, it may be invertible in terms of colour display, but that's a different story; the actual CMYK numbers are the point of the discussion. Notinasnaid 11:54, 20 July 2005 (UTC)

What counts is the size of the sets you're mapping between and/or the domain, range, and codomains of the conversions, depending on what you're talking about, but that is another story. What is being said currently in the article includes the case you just mentioned, I believe, though somewhat implicitly. I could give a mathematical reasoning here to support that or to further clarify that but it may be a little overcomplicated and I would be liable to make mistakes anyway ;) Dysprosia 12:27, 20 July 2005 (UTC)

(resonding to Notinasnaid) This would be true if "K" were really an independent variable. But it's not; it's simply a way of saving ink and accounting for the fack that in the real world, C+M+Y doesn't quite make K. See all the mentions of Grey Color Removal, Under-Color Removal, and the like.

Atlant 12:31, 20 July 2005 (UTC)

Let's try and make it clear what I'm talking about.

1. I am specifically talking about the formulae on the article page, not about the general case of CMYK <-> RGB conversion. These formulae make no allowance for gamut (work through them if you think they do), so gamut is not relevant to that discussion.

Okay. But let's make clear what's real-world and what's theoretical. Atlant 14:56, 20 July 2005 (UTC)

2. When I talk about reversibility I mean, specifically, a return to the same four variable values. I am not talking about whether the colour they represent happen to be the same. While it might in some situations be interchangeable, there are other situations where the conversion would produce bad results. The conversion – this conversion – applies only one rule for GCR/UCR etc.

I understood that. Atlant 14:56, 20 July 2005 (UTC)

3. These are bad formulae but if the article doesn't make it clear the preceding paragraph is only about them, then it should.

Maybe it would clarify the article to give specific examples of inputs and outputs for the formula, showing the non-reversibility. Notinasnaid 14:43, 20 July 2005 (UTC)

Please try and limit your use of the word "reversibility", as it can imply "invertibility" and thus needs to be properly qualified otherwise, and it caught me once already. Thanks Dysprosia 14:52, 20 July 2005 (UTC)

An example or three would be very helpful. Atlant 14:56, 20 July 2005 (UTC)

I am surprised by this whole bad formula discussion. When you convert CMYK to RGB, you are mapping a 4-dimension space onto a 3-dimension space, i.e. you lose 1 dimension (you can compare it to converting a 3D-design of a house to a floorplan -you cannot convert that floorplan to a 3D again). So technically it is not a conversion, but a 'projection' that is not reversible/invertable.

Evenso, one of the questions people have most often concerning CMYK is "but how can I show it on my screen?" so they need the RGB equivalent. So why was the CMYK-2-RGB converter (external link) deleted? Anyway, what comes out of your printer in most cases (most document/image software uses RGB) is an CMYK approximation of an RGB colour, done either by your printer or printer driver. So RGB-2-CMYK inverse projection is far from useless. The extra parameter you use is: do you want to use more black ink (K) or more of the three colour inks (CMY)? So the issue is more economical. Pforret

My argument against this formula is that it is not a good formula. It suggests that RGB cyan (100% blue+100% green) is best represented by 100% CMYK cyan. This is not the case: it is a fallacy. I think it is very important that this article educate people that there is no simple conversion that will produce a satisfactory matching from RGB to CMYK and vice versa. That what they seek does not exist. Otherwise, they will go away from this article and implement things on that basis, to nobody's good. Converting between RGB and CMYK requires knowing what color space is in use for RGB and CMYK, and doing a color mapping based on those color spaces (e.g. using ICC profiles). I have to ask this, I hope it does not come over too negative: is anyone who promotes this formula actually responsible for high quality color matching? Notinasnaid 15:45, 29 September 2005 (UTC)

I would expect that if someone were to develop a high-quality color matching system, he/she would dive into the whole Pantone/RGB/CMYK matter more deeply. For the Average Joe (i.e. 95% of your Wikipedia public) it is enough to know that the best approximation of a certain CMYK colour is RGB #XXXXXX. They don't care about color temperature, gamut and ICC profiles. If they get a color specified in CMYK, they want to know: is it greenish, redish or brownish? Oh, it's pink!

Again: the main purpose of an online converter is conversion from CMYK to RGB, not the inverse. Most of us live in a RGB world, we need our stuff in RGB and all conversion back to CMYK is done for us (by a printer driver or some piece of software). So if "Cyan" is (1,0,0,0) in CMYK, what is the best approximation of that color in RGB? Probably #00FFFF is a pretty good option. If you have a better formula, be my guest and share it. If not, please add the "not for professionals" warning and let Average Joe convert his CMYK to RGB.

--Pforret 08:13, 3 October 2005 (UTC)

No, sorry, I cannot agree. Leaving aside that Pantone has nothing to do with it... you say "it is enough to know that the best approximation" – no. It is not the best approximation. It is not even a good approximation. So the average Joe will go away with a bad approximation, which he does not understand enough to know is bad. So I argue that we should certainly not link to a calculator which people will certainly use in preference to actually reading the difficult and unsavoury facts in the article. "If you have a better formula" – - there is no formula. That's my point.

If this were an introductory piece I'd have no problem with leaving the link in. But this is supposed to be an encyclopedia; that means that I don't think that oversimplifications have any place; if the facts are complicated, it's the job of an encyclopedia to say so, hopefully in a way that can be understood, not to gloss over them. Notinasnaid 08:33, 3 October 2005 (UTC)

By the way, there is nothing impossible about a web site which does this properly while starting the education of people in the issues. For example, a site which offered a choice of RGB and CMYK profiles (ideally with the opportunity to upload a custom one), and then did the conversion on this basis. On the other hand the site to which you posted a link not only does not do this, it does not even carry warnings about how inaccurate the result is. I do not think that simplicity is better than accuracy. Also, please do not change the page while this is still under discussion. I observe that this is your own site (or there is a remarkable coincidence of names): I can understand the emotional attachment you have to it, but I do not agree it has any place on Wikipedia. It may be that this page needs to go to dispute resolution. Notinasnaid 08:48, 3 October 2005 (UTC)

I will try to construct a better formula using CIE-XYZ

--Pforret 10:42, 3 October 2005 (UTC)

[edit] Removing the picture

I propose removing the picture . It contains pure RGB cyan, magenta and yellow, which look very little like printing inks. I propose not replacing it because there is a wide variation in printing inks, and even if there wasn't, there is a wide variety in monitors. In general, I don't think any reference colors belong in wikipedia. Notinasnaid 11:21, 5 September 2005 (UTC)

Don't do that. While everyone knows what "Yellow" is, many people don't know what "Magenta" is and nearly everyone has no idea what "Cyan" is. The picture is a good start.

Of course, you should feel free to add another picture (of 55-gallon drums of ink or what-have-you).

Atlant 12:34, 7 September 2005 (UTC)

Well, that isn't what cyan and magenta look like. That's the problem... Notinasnaid 17:23, 7 September 2005 (UTC)

Funny, when displayed on an additive-color device, I was pretty sure that cyan was 100% green + 100% blue and magenta was 100% red and 100% blue. What do you think they look like (when displayed on an additive color device)? Or maybe you should print the Wiki page and see if your color printer shoots anything other than cyan ink for the cyan swatch and magenta ink for the magenta swatch; it ought to be pretty easy to tell, even without a magnifying lens.

Atlant 14:06, 8 September 2005 (UTC)

The article isn't about additive color. Certainly, additive cyan/magenta look like the picture. But I have never seen a cyan/magenta ink for use in CMYK that was anything like that. Both the cyan and magenta in CMYK are significantly darker, but vary depending on the ink mix. That's why CMYK has a significantly different gamut to RGB. This picture can only mislead and confuse people in an area where there is already substantial confusion about what these colors really look like. Should this article be about real CMYK or an idealised and nonexistent version of it? Notinasnaid 21:59, 8 September 2005 (UTC)

No, you're correct: this article isn't about additive color. Unfortunately, 99.999% of the time, this article will be viewed on a device that uses additive color, and you're just going to have to live with that. As I said, if you don't like those color swatches, find a photo of gobs of ink that pleases you, and add that to the article. But don't remove something that is technically accurate just because you don't like it.

(By the way, you're dodging the suggestion I made about printing the page using an inkjet printer. I assure you that the C, M, Y, and K inks used by those printers pretty-accurately match the color swatches inthis article as displayed on my various CRT and LCD displays.)

Atlant 12:00, 9 September 2005 (UTC)

I did your test, printing on an Epson 880 with Epson inks and default Windows 2000 driver settings. The result was nothing like the strong primaries in the original graphic. They were rather more like but of course that was created with the wrong CMYK profile: SWOP Coated rather than a profile for the Epson inks. The colors were certainly duller than that. And when I printed this new graphic they were duller still, a result of continuing gamut compression which means that they never can match with these printer settings.

Examining with a loupe I can see that in the first printout, the cyan contains white and magenta dots, the magenta has white and cyan dots and the yellow has white and cyan dots. In the second picture, the cyan seems to have only a little white, the magenta only a little cyan, and the yellow is very pure.

I feel that would be a better graphic - it at least looks something like some real world colors. But it is misleading, just like the formula for conversion which I would also like to see removed, because it reinforces wrong ideas. Would the entry for sparrow allow a picture that was sometimes a cormorant? I don't think so. This is why I think color swatches (except RGB swatches designed to show your own monitor's behaviour) have any places in an online resource that has pretensions to accuracy. Notinasnaid 13:07, 9 September 2005 (UTC)

[edit] Do I understand this correctly?

