Technikal Support

Working Spaces

Introduction

This article on working spaces is written for photographers that are preparing images for printing. A lot of the examples are specific to Adobe Photoshop, but most of the principles are broadly applicable. Each section is devoted to one of the four device-dependent color models: grayscale, RGB, CMYK, and spot color. We'll start with gray working spaces, the simplest, move on to the increasingly complex RGB and CMYK spaces, and end by touching on spot color working spaces.

This article assumes a basic knowledge of color management and ICC profiles, but we'll touch on key concepts along the way. If you're looking for a thorough introduction to the basics of color management and working spaces, I recommend this paper written for Adobe by Andrew Rodney. Although his focus is on RGB working spaces (and I disagree with his recommendation of ProPhoto RGB), his explanation of the fundamentals is as good as it gets.

Part 1: Gray

They are less glamorous than their RGB counterparts, but choosing the appropriate grayscale working space shouldn't be an afterthought. Surprisingly, even if you work exclusively in RGB, ignoring gray spaces may have adverse effects on your images. Plus, they are an easy way for us to start thinking about working spaces more generally.

Simply put, a gray working space is two points and a curve. In profile jargon this is a white point, a black point, and a tone reproduction/response curve or TRC. The white point is generally D50 or D65. The black point is almost always 0. The TRC describes the relationship between tones. It is often a power function, i.e. gamma 1.8, gamma 2.2, etc., but it can also be determined by discrete points or another type of equation. One example of the latter is L*, the lightness component of CIE LAB space which makes the relationship between tones perceptually uniform. Gamma 2.2 is actually a decent approximation of perceptual uniformity (you can see Bruce Lindbloom's analysis here) and mathematically it's less complex.

The default gray working space in Adobe Photoshop is "Dot Gain 20%." There are two problems with this. First, unless you are making halftones, you shouldn't use a dot gain based TRC (if you are, consult with your printer before assuming that a TVI of 20% for a 50% tone approximates the press condition). Second, even if you always work in RGB, Ps uses the gray working space to render layer masks for RGB files. So if you do any complex operations like making HSL masks, it's important that your RGB and gray working spaces have matching TRCs. At some point adobe caught on to this issue and now Photoshop does a TRC conversion on the fly when you create a layer mask, but a bug remains. If you change your gray working space at some point, the masks in your file will now all be rendered with the wrong TRC, meaning those adjustments will now be either weaker or stronger than they were before.

So, which gray profile should you choose? In most cases you can safely ignore white points and simply choose a profile with a TRC that corresponds to your RGB working space. If you primarily prepare images for the web and your RGB working space is "sRGB," set your gray working space to Adobe's "sGray." If you've changed your RGB working space to "Adobe RGB" then use "Gray Gamma 2.2." Not sure what your RGB working space should be? Don't worry, we'll take a closer look at that in Part 2. If, however, you fall into the minority of users that use the absolute rendering intent or have some other reason to match profile white points, the following table lists common gray profiles along with their characteristics.

Source Profile Name WP TRC
Adobe Gray Gamma 1.8 D65 γ 1.8
Adobe Gray Gamma 2.2 D65 γ 2.2
Adobe sGray D65 sRGB
Apple Generic Gray Profile D65 γ 1.8
Apple Generic Gray Gamma 2.2 Profile D65 γ 2.2
Elle Stone Gray-elle-V2-g10 D50 γ 1.0
Elle Stone Gray-elle-V2-g18 D50 γ 1.8
Elle Stone Gray-elle-V2-g22 D50 γ 2.2
Elle Stone Gray-elle-V2-labl D50 L*
Elle Stone Gray-elle-V2-srgbtrc D50 sRGB
Roy Harrington QTR - Gray Lab D50 L*

We'd like to call special attention to Elle Stone's well-behaved profiles. She's gone to great lengths to track down some mathematical errors present in many popular profiles and fix them. They are publicly available via her website. And unlike other sources, she offers V2 and V4 flavors of every profile. We suggest most users stick with V2 profiles due to broader support, particularly when preparing images for the web.

Part 2: RGB

Last time we looked at three fundamental characteristics of a working space profile: white point, black point, and TRC. RGB working spaces add three more points to the mix. I glossed over this a bit in Part 1, but the white and black points are each described by coordinates in a color space, as are the RGB primaries. The typical space used is CIE 1931 XYZ. So if we were going to plot our "Gray Gamma 2.2" working space the white point (D65) would be located at 0.95, 1.00, 1.09 and the black point would be at 0, 0, 0. The TRC (gamma 2.2) distributes the values between the points so that they look more or less perceptually uniform. A profile like "Adobe RGB" which shares the same white point, black point, and TRC as "Gray Gamma 2.2" is identical when R=G=B. In other words, they share a neutral axis. That makes sense and also shows why it's wise to choose a gray working space that compliments our choice for RGB.

Now things get a bit more complicated. The addition of RGB primaries expands our one-dimensional gray space into a three dimensions. The distance of the primary coordinates from the neutral axis determines the gamut shape. Gamut is, as you probably know, related to saturation, and some people implement working spaces in such a way as to take advantage of this. See Joseph Holmes Chroma Variant Sets for an example of this (this is not an endorsement). But most photographers do not switch between dozens of working spaces as a means to edit their images. Instead, they use a single space that hopefully does a reasonably good job in most situations. This space needs to encompass a range of typical input and output gamuts—think scans, digital captures, and prints. Kodak developed the ProPhoto working space, which has a huge gamut, to avoid clipping input and output profiles. The problem with ProPhoto (aside from using gamma 1.8 for its TRC), is that the gamut is unnecessarily large thereby necessitating 16-bit files. A portion of the the gamut is used for colors not visible to humans and some of the visible portion simply won't be encountered in normal practice, especially if the intention is to make a print. A better working space is one that closely adheres to "The Gamut of Real Surface Colors" In some recent ICC profiles this gamut, referred to as the Perceptual Medium Reference Gamut (it has evolved a bit since Pointer's pioneering work), is used to calculate perceptual rendering between spaces. This is a pretty big development if it is implemented well.

Bruce Lindbloom has conveniently organized the colorimetric data for many popular profiles along with his calculations of gamut volume and efficiency. You can find that data here along with a ton of other handy information. His BetaRGB profile is well thought out and covers Pointer's Gamut very well (more on that in a bit).

Elle Stone has done some great detective work on various profile distributions and rounding errors. You can read about that here. And you can find her "well behaved profiles" here. Here are some working spaces that I think are particularly well suited to normal workflows.

Source Profile Name WP TRC
Bruce Lindbloom Beta RGB D50 γ 2.2
ECI ECI RGB v2 D50 L*
Don Hutcheson Don RGB 4 D50 γ 2.2
Elle Stone Rec2020-elle-V2-g10 D65 γ 1.0
Elle Stone Rec2020-elle-V2-g18 D65 γ 1.8
Elle Stone Rec2020-elle-V2-g22 D65 γ 2.2
Elle Stone Rec2020-elle-V2-labl D65 L*
Elle Stone Rec2020-elle-V2-srgbtrc D65 sRGB
Joseph Holmes EktaSpace PS5 D50 γ 2.2

Part 3: CMYK

There is a consensus that CMYK color models shouldn't be used for working spaces. I don't meet too many people that need to be convinced of this, but if you're old-at-heart you can read this article by Don Hutcheson from way back in Y2K.