God created RGB. Man created CMYK. Which would you rather use?
I found this quote in an Andrew Rodney's article in his website The Digital Dog (Tips > March 2007: CMYK Part 1, look for a PDF just after middle of page). I think it briefly describes the core of the troubles we face when doing the type of conversion mentioned in the title. Next paragraphs are some thoughts about this and some rules that may prove useful.
Photoshop let us make "unpunished" conversion of images between color spaces. Bearing in mind that every color space has its own properties and limitations, this process should be done on an image very carefully and respectfully. Some types of conversion warn us about severe changes in the edited image: for instance, any conversion from color to grayscale. Nevertheless, some conversions may seem harmless if one judges them just for what we can see on a monitor screen. In this category, the list tops with the famous (and no less uncomfortable) RGB to CMYK conversions.
Once we learn that colors in print will be ultimately described by means of some form of CMYK, we may get tempted to do our own conversions, because a) we know what we're doing; b) it's not at all complicated as long as the converted image looks just right in my monitor; c) I can save time to printer who needs CMYK images in any case.
Despite of all that, I think it's necessary to call for reflexion about this kind of processes, so I suggest to follow two simple rules for image RGB to CMYK conversion:
- Rule 1: DO NOT convert to CMYK.
- Rule 2: If it is not possible to apply Rule 1, then DO NOT convert to CMYK yet.
A reason to be so drastic? Certain issues with the CMYK system; let me be more specific.
CMYK as a color communication system
In color management jargon, a CMYK color system is called device-dependent color, meaning that if we want to talk about a certain color as represented by four percentages of four inks, this description is incomplete unless we specify at least three other things: the printing technology used, the ink type used and the media it will be printed on. (In fact, RGB is device-dependent color as well, although we can expect more uniformity when seeing the same image in different monitors than looking at prints made in different devices. Top reason: current monitor technologies are quite similar; main differences are due to different monitor settings, specially brightness and color temperature).
For instance, two different color patches are shown below. Each one describes ("means") a specific color in a certain printing system; let's take a look at what they "mean" in four selected and different printing systems:
- SWOP (Specifications for Web Offset Publications), version 2;
- European Standard for coated paper (Euroscale Coated), version 2;
- Tektronic Phaser III print, according to its data sheet;
- Epson 4800 on semi-matee proof paper, according to measurements on a real printer.
|50%, 40%, 40%, 0%|
|25%, 50%, 100%, 0%|
Notice: colors shown here are close to printed ones in each system if you have your monitor properly calibrated and it closely matches sRGB standard. There's no way (yet) to make sure colors shown in a browser are accurate, even with a calibrated monitor, unless web pages start to make use of color management, not only in images but in solid colors as well (see El color en la Web), although this simulation let you get an idea of the relative differences among systems.
This "experiment" shows us how a particular CMYK system bind us to a specific numeric description of color, which in general won't be valid on another CMYK system.
Why is it better to postpone CMYK conversion?
Those unaware of classic printing problems might think that using RGB or CMYK to describe image color is irrelevant, because we can establish certain "symmetries" between color described in each system; in fact, applications that handle color in a primitive way use the rule "CMY is one minus RGB", in the sense that a primary color in one system is opposite to a primary color in the other one. But as we already know, that would be true only in a world where a cyan ink reflects absolutely no red from white light (same goes to magenta with respect to green, and yellow with respect to blue). Since those imaginary inks do not exist, conversion to CMYK is more complex than a simple change of units.
On the other side, presence of a black component completely alters this supposed symmetry. Basically, adding black let us replace equivalent quantities of color inks to achieve neutrals with the purpose of cutting costs (by lowering color ink usage), improving quality (by less demanding ink absorption requirements) and getting better neutral gray tones (by not requiring a careful balance of color inks to achieve those neutrals). Besides, there is more (much more) than one way to create a black component, so more conditions are needed to completely define it; e.g. specifying a maximum ink quantity, which is the top limit of C, M, Y, K combined in any point of an image.
Another important fact is even when CMYK is numerically bigger (has more "color combinations") than RGB, that does not mean it has more colors; actually it has less colors (way less). This is usually expressed as CMYK having a smaller gamut than RGB, and consequently many RGB colors do not exist en CMYK.
Last: a CMYK image is a taylor-made suit for a certain printing system; it's not easy to rearrange that suit for adapting to someone else..
Having these ideas in mind, it is time to answer the question asked analyzing the kind of work usually done on an image:
- Global color correction: when image presents generalized color cast (lack or execess of any of them), global correction is required;
- Retouching: removing imperfections or artifacts, detail sharpening, or other kind of editing not involving global color correction;
- Artistic work: cropping, layers, filters, effects, etc.;
- Image preparation for a specific output.
Only the last task requires conversion to CMYK (assuming conventional four-color printing). In this case, conversion is needed to assess the final result of printing the image and make adjustments to compensate the loss of contrast usually found after this type of conversion.
Besides, the other three tasks don't benefit from an early conversion. Much on the contrary, a CMYK image requires 33% more memory than its RGB equivalent, and depending on the amount of RAM available it may take longer to process. Regarding artistic work, many filters and effects only work in RGB and would require convert images back in order to use them. These are enough reasons to delay conversion until strictly necessary.
Despite of the above, some people may find an advantage converting before retouching, because color correction in CMYK is more intuitive than RGB for them. Nevertheless, it is always better to make first global adjustments in RGB (specially dynamic range and color cast) before going CMYK.