I have been reading about the RYB model and the CMYK model. First I'll try to explain my train-of-thought: in my mind, magenta is created by adding red and blue, this works for both paint and RGB values. So I was wondering how CMY is concidered a good base colour triad. So I started reading and found out about the principle of additive and substractive models. I understand these. I can see why CMY is a good model. Then I read this page and my initial thought were confirmed when I read: "dark blue reproduces poorly in CMYK, coming out rather murky (as it has to be approximated by darkening cyan with black and a little magenta)". So here is my theory: RGB works on a monitor because the monitors physically do like the theory: paint with light. The CMY model is a theory that translates poorly to printing practice, because we try to print using physical materials (ink/paint) to mix up a new material that absorbs light like the CMY model would. It is this that results that shades of blue and purple are poorly reproduced? Or do I miss something here?

You're not missing anything. The fact is that it is impossible to get close to a typical monitor gamut on a printer, especially using only four inks. But when four inks are used, although they are *called* cyan, magenta, yellow, and black, the cyan and magenta in particular are very different from the colors that are called by those names in an RGB model. By convention, any roughly blue-greenish ink is called "cyan", and any roughly pinkish ink is called "magenta". For example, here is a rough approximation of the ink colors used in a printer I'm currently working on (with the usual caveats about color reproduction on the web):

As you can see, this is something of a compromise between CMY and the RYB traditionally used with paints. This printer has an excellent gamut of blues, because the "cyan" ink is very blueish. Also, its C+M+Y is not "murky" at all; you can hardly distinguish it from black. It still can't produce anything like the intense blues you can get on a monitor, and it does less well with reds (despite the magenta being more reddish than most).

Using five or six inks can produce a significantly better gamut -- of course with higher costs for heads and ink. Some high-end printers use CMYK+Red+Blue; others use CMYK+Orange+Green. There are also a handful of 8-color printers, generally allowing some of the colors to be replaced by specialist inks for particular jobs. DavidHopwood 04:43, 19 January 2006 (UTC)

When I went shopping last summer, I saw some that were CYMKcm. (lowercase meaning pale colors) The light colors helped to avoid speckles when printing pale things. AlbertCahalan 04:42, 26 April 2006 (UTC)

Just for fun, I converted the L*a*b* values provided by the ink manufacturer to sRGB, using the icctrans utility from http://www.littlecms.com/newutils.htm (with colorimetric rendering intent):

Hmm, the yellow and magenta don't look right at all, and I trust my eyes more than icctrans (although maybe I was overestimating the brightness in the first image). Oh well. DavidHopwood 05:34, 19 January 2006 (UTC)

[edit] Why no formula

I have been asked to explain why I say "there is no formula" for converting between CMYK and RGB. It does on the face of it sound like nonsense. Here's why:

I'll try to explain where this "there is no formula" comes from. There are two distinct phases (a) why there isn't one in theory and (b) why there isn't one in practice if the problem is reduced to one where calculation is possible.

1. There is no formula because the problem is not well defined. Neither RGB nor CMYK are absolute color spaces, so to talk of conversion is not really meaningful until both spaces are actually defined. For instance, if you were talking about sRGB and a particular CMYK profile (derived for ink and paper conditions), then the conversion becomes meaningful. Since each combination of RGB and CMYK profile has a different answer, you can't say anything about the conversion until those are decided. Rendering intent is relevant too.

PostScript version 1 used only CYMK. Was it well-defined? If yes, then I suggest using that with sRGB. AlbertCahalan 04:45, 26 April 2006 (UTC)

CMYK appeared in a supplement to level 1 PostScript, yes. The actual conversion was well documented (I think this is one reason it has been popular) but did depend on what values were passed to setundercolorremoval and setblackgeneration. But whatever values were passed it was a very poor match colorimetrically (this color becomes something like this color). So why use this, particularly poor, conversion? Notinasnaid 21:33, 28 April 2006 (UTC)

CMYK in Postscript is useful (and commonly used) as a way to convey device-specific color, i.e. to tell a RIP or other application not to do any color conversion -- probably because the conversion has already been done. In that case, the conversion documented in the Postscript spec is not used. (It might be used if you preview the image, but when you preview something in a device-specific color space, you're not expecting accurate color, hopefully.) --DavidHopwood 01:15, 29 October 2007 (UTC)

2. Once the color spaces and rendering intent are defined, clearly a calculation can and does take place (e.g. every time the screen is redrawn in Photoshop). But in this case I say there is no formula because nobody has expressed this as a mathematical formula, and it isn't really practical to do so. Let's look at what really happens. To oversimplify, the RGB profile contains information on the white point, gamma, and a matrix. This is used to convert the RGB value to Lab. So far, this would be practical to express as a formula.

The CMYK profile contains 3-to-4 dimensional lookup tables, up to three of them for different rendering intents. Each table is typically pretty large, though not as large as the 4-to-3 table used for going the other way. A color manager interpolates values in this table to convert the particular color. The table is typically generated by sampling colors with a measuring device. I don't see it practical to express this as a formula, nor every useful as each profile would be different.

The formula in the article is an attractive one, but it is based on a fiction. The fiction is that CMYK cyan and RGB cyan (similarly magenta) are equivalent colors. Notinasnaid 13:21, 5 October 2005 (UTC)

[edit] Personal Opinion Used in Technical Description

...The nearest equivalent in CMYK is an ugly purple shade....

The word "ugly" is not strictly appropriate for the context. If there are no objections I will replace this with a more appropriate description, and perhaps add an example of what it looks like, so as readers can draw their own conclusions about whether it is "ugly" or not.

Joseph Crowe 16:50, 6 January 2006 (UTC)

I agree the language is bad. Please don't add a visual example. You CANNOT show CMYK on screen, so you produce an entirely misleading idea that you can, by producing one particular conversion to RGB, which is itself monitor dependent. Notinasnaid 18:10, 6 January 2006 (UTC)

Right. I'll go ahead with the edit then. Joseph Crowe 15:30, 8 January 2006 (UTC)

[edit] Typo in history

I changed the text "a RGB color" to "an RGB color," but when I described what I did in the page history, I fat-fingered and put "'changed "a RGB color" to "an RGM color.'" I know this sounds foolish, but I just wanted everyone to know I hadn't actually changed the B to an M. Sorry about the flub in the page history. —CKA3KA (Skazka) 18:40, 13 January 2006 (UTC)

You don't have to apologize so profusely for this. Dysprosia 04:36, 19 January 2006 (UTC)

[edit] The K-channel image should be...?

According to the convertion,
RGB=(255,255,255) -> CMYK=(0.0, 0.0, 0.0, 0.0)
That is reasonable that there is no quantity of black color in the white color.
Relatively, the color Black (RGB=(0,0,0)) convert to CMYK is (0.0, 0.0, 0.0, 1.0)
So, if a image in RGB is converted to CMYK domian.
The K-channel image should be like the invert of image of luminosity.
But the K-channel image of the picture on the right side of this page, the CMYK-domain image, seems to be like the image of luminosity, but the invert of image of luminosity.
Is something wrong?

Ignore the formula. It's nonsense. There is no simple formula, and the idea doesn't even make sense unless RGB and CMYK are defines as specific color spaces. However, it is not the case that the K channel should be a mirror of the luminosity. K is there for mixing with the other inks for the desired affect. Consider an image that is entirely magenta. It has luminosity, but there must be nothing in the K channel. Notinasnaid 09:20, 31 August 2006 (UTC)

[edit] Drying

Does the pigment dry before the next color is applied or are they all applied wet and dried together?--Energman 22:34, 17 December 2006 (UTC)

Typically they are all applied at almost the same time, and so there will be some mixing of inks of different colors before they have dried. (To dry fully takes hours; here I mean before the initial "to the touch" drying.) The amount of mixing will depend on the absorbency of the substrate and the ink formulation. If the ink takes too long to dry (and definitely if ink from one pass has not dried before the next pass), then differently colored inks will bleed into each other -- especially black into areas of lighter color.

Note that the ink "lay-down order" -- i.e. the order in which ink from different color heads is printed -- has a quite significant effect on color reproduction: the color that is laid down last tends to be emphasized. The effect is consistent for a given lay-down order, and so color management can sort it out, provided that the order is always the same. (It might not be in bidirectional mode, unless the heads are in a symmetrical arrangement, e.g. "K C M Y Y M C K".)

For a printer using UV-cured inks (cured by ultraviolet light from a lamp that trails the print heads), the ink stays wet until the lamp reaches it, and so it's possible to test the effect of delaying drying by various lengths of time. It turns out that a delay of a couple of seconds is enough to have a disastrous effect on print quality, due to ink bleeding and spreading. For UV-cured ink, the delay time also significantly affects the surface texture of the print. --DavidHopwood 01:58, 29 October 2007 (UTC)

[edit] Other meaningsof CMYK

I don't think the Spanish record label needs to be mentioned at the very bottom since it's irrelevant to the color model.

