Out of Gamut: Finessing Photoshop Color
This month BruceRGB celebrates its second birthday. For those who haven’t heard of it, BruceRGB is my attempt to provide the world with a reasonably safe RGB working space for use with Adobe Photoshop 5.0 and later. You can download a complete description of BruceRGB, along with a critique of the various working spaces that shipped with Photoshop, at my Web site. However, a recent, lengthy thread on the Epson inkjet users list shows that there is still plenty of confusion surrounding the topic of Photoshop working spaces. As an upshot, some people who probably shouldn’t be using BruceRGB are, and some who probably should be aren’t.
First, let’s kill some myths.
- There’s no such thing as an ideal working space: There are always trade-offs involved. If you understand the trade-offs, you’ll be in a much better position to make an intelligent choice of working space.
- No working space will, by itself, magically make your prints match your monitor, or your monitor match your original. To match up your colors, you need to use the appropriate ICC color profiles for your capture device, your output device, and your monitor, to convert the data going into or out of the working space. The working space is simply a safe place for your data to live as you take it from capture to output.
- BruceRGB has not been revised since its inception (an early version of NikonScan, Nikon’s scanner plug-in, implemented BruceRGB incorrectly, but this has been fixed).
A Study in Contrast
There are two philosophical extremes in digital imaging, which we’ll call input-centric and output-centric. The input-centric philosophy seeks to preserve as much of the original capture data as possible, even though much of it lies outside the range of colors that printers can produce. In contrast, the output-centric philosophy values only colors that can be printed, and it tries to bring the image into the printable gamut as early in the process as possible.
Photographers who print their work to a wide range of outputs tend to be input-centric, as do museums and archivists. Their thinking goes something like this: “We may not be able to print all the colors now, but perhaps someday we will.”
The prepress business, on the other hand, has always been relentlessly output-centric, scanning directly into output CMYK and letting the out-of-gamut colors fall where they may. Their thinking: “If we can’t print it, what use is it?”
This philosophical difference has contributed in no small part to the tension that often exists between photographers and the prepress folks charged with printing their images. There are plenty of shades of gray between the two extremes, and in practice most Photoshop users come down somewhere in the middle. But they often don’t realize that their choice of RGB working space is one of the main factors that dictates where they fall between the two extremes, because the working RGB space defines the palette of available colors for their images.
The Gamut Trade-off
Most digital imagers have now grasped the idea that the main difference between the various working spaces out there is the color gamut — the range of color — they represent. So why not simply choose the widest possible working space, so that all your colors are preserved as long as possible? That’s where the trade-off comes in.
Typical RGB images are made up of three 8-bit channels — one each for red, green, and blue — and each of these can define 256 (and only 256) shades. When you edit an image, you change the RGB values of pixels with whatever software you use, hence altering the pixels’ tone and color. But you can only alter the pixel values in the 256 increments imposed by the 8-bit channels: A pixel can take on a value of 128 or 129, but never a value in between.
The size of the working space’s gamut determines the spacing of the 256 possible values of each channel. In a large-gamut space, the values are spread farther apart than in a small-gamut space. Hence the trade-off: A wider gamut gives you a wider range of color, but it doesn’t give you more colors. The same number of colors are simply stretched over a larger color range. This means you don’t have as fine control over the color as you would in a smaller-gamut space. That’s one of the main reasons we don’t recommend doing everything in Lab color, which encompasses all visible colors: With such a wide range of colors, Lab makes for a very wide spread in data points indeed. More practically, if you need to make large editing moves, you’ll find that your ability to do so without introducing posterization or color banding is more constrained by a large-gamut space than it is by a small-gamut space.
In a smaller-gamut space, you have finer control over color and tone, because the data points are packed closer together, but you lose the ability to specify the very saturated colors that would be available in a larger-gamut space.
Note that you can avoid this trade-off entirely if you work with 16-bit channels instead of their smaller, 8-bit counterparts. But with this approach you run into another trade-off: Your files become twice as large, and to make it worthwhile you need a capture device (a scanner or digital camera) that provides at least 12 bits of data per channel. If you’re willing to make this trade-off, using a large-gamut working space probably makes sense for you, and we’ll discuss this in detail in a future column.
