Color spaces
How do colors always look the same?
When you take a photo, the image is generated by detecting photons striking a silicon CCD, using red, green and blue filters to capture color. When you view on your iMac, the photo is displayed by shining a backlight through IPS LCD RGB subpixels. After tweaking the RGB values in a photo editor, you can print it using dots of cyan, magenta, yellow and black (CMYK) inks, which each absorb different wavelengths.
These are different processes, and each piece of hardware will differ from others, even if based on the same tech. Yet the photo looks the same throughout. There’s a lot of color science making that happen.
Squaring the circle
For thousands of years, philosophers worked on models of color that assumed it had the perfection of geometry. Spoiler: it didn’t. After key discoveries in the 19th century about human vision by Thomas Young and Hermann von Helmholtz, in the early 1900s Prof. Albert H. Munsell came up with the first proper model of the colors we actually see. He painted patches of each hue (color of the rainbow) with steps in value (brightness) and chroma (saturation), then, by asking people which ones looked equal, arranged them according to ‘perceptual uniformity’. The result was a lumpy blob. CIE color gamut diagrams (see below) are
a 2D version of this. A color space is the blob, or gamut, of colors that a system can reproduce, or a device-independent space, such as sRGB.
Formulae specified by the International Color Consortium (ICC) convert between color spaces. Numbers that represent color are only meaningful in terms of a color model, so ideally an ICC profile should be stored with each image – otherwise whatever device reads it has to guess. For web images, though, we don’t want any unnecessary bytes. So web designers save images in sRGB but without a profile, and web browsers assume they’re in that space.
Room with a view
Saving an image in a color space imposes limits on its fidelity, and the color space in which your editing software works affects the accuracy of your edits; pushing colors outside the gamut will clip information. So you should aim to work in a suitably large color space. The most popular working space for print designers is Adobe RGB (1998), which extends sRGB at the blue-green end, while video editors have used Rec709 but are moving to wider spaces like DCI P3 and Rec2020.
Premium smartphones, tablets and displays cover all of the sRGB gamut, while cheaper displays usually manage at least 70%. Within a space, colors may be mapped more or less accurately, expressed as a Delta E value: below 1 is excellent, below 2 is decent. As color technology improves, wider gamuts are getting more common. Apple uses P3 in recent iMacs, iMac Pros and iPad Pros.
A wider-gamut screen means you can edit photos and videos more confidently and see the benefit of HDR (high dynamic range) content, but it can also bring problems. Software that assumes images are sRGB will make colors too vivid in P3. And although these displays cover more of the Adobe RGB gamut than before, quite a lot of it falls outside P3.
In editing software, your working RGB space should be a generic space such as sRGB, not a display or printer profile. Many designers will prefer Adobe RGB. Save your work in the same space, with a profile; copies can be saved as sRGB if necessary. The display profile set in System Preferences > Displays > Color should take care of on-screen color.
Respecting profiles
iOS also manages color, and app developers are encouraged to use profiles and cater properly for screens that use the P3 color space, so color should ‘just work’. On both macOS and iOS, Safari now respects display profiles and will make images that are assumed to be sRGB appear correctly on wider-gamut displays.
If you work in CMYK for print, a CMYK ‘press characterization,’ such as US Web Coated (SWOP), can serve as both working space and press profile. With photos, though, you’ll normally work in RGB and convert to a CMYK profile on output.
Larger color spaces need more bits per pixel to avoid posterization (banding). The huge ProPhoto space requires 16 bits. While apps can work in 16- or 32-bits, most displays are 8-bit. 10-bit color (also known as 30-bit, since there are three channels) is supported by OS X 10.11 and later, but only Mac Pro graphics cards support 10-bit displays.