It should be it's own article under CMYK disambiguation heading. 206.45.81.74 21:55, 12 February 2007 (UTC)

[edit] Useless conversions

The conversions on this page are practically useless, as they don't do anything like an accurate transformatoin from a display's rgb ⇔ a printer's cmyk. Their presence here only serves to confuse and mislead readers. In my opinion, they should be removed. Instead, some mention of CMMs and various choices in converting from one model to another should be made, including a discussion of dot gain, separations, etc. But the formulae should be left out. --jacobolus (t) 03:25, 23 April 2007 (UTC)

I agree. I have tried and failed in the past, because some editors insisted they were "useful": I recommend reading this whole page to get a flavour of the arguments on both sides. I observe a recent edit, however [1] which suggests that half of the information has always been wrong. This seems to destroy the idea of usefulness. Times and editors change, so can consensuses. Notinasnaid 10:49, 23 April 2007 (UTC)

Yeah, all the arguments made against your position are completely missing the point though. It's as if there is a lack of understanding that these formulae have absolutely nothing to do with a real conversion. They aren't even good as a first order approximation. It's quite unfortunate that the color separation article uses similar "invert the colors" logic. But maybe if some explanation was added about the choices involved in converting from RGB → CMYK, then it would be easier for other wikipedians to understand how bad these formulae really are. --jacobolus (t) 16:10, 23 April 2007 (UTC)

Okay, well if no one answers soon, I am going to remove the conversions section. --jacobolus (t) 15:04, 17 May 2007 (UTC)

[edit] THERE IS NOOO FORMULA

Hi guys... you're struggling with my every day hell, I am a color matching professional in a printing facility, what your are missing is that there is actually no conversion between RGB and CMYK, because both of them are DEVICE DEPENDENT. The same RGB values will display different colors on each device (just watch two TV sets together, the color looks different but they are outputting the same RGB values), it will also happen with CMYK specs each printer will print different colors usting the same CMYK values.

What is used to bring them together is called ICC profiles. This are "maps" measures the interpretation of specific colors of a device and generates a "map" to a theoretical colorspace called Lab (Luminosity, alpha and beta)

You can convert an RGB image from your camera and convert it to an RGB image for a projector or television. Also, you can convert a CMYK file that was destined to be printed on press and convert it to a CMYK that will be printed on a plotter.

What happens is that it's impossible to reproduce color consistenly without being device independent. The same values will give different outputs on different devices.

Lab colorspace is used between all the conversions, an RGB file will be "mapped" into a LAB format and then "mapped" again into a CMYK. You must provide the profiles for both of the files, (the ICC profile of the digital camera for example and the ICC format of the Printer), THEN you will have your conversion. The problem is that the RGB color space is an additive color reproduction, wich can output a far greater gamma of colors than the CMYK color space. So when converting from RGB into CMYK the colors are "shrunken" into the CMYK space, so a picture that looked really bright on the screen will look somewhat dull on paper.

There are several types of color space "shrinking", there is "perspective","relativer colorimetric" "absolute colorimetric" and "Saturation" depending wich kind of "shrinkage" is more desireable. --- Reuben von Insel

Regarding the Black generation, this is done by the converter of the spaces and is user-defined, depending on the type of press and the image the black generation will be greater or lower but that depends on the person converting the file, if it's got large areas of light gray, the black generation could be small if its got great areas of black, then more black generation is desireable.

check this page, this shows the two color spaces of RGB, CMYK and LAB http://back.iccview.de/index_eng.htm

Hi, and I beg to difer. I know I'm up against a "heavyweight" (hope I don't offend you there) here, but you have to be able to create a model for this conversion. Of course it's "device"-dependent, take a piece of pink paper and a piece of white paper, print one with RGB the other with CMYK, and "voila" two different devices.

My point is not to insult you, my point is that a model is no perfect solution, it is an APPROXIMATION (any phycisist will tell you this) to the desired ACTUAL output, in any given frame of reference. A model is only an approximation is only useful if it, within the same frame of reference, yields the same result BOTH ways. Low budget versus expensive paper, good vs. bad ink and comercial grade vs. home-office color printer. A little example of what I mean:

On my screen I have to make a model where "Input -> Model = Output" & "Previous output -> Inverse model = Previous input". That is the definition of a model. It's true the tranfer between the computer's screen to [insert color reproductive device here] is device-dependent. But do not confuse this with the model, of course the model has to fit generically to the desired output (magenta should not turn cyan), but imagine if the model was to include patterns or randomly miscolored spots on the paper as well. It's just not possible, or at least not feasable.

Now this next rant is not aimed at this sections original author, but a more general rant at some of the printed, and then scanned images posted here. Please don't post scanned images here as an example of how it really should look, it's the worst idea EVER:

Draw the nice perfect colors on YOUR device-dependent screen (cyan is defined to be white, completely without any red (#rr gg bb), ergo #00 FF FF:
Print the colors on your device-dependent printer
- What's the grade of the paper?
- Is the ink correct for that particular type of paper?
- Is the printer properly calibrated for this particular paper/ink combination?
Now scan the picture using your device-dependent scanner
- Is the scanner properly calibrated?
Finally place the picture on Wikipedia
I download the drawn-printed-scanned picture, from Wikipedia, to MY (uncalibrated) monitor (colors look nowhere NEAR Cyan Magenta, Yellow... but hey black looks ALMOST acceptable).
I print YOUR drawn-printed-scanned picture on MY device-dependent printer to see if you are right, and now OMG, the colors are WAY off. I do the same you did and claim that YOUR colors are way off, and MINE are the perfect value.

If you are not convinced then try printing your scanned image you posted here, scan the new picture, print that, now scan the latest printed image, and keep doing this a couple of times, and you'll see how perfect your example is!

Please, there is no way for me to stress this strongly enough. Theory is ONLY a model of the perfect world, not our REAL world. With a few changed parameters (I call them calibration), it ALMOST looks correct and it almost looks right, but it will never ever be 100% correct. A little perspective here, please. This article is about a MODEL, not screen to printer conversion... I'm not angry, just a bit miffed that this has not even been COVERED here yet!

—Preceding unsigned comment added by Zero2ninE (talk • contribs)

It's not clear whose old talk item you are responding to, but in any case there's no reason to be "miffed"; just work on improving the article since you have a clear idea what it needs. Dicklyon 15:20, 18 September 2007 (UTC)

I understand it perfectly... but actually the point of the model IS to be device-dependent. same CMYK values will print differently on different devices. same RGB values will display diferently on different Monitors. LAB values will display THE SAME on all devices (If they are properly calibrated). The color cyan which you stated "is defined to be white, completely without any red (#rr gg bb)" is being defined in an RGB model, again RGB is device dependent. it will be "cyan" in your computer and also "cyan" when printed, and if printed on a different printer will produce a different "cyan", however all will have different LAB values.

Wich you sated early is true, there is a model conversion between the two. And they should be aproximate, but as well as physics they are many methods to achieve that same aproximation, (which I stated earlier... perspective, relative colorimetric and so forth.

I mean they are made in two steps because for RGB to CMKY the process is RGB-LAB-CMYK and reversely is CMYL-LAB-RGB. You might be wondering "Hell if that's true then why don't they make everything in LAB".... Well the reason is that a file that is already in RGB doesn't need to be processed again by the computer to be displayed. For the CMYK model, the black generation information the "K" channel is usually kept apart.

Regarding the colors... of course if you print a color, then scan it again, and then printing it again you will have different results on each step, but that's because there is no perfect way of reproducing colors, as you say correctly, only an aproximation. the profile tries to aproximate those values more... but still they are not perfect. However, if you measure those values with a colorimeter (That will give you LAB values) you can see exactly how way off the difference is)

There is actually a formula for calculating color differences... it's called deltaE if i recall correctly and I think is calculated the following way abs(thirdroot(L*L*L + A*A*A + B*B*B)) (I don't now how to wirte wikipedia formulas... sorry).. but if you look it up you can see the formula..

anyway... its great to have real discussions with smart people...

Ruben von Insel.

Yes, everything you just said is accurate. What's the disagreement? --jacobolus (t) 09:15, 26 February 2008 (UTC)

[edit] Suggested merge from process color

These two articles are discussing essentially the same process and color model. They should be merged, and one should redirect to the other. --jacobolus (t) 04:03, 6 May 2007 (UTC)

I agree. And nobody has objected. Go for it. Dicklyon 23:54, 25 July 2007 (UTC)

I agree. —Parhamr 05:53, 26 July 2007 (UTC)

Okay, I did the merge. :) --jacobolus (t) 00:02, 29 July 2007 (UTC)

[edit] Original research / poorly worded

I removed this from the article, as it appears to be a combination of original research and a poorly executed explaination. 216.148.248.63 14:43, 16 May 2007 (UTC)

To perform a live example of this effect reduce the color depth of your screen to its lowest property and note the effect. Your computer will still show colors in RGB but the quality is far less. This commonly overlooked fact is why an RGB printer will have a different spectrum of colors it can reproduce compared to a RGB monitor. This is also why conversions based on color hue alone such as those shown below have absolutely no relevance or meaning to converting RGB on a monitor to CMYK on a printer.

By the way... they are NO RGB printers... only CMYK... that's the point of the model. -- Ruben von Insel.

[edit] tagged with "rewrite" template

A very large proportion of this article is similarly confusing for me. I think “combination of original research and a poorly executed explanation” pretty much sums it up. For instance:

A common misconception is that this color model is based on actual mixing of pigments, however when seen under a microscope you will see that the dots are printed next to each other and there is very little mixing. This is how a printer is able to reproduce rich red and blue from cyan, yellow, and magenta.

Why does mixing or not affect how "rich" the colors are?

CMYK works on an optical illusion that is based on light absorption. The colors that are seen are from the part of light that is not absorbed.

What does "optical illusion" mean in this context? Isn't this just a normal part of color vision?

To improve print quality and reduce moiré patterns, the screens for individual colors are set at unique angles.

This is not adequately explained. For non-experts, it is nearly useless, and the without explanation the associated table does nothing but clutter the article.

On the press: Though it varies by print house, press operator, press manufacturer and press run, ink is typically applied in the order of the acronym—cyan, magenta, yellow and then black.

Why is this deserving of an entire section?

A mixture of practical cyan, magenta, and yellow pigments rarely produces pure black because it is nearly impossible to create sufficiently pure pigments.

Is that really why? Would "pure" pigments mix to produce black? What does "pure" mean in this context.

so to reproduce text using three inks without slight blurring would require impractically accurate registration (i.e. all three images would need to be aligned extremely precisely)

Why is registration not defined before this, or linked? How is a non-technical user supposed to understand what it means?