It’s Not Just the Size
Gamut size isn’t the only important factor; You also need to consider gamut mismatches. For example, we typically think of the monitor as having a larger gamut than an offset press, and this is true. But the press can print colors (even with CMYK inks, leaving fluorescents and metallics out of the picture) that your monitor can’t display. Ideally, the RGB working space should encompass the gamut of your eventual output, without also containing large numbers of colors that you can’t print. In practice, this is hard to achieve, which is why there are so darn many RGB working spaces in the first place.
I designed BruceRGB from the outset as an output-centric RGB working space. In short, if your goal is to retain as much of your original color information as you possibly can, you shouldn’t be using BruceRGB (and if you really want to keep all the color on, for example, E6 transparency film, you need a very large RGB space that will almost certainly require 16-bit channels).
BruceRGB is essentially a compromise between two spaces shipped with Photoshop 5.x — ColorMatch RGB and Adobe RGB (1998). ColorMatch RGB is a high-quality monitor space, but it is a monitor space nonetheless, designed to accommodate the color range and spectrum of light-emitting RGB devices. Adobe RGB (1998) is a considerably larger space that grew out of wishful thinking for a future generation of video monitors.
ColorMatch RGB and Adobe RGB (1998) are in common use in output-centric workflows, but neither was designed with the idea of color-accurate output as the paramount concern. As a result, both spaces suffer from something of a mismatch with typical hard copy output, whether from a CMYK press or a photo-realistic inkjet printer. Both spaces clip (drop out) the saturated yellows and oranges achievable in sheet-fed printing and on photo printers: You’d have to resort to a very large space such as Adobe’s Wide Gamut RGB or Kodak’s ProPhoto RGB to encompass those. But ColorMatch RGB also clips cyan, as well as the blues and greens that lie adjacent to it, quite significantly. Adobe RGB doesn’t clip printable cyan, but it contains a fairly large number of colors that few if any output devices can reproduce, so it wastes a good number of those precious 256 data points in each channel.
BruceRGB, in contrast, was designed with output in mind from the start. It clips fully saturated yellows by about the same amount as Adobe RGB, and quite a bit less than ColorMatch RGB. It may clip cyan slighly with very high-quality sheet-fed printing, but not by more than a few percent — much less than ColorMatch RGB. Equally important, it wastes far fewer bits on unrealizable colors than Adobe RGB.
We can represent the relative size and shape of the various color spaces graphically, but only with one major caveat: Color gamuts take on complex three-dimensional shapes, and two-dimensional gamut plots are not a terrifically accurate way to judge actual gamuts. In short, then, don’t take the following 2D figures as literal representations of the gamuts involved. Use them as visualization aids instead. (I made the gamut plots using the ICC 2D Viewer, one of several handy, free tools you can download from icctools.com.)
In practice, the differences between the color spaces we’re discussing are more subtle than the plots above would suggest. Still, though subtle, the differences are real, and using the wrong color space for the job at hand can easily undermine your attempts to reproduce color accurately.
Draw Your Own Conclusions
Imagine the following spoken loudly and clearly: BruceRGB is not the be-all and end-all of RGB working spaces. Rather, it’s a useful, safe space for the purpose for which it was designed — generating hardcopy. If your output is destined for very-wide-gamut processes, you should probably choose a larger space to work in. However, I’ve yet to find an output device that benefits significantly from using Adobe RGB rather than BruceRGB. Both spaces are deficient in the same area — the yellows — and wide-gamut output devices such as film recorders, Durst Lambdas, and so on, tend to extend their gamuts primarily into the yellow region, not the greens and blues. I’ll address large-gamut work in my next column.
In short, BruceRGB is a good choice if you’re primarily concerned with ink-on-paper output, you work in 24-bit RGB, and you need a decent amount of editing flexibility. If you’re using one of the other spaces without encountering problems, by all means continue to do so: Don’t try to fix what isn’t broken.
Whatever space you use for editing, if your imagery is precious to you, archive the raw capture file, whether taken with a scanner or digital camera. Then convert a copy into the working space of your choice and use the copy as the work file. One day we may be able to do things with color that we can only dream of today, and the raw capture file represents the image in its most pristine state.
Having been studying the issues of digital color for a couple of years now, it’s great to have the working space issue clearly described, summarized, and put into related contexts.
This is a wonderfully clear article; Fraser does an excellent job explaining the trade-offs one makes when choosing a working space. However, I do wish that he analyzed the specific differences between the various working spaces he referred to in greater detail.
Thanks, Bruce, for giving such a lucid explanation of the often confusing subject of color space.
still leaves me undecided as to what color space would suit my work (photography) best. I now understand some of the pros/cons. Thanks