The amount of black to use to replace amounts of the other ink is variable, and the choice depends on the technology, paper and ink in use. Processes called under color removal, under color addition, and gray component replacement are used to decide on the final mix, so that different CMYK recipes will be used depending on the printing task.

Why is this not explained more clearly in the article? These terms should be defined and described, not just linked.

Although the two acronyms have a similarities they are very different in many ways.

This sentence is so vague as to be useless, even for setting up subsequent sentences.

The whole “Conversions” section

Given that this section has no basis in reality, and is misleading, I think it should be scrapped.

Conclusion: this article requires a complete rewrite by authors who knows what they're talking about. I suggest that merging process color into here might go some of the way towards solving these problems. But it's still not enough.

--jacobolus (t) 14:46, 17 May 2007 (UTC)

[edit] A good book ref

I agree it needs a rewrite. Here is a good source to work from. I have the book in case you can't see enough online. Dicklyon 19:17, 29 June 2007 (UTC)

I didn't incorporate the good model from the book yet, but I did take out or rewrite most of the nonsense. So I took out the rewrite tag. Put it back if you think it's still needed. Dicklyon 05:40, 28 July 2007 (UTC)

Okay, I think my cleanup of the introduction, and addition of a "halftone" section helps a little bit. The article could still use a great deal of work, but as you said, at least it's no longer full of nonsense. :) --jacobolus (t) 00:03, 29 July 2007 (UTC)

[edit] error

Theres an simplification error in the conversion formula {1 − (C(1 − K) + K) =! {(1 − C)(1 − K) similar error in the other components value. It seems (1-C)(1-K) works to convert C to R though. —Preceding unsigned comment added by 72.185.40.123 (talk • contribs)

Given that the formulas are as far as I'm concerned completely useless, I'm not too worried about it ;). Anyway, feel free to fix it. --jacobolus (t) 21:07, 25 July 2007 (UTC)

[edit] Why don't we print with RGB or RGBK

Simple question. The only thing I could think of is that RGB can't be so easily be produced in very saturated quantities. So it would be expensive (or non existing). So it's more technical than physical. Right? (Please don't start about additive and subtractive 'cause that would only proove you don't inderstand - let me startle you: most printers print dot's next to each other: so though the CMY colors themselves are made by a subtractive process on a white sheet all the other colors are made by additive process: the CMY is reflected from the dots. So it it's CMY colored light that is mixed = additive. You could easilt replace the CMY dots with small CMY lights. The eye would not see the difference. So another question is why do monitors work with RGB. Again probably more technical than physical. BartYgor 14:11, 13 August 2007 (UTC)

No, you are confused. Although you can see the dots in color prints, the colors are in general on top of each other. The rasters in CMYK offset prints are angled with each other that some dots overlap while other dots end up between each other. Additionally, the grain of the paper smears out the ink. The net effect is roughly equivalent to smearing out a thin layer of ink as in conventional substractive mixing. Mechanically, it is not possible to get make the rasters to control exactly the amount of non-overlapping versus overlapping ink; this would require an insane amount of precision. You can actually often see in newspapers that the rasters for the different components are misaligned by more than 1 mm. If you were able to control the alignment of non-overlapping RGB ink dots, you would not be able to produce anything darker than a 33% grey (think about it). Han-Kwang (T) 16:30, 13 August 2007 (UTC)

Another response. It would indeed be possible to print with RGB inks; you'd want to make them somewhat pale, so that like CMY they each absorb only about 1/3 of the spectrum, not 2/3 like saturated RGB would. So just shift the spectral peaks around and go. But you are indeed confused about additive versus subtractive. When you put inks on white paper, the result is subtractive, whether they overlap or not. CMY works nicely because it's easy to convert from additive RGB. E.g. the C ink density is approximately 1-R. Plus the gamut works out better with these color, according to long experience; adding some Green and Orange inks helps to expand the gamut, illustrating the point that almost any set of inks can be used. Dicklyon 17:57, 13 August 2007 (UTC)

Indeed, there is nothing *special* about cyan, magenta, and yellow inks as primaries. They just happen to produce a nice gamut (the three colors are approximately evenly-spaced around a perceptually-uniform hue wheel, and mixtures do a reasonable job at making saturated red, green, and blue), without requiring too many inks. Like all technology, they are just a compromise. (I'm not sure though that ease of conversion between CMY and RGB has much to do with the decision: I'd have to see a source before believing that.) --jacobolus (t) 23:09, 13 August 2007 (UTC)

Thanks, everybody. I don't think I'm confused with additive; subtractive. But I do admit that I may be confused about what printing really is about. Though I do wonder if the small dots are so fast drying that they don't mix and if they don't mix, if they are translucent enough to act as layers on top of each other (subtractive color forming see http://www.rgbworld.com/color.html). If they mix then the only subtractive proces that is going on may be ink on white paper; or if the ink is not tranlucent just the color of the (mix of)ink is formed subtractive. In addition (as I do believe I know what subtractiev / additive is about - sorry if I sound pompous) in the first sentence it is stated "CMYK is a subtractive colormodel". This is incorrect: color models or spaces are never additive of subtractive. If you think it is: could you please describe, define when a colormodel is subtractive/additive. What properties should such a model or space have to call it subtractive/additive? --BartYgor 18:45, 14 August 2007 (UTC)

"Subtractive" is too simple a model, I think is what you're saying. That's true. See the article section on Neugebauer primaries, where it explains that the net color is caused by a mix of the light from the different bits of drops or spots. But that doesn't make it additive exactly, either. Dicklyon 21:37, 14 August 2007 (UTC)

You may be thinking of paint, that contains colorless `white' pigment (TiO2 particles) that reflects back the light. With paint it matters whether layers of paint are mixed. Ink is translucent and doesn't scatter: a photon either passes through or is absorbed forever. I can't comment on your remark about color models. Anyway, since the issue about overlapping dots comes up every now and then, I've added an actual close-up of a CMYK print so that you can see for yourself how the colors mix. Han-Kwang (T) 22:31, 14 August 2007 (UTC)

Thank you, I do acknowledge that colormodels mostly are specific to their application. Yet the application generally involves additive and subtractive processes. And almost always additive processes. Those colormodels predict what color you will see by applying primaries in that process. So please refraign from calling colormodels additive or subtractive, 'cause many wiki think that by any model based on CMY would be subtractive; yet as i stated this has nothing to do with the choice of primaries but with their application. For example http://en.wikipedia.org/wiki/RYB_color_model they are just talking about a colorwheel, this is neither subtractive nor additive yet they call it subtractive for some reason. Also the picture (right above) on the page shows that mixing RYB would give you some black. Wrong! It would give you some greyisch color with the same lightness as the three primaries. In this wiki article (CMYK) sentences like "Such a model is called subtractive because inks “subtract” brightness from white" are nonsence! Any color of a (no light emitting)object is formed by subtraction! So if I would mix non translucent (meaning regular pots) CMY paints and make up a colormodel that describes me what color i will get from what quantities of primary paint then this model would be subtractive?? Mixing paints is not additive nor subtractive; if the lightness of the three primaries are the same the result will have the same lightness! --BartYgor 10:30, 15 August 2007 (UTC)

About Neugebauer and mixing the light of spots. If I would put spots on black paper it would be additive, if i would put spots on white paper it becomes subtractive. I'm talking about the mixing of the lights of spots and paper. The colors of the spots themselves are made by a subtractive process if their non translucent. If they are it is the pigment color that is formed in a subtractive way. The color of the dot can be additive of subtractive according to the color (black of white) of the paper. --BartYgor 11:08, 15 August 2007 (UTC)

“If I would put spots on black paper it would be additive, if i would put spots on white paper it becomes subtractive.” – No, this is wrong. --jacobolus (t) 16:27, 15 August 2007 (UTC)

The Neugebauer method doesn't care what the substrate is, or whether the layered paints or inks are reflective or absorptive; you just characterize each combination of layered inks and their areas. Typical inks on black paper will leave you with black. Reflective RGB dots that don't overlap much might give you a sort of additive effect, but there's no way to get a decent white that way. Dicklyon 21:08, 15 August 2007 (UTC)

Of course I can't get white on black paper my primaries aren't bright enough. But with each spot on black paper I do add light hence additive. Whith each spot on white paper I take light (brightness, lightness) away hence subtractive.--BartYgor 21:35, 15 August 2007 (UTC)

The real answer to my question can be found here http://www.handprint.com/HP/WCL/color5.html#theorysub scroll further down until it starts to talk about subtractive primaries. I haven't read it completely through.--BartYgor 21:48, 15 August 2007 (UTC)

[edit] Why black is used

Again: wrong, wrong, wrong! If using only CMY you can only make colors with the lightness in between lightnesses of the primary colors. You can't make white or black. And if the primaries all have the same lightness you can't even change the lightness (if you mix CMY you won't get anything near black you will get a grey (ideally - maybe in reality some brown but not a dark brown!) of the same lightness as the primaries. But if you print on white sheets: you can adjust the lightness in one way: you can lighten the colors (in fact you are really entering CMYW color space). Now to darken the colors (darken is not really correct I mean lessen the lightness) you need black. It is not something out of covenience! Even if the primaries would mix in an ideal way (grey instead of brown), even if inks wouldn't soak the paper, you still would need black. With the black you enter in CMYKW space wich gives you a satisfactory gamut. (You sometimes see software that makes you alter CMY and if you put CMY full force you get black. To get the real life equivalent of this model I would have to imagine some monitor with CMY primaries and putting a primarie full force would be telling the monitor not to give any light of that primary. Shutting down all primaries gives you no light e.g. black. Make no mistake this model can be in no way be applied to mixing paint, printing with CMY so don't draw conclusions from it- maybe that's where the confusion arose?)--BartYgor 11:35, 15 August 2007 (UTC) So in short: CMYWK gives you a satisfactory gamut. But if you are printing on white sheets you don't need W as a primary, this leaves you with CMYK.--BartYgor 11:46, 15 August 2007 (UTC)

CMYW as well as RGBK are inocrrect.... you are substracting from the reflection... hence you should use black. on RGB you are adding to the reflection. --Reuben von Insel

No, you are incorrect. Printing 100% c, m, y gives a very dark color, and printing 0% of all three, on white paper, yields white. The gamut of just a three-color process is still three-dimensional, and roughly similar to the gamut of a four-color process (go try it out on your own printer if you don't believe me!) --jacobolus (t) 16:22, 15 August 2007 (UTC)

Again: how can I tell my printer to print equal amonuts of CMY and even alter the amounts. I'm no computer wizz who writes drivers. On this page http://www.handprint.com/HP/WCL/color5.html#theorysub it is explained how yellow and magenta is mixed (scroll down a bit you will see the wavelength figure). From the figure it is clear that lightness of the mix is the mean of the lightnesses of the base paints. I'll explain more basic. Imagine a red and green sheet both lettre size. I cut them up in small pieces mix the pieces put close together so no of the table colour gets through. What you are basically saying is just by mixing up the pieces lightness (the power of the light so to say, will change). I say lightness stays the same: the same amount of light is reflected if pieces are mixed or not. THis reasoning seems logical to me. Were do I go wrong? And if you define a CMYK model as a model that gives you the quantities of ink you need to produce a certain color then one color will be represented by 4 parameters (from 0 to 100%), 4 quantities. You are in 4D.--BartYgor 21:41, 15 August 2007 (UTC) You might be wright that printing 100%CMY gives you some black; but only in a printing process that uses tranculent inks and prints them on top of each other in layers and when the first layers have dried. Yet this has nothing to do with CMYK color model as a general explanation. Because you can also get black by applying enough layers of yellow on top of each other in this particular process. You would take this into acount in a particular colormodel for this particular printing process.--BartYgor 21:57, 15 August 2007 (UTC)

Again, the key is understanding the differences between inks and paints; inks don't reflect any light (almost); they just absorb; whereas paints mostly reflect, and obscure what's underneath them. But as pointed out below, this is not the right way for you to test your strange color theories or learn about color; read the aricle and some sources and let us know if you find conflicts between them. Dicklyon 23:55, 15 August 2007 (UTC)

The paragraph you wrote above is incomprehensible to me (check my user page for my background if you are curious). In any case, this discussion is resembling more and more a forum discussion. Check WP:TALK and WP:NOR. Please focus on the article, and in the case of a controversy (in this case it's you against three other editors) provide references for an alternative description. The only thing where I think you have a point is whether one can talk about a color space or color model being substractive or additive. Han-Kwang (T) 23:18, 15 August 2007 (UTC)

In explanation of the paragraph above: I have a red and green sheet size A4 (I'm European). I neatly divide the sheets in small pieces, and mix the pieces but the total surface is still 2xA4. The amount af recflected light from the 2xA4 surface can't change can it? So you can't produce black nor white this way. The process I describe is additive nor subtractive.This in explanation of my paragraph above--BartYgor 00:21, 16 August 2007 (UTC)

I didn't say the webpage I mention is accurate. I only want you to look at the picture. It is explained in more detail in http://www.handprint.com/HP/WCL/color3.html#mixprofile (this is a reference for you). And my simple point is: if this curve is correct in representing the result of a mix of two paints or inks (not applied as filters - just mixing) than it is obvious you never can get black from three colors this way, as the mean never can get lower the the lowest value of the three colors. I'm sorry that this seems more like a forum. But i'm only pointing out the mistakes. I guess this is the place. ON references: the first section of this wiki article and the why black ink is used section has no reference whatsoever. And I'm the one having to provide references?--BartYgor 23:56, 15 August 2007 (UTC)

First of all, it's not a mean; it says approximately the geometric mean; this is about what you'd expect for paint mixing, where you lots of reflective white bits and mixture of absorptive bits. But inks, the basis for the standard subtractive model, don't have the white bits, and you can put them on top of each other instead of mixing, so their densities add and you can get make a very dark result (not a great black, but a fair black). You're right we need to get some refs... Dicklyon 00:06, 16 August 2007 (UTC)

Yes it's a geometric mean but the argument holds. If you would use inks (or paints) as filters then you can get black. I have aknowledged this before. Though I do think that if the ink would be more absorbant than reflectant you would never be able to apply them on top of each other because one would block the other. I guess you mean inks are more translucent than paints. But to the point: http://www.tomshardware.com/2007/07/27/one_dozen_color_laser_printers/ specifically says that laser printers use CMYK color model and from http://en.wikipedia.org/wiki/Laser_printer it is clear the colors aren't applied as some filter on top of each other.So again the generakl understanding of CMYK color models is braoder than the specific process you are refering to.--BartYgor 00:33, 16 August 2007 (UTC)

The inks are transparent to some wavelengths and absorb others, and are applied right on top of each other (partially overlapping when less than 100% coverage). I don't see where the laser printer page suggests otherwise. Dicklyon 00:39, 16 August 2007 (UTC)

Laser printers, ink jet printers, half tone printers (http://en.wikipedia.org/wiki/Halftone) put the dots of ink next to eachother (and maybe partially overlap and sometimes on top of eachother). They don't put colors CMYK on top of eachother as filters. Imagine if you could choose if dots would more be on top or more next to each other in a printing process. Why would you choose 'on top': just because you can save on black? And with the downside that pictures would fastly become dark? And putting them more on top of eachother would require more accuracy. Look at http://www.micrographia.com/articlz/artmicgr/inkjetpr/inkj0100.htm--BartYgor 01:03, 16 August 2007 (UTC)

I would say the opposite; more often than not, the inks are on top of each other, at least parially, and act as filters, pretty nearly. That's the first-order model, and it's really only good to first order. The Neugebauer model, however, works for all combinations of partially overlapping and partially transparent inks or paints. Dicklyon 01:07, 16 August 2007 (UTC)

Come on: the dots are clearly randomly scattered. Rarely four inks CMYB are on top of each other. About halftoning, you might be right that they try to print the colors just next to each other so when more ink is used it's almost one big black dot. So yes there you wouldn't really need black.--BartYgor 01:38, 16 August 2007 (UTC)

Depending on the device dots will be scattered. They are two types of order that the dots can take...a pure screen which varies the size of the dots or a FM (Frecuency modulated) screen wich varies the frequency of the dots. actually yes... when more ink is printed it IS a big black dot it can even be completely black.--Reuben von Insel

Erm… it seems pretty clear to me that if you print anywhere near 100% coverage of all three inks (as you'd do if you wanted a very dark color without including black ink), they are going to be overlapping, not next to each other. Quite honestly, I suggest you try making a TIFF image (or PSD or whatever) in CMYK mode with 100% C, M, and Y, and then printing it on any printer. You'll see that you can get a color which is quite clearly darker than any of the three inks. --jacobolus (t) 00:31, 17 August 2007 (UTC)

I'm sorry Dicklyon but it is you who is talking about some specific printing process NOT the CMYK color model. Also the difference between inks and paints is irrelevant it is how you apply them: as filters (layers on top of eachother) as dots next to eachother or some mix.--BartYgor 00:01, 16 August 2007 (UTC)

The model is a good fit to the typical ink-based printing process, both on press and inkjet, as well as to color photographic material. It's less appropriate to paint mixing, but can be applied with a little modification. And watch your tone. Dicklyon 00:06, 16 August 2007 (UTC)

The model is only really used for the printing process (I'm unaware of other uses at least), so having the article discuss the printing process seems perfectly reasonable to me. Indeed, so much of the discussion of the process and the color model would necessarily overlap, that making them separate articles would be foolish IMO. --jacobolus (t) 00:34, 17 August 2007 (UTC)

Printing certainly is the "primary" use. A "key" plate doesn't make much sense anywhere else. Dicklyon 01:32, 17 August 2007 (UTC)

It's not a "key" plate.. its a "Black" plate... they used the "K" in order to avoid confusion with the "Blue" flim... before computers you needed to make three positives a "Red" negative, a "Green" and a "Blue" and then make three negatives "Cyan", "Magenta" and "Yellow", then the "Black" was made.. they used the "K", otherwise they would have two "B" "B" films. --Reuben von Insel

[edit] Structure of the article

Two concepts are mixed in this article. In my opinion the heading should read CMYK color models and it should explain that it's use is primarily in printing because to get a good gamut you would need CMYWK but because printing is generally on some kind of white paper the gamut can be produced using CMYK inks. Then it can be stated it is used in halftone printing or in spot color printing and maybe more printing technologies. I don't see why both types of printing technologies (halfton and spot color) are not treated equally. Then some alterations of the model can be discussed: CMYKOG aso. The article should clearly state that ever type of printing process and even every type of paper used would yield a different CMYK color model. Because to preduce the same color in different printing preocesses (or even with different papers) different quantities of the inks are used. It should also make clear that CMYK color models are 4 dimensional. It is not some (double) piramid. Though you can transform the 4 dimensional CMYK model into it (for example - CMYK to CIELab). The CMYK model shows the quantities of ink needed to produce a color; 4 inks; 4 paramters: 4 dimensional. In fact the very first thing that should be stated is "The CMYK color model describes the quantities needed of CMYK inks to produce a particular color."--BartYgor 12:27, 15 August 2007 (UTC)

[edit] Why we use CMYK in printing

Ok, I would like to conclude the above discussion (why not RGB and why K) with a full explanation. I admit I was partly wrong (but partly right) about things. But I can't say that other authors provided the correct answers either.

Printing and mixing painting is a mixture of additive and subtractive forming of color (see http://www.t2f.nu/s2p2/s2p2_ipq_13.pdf - expression 4 and 5). With additive it is ment that brightness rises (light is added) when mixing colors (for example adding colored light to another light). With subtractive light it is ment that brightness is lowered (light is subtracted) when mixing colors. Yet with paints and inks mixed the subtractive component is (much) stronger than the additive component (http://www.t2f.nu/s2p2/s2p2_ipq_13.pdf - see figure 1). To understand: think of ink or paint as pigments floating around in a substance. Two things can happen with the incident (day)light: 1. it hits a pigment (a part of the daylight is absorbed) and then is reflected to our eye. 2. it hits a pigment (a part of the daylight is absorbed), is reflected to another pigment (a part of the daylight is absorbed - a different part if it is pigment with different reflectance and absorption curves). It is now easily to see that the first will only happen with pigments on the surface of the ink bubble. The second (boucing of on several pigments) is much more likely with pigments deeper inside the ink layer. (See http://home.physics.wisc.edu/gilbert/publications/98.PDF fig 9). (Only if pigments would be very rare in the substance the first way will be more apparent or even dominant - see watercoloring below). If all the pigments have the same reflectance and absorption curves (all the same pigments), you get a nice bright color. Now if different pigments are mixed the mix would rapidly loose brightness of all the reflectance and absportion between the pigments. If you would apply the inks/paints not as a mix but even as filters on top of each other this would probably not make all that much difference. (own conclusion). With watercoloring pigments would be more rare and additive would become more apparent or even dominant (http://www.handprint.com/HP/WCL/color3.html#mixprofile).
Subtractive primaries. Now if we agree that subtractive proces is dominant than we can look for a set of subtractive primaries. Which properties should they need? 1. Bright (I don't mean saturated but high overal luminance)! If the inks are bright, they can take some loss of brightness because of the subtractive process. 2. Wide range of wavelengths. As a subtractive process subtracts wavelengths (by absorption for example) from the light it proves usefull to have some broad band otherwise one mix with onather pigment would get you immediately black. 3. Wide chromaticity gamut! (Which together with point 1 would deliver you a wide overall gamut). The colors can be chosen from a CIEcy chromaticity with the surface of the surface formed by the primary colors as big as possible. Which means choose saturated colors. From this point 3 we would immediately take bright RGB. But now look at http://home.physics.wisc.edu/gilbert/publications/98.PDF fig 3. There you can see that with an equal reflectance per wavelength CMY have more overal brightness than RGB because the cyan on the figure is not pure cyan wavelenght but all redcyanblue wavelengths! So by choosing a more broader reflection curve than a pure small reflection curve we get more brightness. (In fact we may not have the choice for pure CMY because Y and C are really small wavelength bands (http://en.wikipedia.org/wiki/Color) in the spectrum so you would need pigments that reflect very very specific wavelength bands - which may not exist but this doesn't add to the discussion); and for magenta: well magenta isn't a pure wavelength - is always a mix of RB). So to get bright colors it would be easier with a broad reflectance curve yet still a pure (saturated colour)for maximum chromaticity gamut. So we conclude that there is a tradeoff between brightness gamut and chromaticity gamut with subtractive processes. I can only assume that brightness has proved to be more important. Now why CMY? Surely we can make more saturated colors with a broad reflectance curve? Well yes but the best are CMY! Look at the The CIE 1931 color space chromaticity. The resulting perceived color of a reflectance curve can be found as the mass centre (centre of gravity) of the envolved (uniform - otherwise a weighted mean should be used)wavelengths on this chromaticity diagram. Now you can see that an ink with a reflectance from 700nm to 540 or even 520nm would be 1.very bright 2. very saturated - the mass centre of alle wavelengths would be on the straight line from 700 to 540 3. Be perceived as YELLOW!. Red wavelengths alone would also produce saturated colors yet it would be not so bright with the same reflectance per wavelength. The same thing as with yellow could be done with cyan from 380 to some 500nm. We could also do this with green 560nm to 500nm: the result would not be the most saturated green but it would be approximately as saturated as the just formed yellow en cyan. We could equally make magenta very bright and keep it saturated by combining wavelengths of 620-700 (red) with 470-380 (blue). So if we were the choose only three of these primaries I would choose CMY which would give you a very nice gamut. For a fourth primary I would add green because that would give you even a bigger gamut of bright saturated colors. So now we know why CMY are the best subtractive primaries. (Another way of explaining is given in http://www.handprint.com/HP/WCL/color5.html#theorysub (look for ideal spectral reflectance curves for subtractive primary colors) but this makes use of going back to the cones of the eye but this is not necessary because with the CIExy chromaticity diagram we incoporate this as it is based on how we see color).
Why we need black. So in an ideal subtractive process we would get black be combining CMY. Yet 1. mixing paint or ink is not purely subtractive there's also that small additive component which makes you don't get 100%black but a dark grey. 2. Well more a brown than a grey because also the inks are not the ideal subtractive primaries that you would wish. Yellow is high reflecting in 700 to some 520nm but not 0% in the other regions it is these smalle reflections that interact as the high reflections are cut off by the subtractive process. This is explained in fig 10 of http://home.physics.wisc.edu/gilbert/publications/98.PDF.
Don't we need white? Artists generally use white to give more brightness (but they loose saturation). (http://home.physics.wisc.edu/gilbert/publications/98.PDF p 316 - adding white). Well if a printing process uses white paper, white might shine trough the inks or through spaces between the droplets. So the CMYK model will take this into account for printing the right color (that's why you generally can choose which paper you use in the printer driver - but this also will have to do with absorption of the inks by the paper aso). If you where to print on a black sheet you would need white. -- BartYgor 12:06, 19 August 2007 (UTC)

OK, that's not too far off from how it works, though a rather idiosyncratic description. In particular, your notion of "pure cyan" or a "pure yellow" as if they were spectral colors is not really right. Cyan is not a spectral color; it's a broad spectrum missing red. Similarly, yellow, though it can be made, not vary satisfactoraly, as a pure spectral color, is really more generally a broad spectrum missing blue. Similarly magenta is a broad spectrum missing green. This is what you want in subtractive primaries: colors created by absorption in only a small band, like about 1/3 of the visible spectrum, for each primary. Dicklyon 06:37, 20 August 2007 (UTC)

I wonder if we could get some diagrams of spectral reflectance curves for some printer's C, M, and Y inks to add to this article. Any ideas where to go about looking for such a thing? --jacobolus (t) 09:42, 20 August 2007 (UTC)

I would suggest to draw up ideal subtractive primaries for the the wiki article about (subtractive)primaries. Then just mentioning here that inks a. are additive + subtractive b. are not ideal (not 0% in low region), will be sufficient (well to me anyway -yet more diagrams will help of course).--BartYgor 10:18, 20 August 2007 (UTC)

Thank you.

I would like to suggest to remove the above sections who started the discussion as they will only lead to confusion.
"idiosycratic" maybe but can you point me to ANY other explanation that in a (some) scientific way explains that CMY an why CMY are the choice for subtractive processes so thoroughly? I haven't found any such explanation and I would suggest that the wiki article about subtractive primaries (or primaries in general) would use the above explanation as a start. (or is my brain really that crooked ;-). Sorry if I have annoyed anyone but I really had to cook up the above explanation myself from bits here and there. I haven't found any such good explanation on the internet of subtractive primaries and the fact the painting and printing is a combination of subtractive and additive processes.
I use the word pure in two meanings (sorry) in the first part 'pure cyan' is light with a small bandwith of around 490nm (this was my mistake: I thought cyan paint would reflect something like that, while it really is a more broader spectrum). Are you saying that such light would not also appear as cyan to us? (In the rest of the explanation 'pure' means hihgly saturated).
One thing still: it is not good to talk about subtractive colormodels. 1.real world models are generally additive + subtractive 2. what would it's properties be? In other words if I place the origin of the RGB space on the point that represents black and the axes on CMY I would get the subtractive CMY space yet nothing fundamental changed about the space. none of it's fundamental properties are changed; so IMO you can't tell from the space itself if it is additive of subtractive: so again a colorspace is not additive or subtractive it is how you interprete it. If you disagree could someon please define the properties of such a additive /subtractive space (other than the origin is white or black). --BartYgor 10:18, 20 August 2007 (UTC)

Suggestion: BartYgor, why don't you look for some good sources and propose modifications based on what is verifiable? It still seems like you are trying to re-invent color theory on your own. Your conception of the subtractive primary colors like cyan was obviously all wrong, so you came to the conclusion that subtractive color doesn't make much sense; that's not really a supportable conclusion. If you want to understand it at a physical level, I recommend Judd & Wyszecki's Color in Business, Science, and Industry, 1952,1963, Part III, Physics and Psychophysics of Colorant Layers. Dicklyon 18:46, 20 August 2007 (UTC)

Thanks for the suggestion. I do think I gave alot of references but I admit that I had to put one and one together because, like I said I found no good explanations. I was partly wrong (I'm not scared to admit), and I may be still on the wrong track (so I do thank you for you patience). And in my defence there is so much misinformation on the internet that my own critisme and detuction was the only thing to go by. And also you ask me for references before I shoot but at least I do try it out on the talk pages. Many basic color articles on wiki use no references and are very wrong (but again it might be me :-). For example: do you truely think that de definition on wiki http://en.wikipedia.org/wiki/Additive_color is correct? THings like "It should be noted that additive color is a result of the way the eye detects color, and is not a property of light. There is a vast difference between yellow light, with a wavelength of approximately 580nm, and a mixture of red and green light. However, both stimulate our eyes in a similar manner, so we do not detect the difference. " No references, so I'm thinking logical: what is really said here is "additive color is the result of the way we see yet we don't see a difference" This is against elementary logic. My idea: Yes, there is a difference but 1. as we don't see the difference we can't talk about additive of subtractive color as the color is the same; the same perceived color can be the result of an additive or subtractive process but their is no way we can tell just by the color hence additive or subtractive color is bad english. 2. The difference has nothing to do with additive / subtractive as I can apply both types of yellow light in additive and subtractive processes. 3. The only correct definition of additive is that an additive process will makes the luminance rise and with a subtractive process luminance is lowerd. This to illustrate that the only thing I had to go on to get to the bottom was my own judgement. So I would practically have to take on tens of wiki articles to get things straight. And if you now say that the wiki article is right without any explanation or even what your defintion is then I think that I will leave it to that because that would mean that for any sentence changed in wiki I would have to take on a similar discussion ... and I still believe that science is not done by democracy. --BartYgor 14:31, 21 August 2007 (UTC)

Sadly, you are right that too many wiki articles are too short on refs. I've done a lot to help that problem, but I still need tons of help. So, since you're interested, get some books and chip in. I would argue, however, that the statement above that you think is "against elementary logic" is actually pretty nearly correct. Additive color works just because your eyes don't distinguish between a wavelength that looks yellow and a pair of wavelengths that look red and green (this is very nearly true, not necessarily exactly true). Dicklyon 19:52, 21 August 2007 (UTC)

I'll try to help. About additive color: you can say "we can't see the difference if a color is a result of a subtractive or an additive process" and of course "color is the way we see , perceive light". If you admit that we can't see the difference. In my logical thinking: if we can't see the difference is saying it is unrelated. Maybe I'm wrong. Could you explain in full that "additive color is the result of how we see"? I mean, if that is true it would mean: if we would see differently (maybe animals, bees aso) then additive would be /mean something else? So if you think the explanation in the wiki article is correct you should be able to give me an example as proof that a different way of seeing would redefine additive color or the color of an additive process. IMO additive is independent of how we perceive the light (=color), it is just a manipulation of light that add luminance / brightness / energy.--BartYgor 15:41, 22 August 2007 (UTC)

The easiest example is to compare human dichromats ("color blind" people) and trichromats ("normal" color vision). Both of them have additive colorspaces, using two or three linear sensing cell types. It is easy to find different spectra that are indistinguishable to the dichromats, but different colors to the trichromats. The same is true between any pair of animals whose spectral sensitivity curves don't span the same subspace of spectral space. Dicklyon 17:17, 22 August 2007 (UTC)

This is just saying: the perception of color is related to the way we see. Where does the additive come in? (It's like you're saying if it's a color we can see it's additive color. And subtractive color would then be the color we can't see?)--BartYgor 23:52, 22 August 2007 (UTC)

I think I understand what you're getting at. I can agree that if we would see the full wavelength spectrum of light, we would not need the concept of additive or subtractive. That I can second. But to use that as a definition or explanation would be like explaining "colorspace" with "a colorspace is the result of the way we see". Yes, because every conecpt about color has to do with how we see light. I can't say it's incorrect but in a first explanation I think I would rather read that a colorspace is a way of organising colors in some systematic way. All concepts are a result of the way we are, be it temperature, space, time, naming things, understanding the world around us ... --BartYgor 00:41, 23 August 2007 (UTC)

Don't be seconding things that nobody said. Get a book, and read it, please, instead of using the article talk page as a forum to try out your ideas. Dicklyon 00:45, 23 August 2007 (UTC)

So how is 'additive color' more (in another way) related to the way we see than just 'color'? (At least I use the talk pages - not complete wiki articles without any references)--BartYgor 00:49, 23 August 2007 (UTC).

At least I have a reference for my definition: "Additive Color - Color created by superimposing light rays, adding(superimposing) the three physical primaries (lights) - red, blue, and green - will produce white." http://www.khsd.k12.ca.us/bhs/Perry/art%20vocabulary.htm.--BartYgor 00:51, 23 August 2007 (UTC)

Can we have a reality check here? Just try printing CMY without any K component, on any printer that allows you to do so. Unless the CMY inks are really poorly chosen, or the driver and/or firmware are doing too agressive ink limiting, the result will be black. Not "muddy brown", and not "dark gray" -- except to the extent that any black is a dark gray. (The proportions of C, M and Y needed may not be exactly equal.)

More generally, the subtractive model is a very good one. Like any model ("the map is not the territory"), it has some inaccuracies. That's why we need big look-up tables, in the form of color management profiles, to get more accurate results. The generation of these profiles relies on the subtractive model working as a fairly close approximation.

If you want a more accurate model without resorting to look-up tables, see the Neugebauer and Demichel equations. Unfortunately the WP entry on that is just a stub. --DavidHopwood 17:08, 30 October 2007 (UTC)

[edit] RGB to CMYK MORE

The article states:

Since RGB and CMYK spaces are both device-dependent spaces, there is no simple or general conversion formula that converts between them. Conversions are generally done through color management systems, using color profiles that describe the spaces being converted. Nevertheless, the conversions can not be exact, since these spaces have very different gamuts.

There is a serious mistake here. As others have stated above, if we presume 100% blacK removal, there are simple conversion formulas between the two ways of representing the same color.

The serious error is the confusion of two issues which are:

Conversion from RGB to CMYK color representations

Conversion from one color space to another

Color space is device dependent, color representation isn't.

Tyrerj 23:35, 11 October 2007 (UTC)

Do you have a source for this viewpoint? If so, we can represent it in the article. You can also place a fact tag on the part you consider to be a mistake, if no source is cited there, and if a source is not provided in a reasonable time, you'd be justified in removing it. I don't agree with you, but a little action to force us to get a solid referenced statement there would be a good idea. Dicklyon 01:07, 12 October 2007 (UTC)

[edit] CMYK jpgs

The streams embedded in pdfs are sometimes jpgs (/DCTDecode) that can be extracted by hand into a file and viewed. But they are not always RGB; they may be /DeviceCMYK -- 4 bytes per pixel instead of 3? Which image viewers can properly handle CMYK jpgs? Irfanview sort-of recognizes the file type but shows false color, and then saves as a wrong RGB. Image Analyzer says it can handle CMYK, but does not seem to do it by default -- maybe with the ImageFileLib plugin? -69.87.199.97 14:22, 19 October 2007 (UTC)

Good questions; why don't you go ask in an imaging forum some place, find some sources, and come back and add anything relevant to the article? This page is for discussing the article. Dicklyon 15:44, 19 October 2007 (UTC)

You might try GraphicConverter if you have a Mac available somewhere. It generally knows how to open every image file known to man. :) --jacobolus (t) 16:07, 19 October 2007 (UTC)

[edit] Why CMYK?

First and foremost: I am new to wikipedia, so if my editting/coding is horrendous I do apologize :)

But how did Cyan, Magenta, Yellow, and Black get picked for printing? How were these 3 colors (black not counted since I find it quit self explanatory) decided upon for printing? I find the article explains what CMYK does, but not why CMYK is the way it is. Did some random guy sit in a lab and pick 3 colors on a chart and decide they should be used for printing?

Bvlax2005 09:20, 7 November 2007 (UTC)

I don't have a description of that history at hand, but I'll tell you what I think I recall. Basically, the "process" colors evolved over time, probably starting with red, blue, and yellow, until color theory developed enough for people to start understanding what makes good subtractive primaries, and they renamed process blue to cyan and process red to magenta, to recognize that the colors that work well are a long way from the additive primaries. The colors used are certainly not unique, but I think you can see how they are "logical" in each absorbing one-third of the spectrum. Black was added for lots of good reasons already discussed; or, historically, black was there already and the color were added. Dicklyon 15:35, 7 November 2007 (UTC)

They didn't evolve... our eyes only see three colors "Red" "Green" and "Blue" and their combinations. "Cyan" "Magenta" and "Yellow" are the negatives of these colors. If add to a piece of white paper yellow and magenta.. you'll get RED.

yellow and cyan....GREEN cyan and magenta...BLUE. That way you can have all the colors you want.

on the other hand.. if you add RED and Green ink you'll get a dark grey... same with any other RGB combination... RGB inks cannot work with each other when printed... they can work with each other when emmiting light.--Reuben von Insel —Preceding unsigned comment added by 189.163.196.120 (talk) 07:26, 26 February 2008 (UTC)

Here's a book showing an intermediate state of development of subtractive color terminology, with "cyan blue" and "magenta red". Dicklyon 15:39, 7 November 2007 (UTC)

[edit] How to pronounce this Kmuc?

Ha? Thx! —Preceding unsigned comment added by 77.122.114.34 (talk) 14:48, 19 November 2007 (UTC)

[edit] Origination of the last letter in the CMYK abbreviation

Answering why the abbreviation originates from "cyan, magenta, yellow, black".
When the subtractive CMY approach had been decided on as one best fitting white-paper printing purposes (it would be RGB if most of paper were black) and the problems of ink mixing and ink economy had arise, the only accompanying color could be and was chosen is the black color for the reason that it is the only antipode to the white color and this way is the best color of contrast for a white sheet of paper. Next, the abbreviation CMYB was assembled (still being encountered), but seeing that "B" may be confused as standing for "blue", like it was for the existing RGB color model, the last letter "K", color-unoccupied at the time, was picked to represent the color.
Those folks who have any doubts can google the text cmyk "cyan, magenta, yellow" [2] and for several result pages compare by themselves how much predominantly this color sequence ends with "black" making out something like "CMYK stands for cyan, magenta, yellow, and black.". So almost everywhere, including this Wiki article, when it is placed "K", it is meant black and only black.
In addition, here are some quotes and references:
"Short for Cyan-Magenta-Yellow-Black" [3]
"CMYK - Cyan, Magenta, Yellow, blacK" [4]
"The C stands for cyan (aqua), M stands for magenta (pink), Y is yellow, and K stands for black." [5]
"CMYK is an abbreviation for Cyan, Magenta, Yellow, and blacK; the primary colors of a printer's pallette." [6]
--SunnySideOfStreet (t) 20 November 2007

Do you have any more credible sources? Perhaps some book about printing processes, instead of entries in random online glossaries? --jacobolus (t) 13:44, 20 November 2007 (UTC)

That "k" stands for "key" is pretty well established (see, for example [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], etc. Here's what Yahoo Answers has to say. Note that indeed, "key" and "black" are basically interchangeable, and calling this the "cyan, magenta, yellow, black" color model is perfectly legitimate. That doesn't mean that "k" stands for black though.) --jacobolus (t) 13:50, 20 November 2007 (UTC)

Thanks to your first Adobe-related book examples, Jacobolus, I finally have remembered where I read the "story": it's from Adobe Photoshop Help — I'm sure a very reliable source hardly can be rejected. So, Adobe Photoshop CS Help says (a paragraph): "In theory, pure cyan (C), magenta (M), and yellow (Y) pigments should combine to absorb all light and produce black. Because all printing inks contain some impurities, however, these three inks actually produce a muddy brown. For this reason, black ink (K) is used in addition to the cyan, magenta, and yellow inks in four-color printing. (K is used for black instead of B to avoid confusion with blue.)" [17] End of the quote (the bold is mine). Does anyone see a "key" in it?

Now about books. I still think that Google's statistics [18] is really noteworthy, but if you like books, then Ok, here are the book examples: [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], and so on. But I wouldn't want the discussion to come at "who is to give more examples", so the next I hope will end the book theme. Here are the comparison of "key"-vs-"black" results from Google Book Search ("All books" mode):

phrase cmyk cyan magenta yellow key
books found: 275

phrase cmyk cyan magenta yellow black
books found: 660

660 of "black" against 275 of "key". But you may rebuke that "black" is just more frequently encountered word than "key" as regards the subject of CMYK color model, and where is "black", there may be "key" as well. So here comes the final experiment on Google Book Search. Concerning CMYK, let's find cyan magenta yellow black combination but without key:

phrase cmyk cyan magenta yellow black -key
books found: 650

I will do my edition for the last time and leave the discussion, but before you, Jacobolus, decide to redo yours, wait for and pay attention to somebodies' extra opinions. --SunnySideOfStreet (t) 20 November 2007

You completely miss the point. CMYK does indeed imply "cyan, yellow, magenta, and black", but the letter "k" is used because black was the "key" plate, not because black has a "k" in it. This is not a matter of opinion, but is simply a well-documented part of the history of printing. --jacobolus (t) 17:18, 20 November 2007 (UTC)

the "Key" plate was used to change the language on the books... images remained the same but the text changed with languages.... soooo a "key" plate was made... this process had actually five plates "CMYK" for printing the images (Black is needed in the images).... and the "Key" plate to print the texts in different languages and to separate text from images. --Reuben von Insel —Preceding unsigned comment added by 189.163.196.120 (talk) 07:30, 26 February 2008 (UTC)

Just because "black" has a "k" in it. See carefully [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48]. --SunnySideOfStreet (t) 20 November 2007

The "K" comes to differenciate it from the "B" (Blue), before computers you needed seven fimls the R(Red), the G (Green) the B (Blue) and then from those you made another three C (Cyan) M (Magenta) and Y (Yellow), then you took a picture of the original with the text and came up with the Black... but then you would have RGBCMY"B"... and to avoid confusions the black was labeled "K". then you used the CMYK films to "burn" (with uv-light) the plates that were later used in the offset press. to print color with good quality you need four plates.

- - ::::There was sometimes another plate called the "Key" plate and was used to make quick changes on the press. You would print 100,000 brochures with the same pictures, and then change the "Key Plates" to print 10,000 for German customers and another 90,000 for the american market. Printing used to be very expensive and using a "Key" plate to change text was more cost effective... to make it short. There was a "Key" plate... but it it wasn't the "K" plate... That's a common misconception with graphic designers that don't really know what's going on on the printing business --Ruben von Insel —Preceding unsigned comment added by 189.163.196.120 (talk) 07:42, 26 February 2008 (UTC)

Of the books you just linked to, 3 suggest that the "k" comes from the term "key plate", 2 suggest that it comes from the last letter of "black" (specifically, the first and sixth links, Digital Imaging: Essential Skills and Computer Graphics: Theory Into Practice), and the rest fail to mention the derivation of the "K" in CMYK at all. I'm not sure I find these two sources credible, however; both are quite simplistic, and are mainly about digital imaging, only briefly discussing CMYK. For those unfamiliar with printing, "k" as the last letter of "black" is a reasonable assumption for the derivation of the name. I place more trust in Handbook of Print Media: Technologies and Production Methods (your second-to-last link), which explains various printing processes in great detail, and explains quite clearly on page 82 that "It is known that in subtractive color mixing the color "black" (letter symbol "K" for "Key Color") is produced by the superimposition or the overprinting of the colors red (R), green (G), blue (B) or cyan (C), magenta (M), yellow (Y). PostScript's architects had assumed that a combination of the same proportions of the three primary colors would produce an ideal neutral gray and that consequently the largest possible proportion of black in the color separation could be calculated from the largest common proportion of primary colors. These assumptions were so far removed from actual practice that during activation of the first generation of PostScript color printing systems unsatisfactory color print results appeared. This was rectified in later versions." Which (along with the following couple of pages) would be a great source for the expansion of this article. --jacobolus (t) 17:50, 20 November 2007 (UTC)

So, for those keeping score at home, we've got at least 15-20 sources, including some hard core printing books, claiming that "k" comes from "key plate" or "key color", against a few web glossaries no one has ever heard of, and a couple of digital imaging books, claiming that it is from the last letter of "black". I'd like to get a definitive source saying something like "it is a common misconception that 'k' comes from the last letter in 'black'", but I doubt such a source is going to turn up. If you want though, we can maybe add a footnote about it to the article, as you seem adamant that the "k" from "black" theory be represented. --jacobolus (t) 17:57, 20 November 2007 (UTC)

Hah! Ask and ye shall receive: Getting it Right in Print: Digital Pre-press for Graphic Designers: "It is a common misconception that black is assigned the letter 'K' because if it was called 'B' it could be confused with blue. While plausible, this is not the case". --jacobolus (t) 18:01, 20 November 2007 (UTC)

The first three book examples say that "k" originates from "black" explicitly: Digital Imaging: Essential Skills, Artist's Color Manual: The Complete Guide to Working with Color, and Video Demystified: A Handbook for the Digital Engineer, and also the sixth Computer Graphics: Theory Into Practice and the eighth one Rapidex DTP Course (DTP is Desktop Publishing). The remaining books decode CMYK without use of "key" (at all) but placing "and black" after "cyan, magenta, yellow". It's an opinion of the author of a book, yours or mine (you can write a book, I can do it as well). But it's not a matter of source/book reliability -- it's a matter of statistics of the most prevalent interpretation. See with consideration the Google examinations I had done. I cannot waste my time for this no more.
--SunnySideOfStreet (t) 20 November 2007

I'm sorry, but the statement that "CMYK stands for cyan, magenta, yellow, black" is completely correct, and does not indicate anything about the derivation of the "K", one way or the other. You are right however that your second link, which seems to be mostly directed at painters, with a brief explanation of process color, used partly as a lead-in to explain why pointillism works, also claims that "K" comes from the last letter for black. I really don't think this source is as credible as the several technical printing books which claim otherwise, but you're right: You tracked down 4 books which claim that "k" comes from the last letter. --jacobolus (t) 18:42, 20 November 2007 (UTC)

Here's a source from 1872 (entitled Photo-engraving: A Practical Treatise on the Production of Printing Blocks by Modern Photographic Methods) which states: "For color work the design of the black or key plate is generally transferred to the surface of the zinc…", and one from 1910 (Colour Printing and Colour Printers) which says: "…only a few prints seem to have been produced in Paris by Le Blon himself, who died in … 1741. … His own work, at any rate in the earlier years, seems to have been strictly on the three-colour basis, but there has been a good deal of controversy as to whether or not he used a fourth or black plate as a key for the colours." And here's a 1903 source about Le Blon (American Architect and Building News): "He engraved a separate plate to carry each color, and used red, yellow and blue, with sometimes a key-plate in neutral tone." (Of course these sources are all from a time before one-letter abbreviations seem to have been used for any of the plates. That is, the abbreviation “CMYK” wasn't invented until later). --jacobolus (t) 19:32, 20 November 2007 (UTC)

SunnySide, the big difference between your sources and Jacobolus' sources is his are much older some going as far back 1800's. Which in the case of the original meaning of a word or acronym is more valid than the a newer source. Wikipedia is not about finding out what most people think about a particular issue which is what you have done, but instead finding out the truth of the matter regardless if it is well known or not. 68.50.168.239 (talk) 00:53, 22 November 2007 (UTC)

[edit] need for a history section

All this discussion of the origin of the "K" in CMYK has me realizing that this article (and the color printing article as well), desperately need some expert attention, including adding lots of sources, and creation of history sections. This 1896 article “How to Study Process Chromatics” on pages 202-205, 248-250, 322-324, 393-395, and 436-439 of Wilson's Photographic Magazine, seems like a pretty decent source on the printing process of that time. --jacobolus (t) 19:44, 20 November 2007 (UTC)