Chromaticity Diagrams Without the Intimidation

A chromaticity point is not just a dot. It can be checked against device primaries, converted into u'v', pulled toward or away from a reference white, compared with a blackbody locus, and summarized as dominant wavelength and excitation purity. This lab puts those operations in one place so the rest of the article has something concrete to attach to.

A chromaticity diagram is a two-dimensional plot of color without brightness. The horseshoe is the boundary of human-visible chromaticities - everything inside it is a color a standard observer can perceive; everything outside is either physically unrealizable or out of the visible band. The horizontal axis is x and the vertical axis is y, each derived from CIE XYZ tristimulus values.

Each point on the diagram is one chromaticity - one combination of hue and saturation. Brightness has been factored out by the projection. Two physical stimuli of the same chromaticity differ only in luminance; a deep red sun and a deep red LED can sit at the same point on this plot at very different intensities.

The diagram is what you get when you divide XYZ tristimulus values by their sum. That single normalization throws away brightness and keeps only the chromatic information.

Because the three coordinates always sum to one, the third is redundant. Drop z and you have a two-dimensional plot of every visible chromaticity. The remaining brightness information lives in Y, which is why the full triple "xyY" reproduces the entire stimulus.

The horseshoe-shaped boundary is the spectral locus. Each point on it is the chromaticity of a monochromatic light - a single wavelength, the most saturated color of that hue a human can see. The locus runs from short wavelengths (380 nm) at the bottom left corner, around through green at the top, then down through yellow, orange, and red at the bottom right corner (~700 nm).

The straight line closing the bottom of the horseshoe is the line of purples. It connects the deepest visible violet at one end to the deepest visible red at the other. Colors along this line correspond to no single wavelength - they are mixtures of short and long wavelengths that the eye synthesizes into a unified purple-magenta percept.

Magenta has no place in a rainbow because a rainbow is a continuous spectrum, and magenta requires combining two non-adjacent wavelengths. Yet magenta is a real, stable, named color experience. This is one of the cleanest examples of how color lives in the brain, not in the light.

Somewhere near the center of the horseshoe sits the white point - the chromaticity declared neutral for a particular system. Lines through the white point define hue circles. Distance from the white point measures saturation. Most published diagrams pin a D65 white point at (0.3127, 0.329); print workflows use D50 at (0.3457, 0.3585); the more amber Illuminant A sits at (0.4476, 0.4074).

Move the white point and every color on the diagram is reinterpreted. A point near the spectral red, viewed against a D65 white, is "deep saturated red." Viewed against a candle-warm white, the same point is closer to neutral - it sits along the hue axis the observer would consider least chromatic. This is why color appearance always specifies a reference white.

The horseshoe is the limit of human vision. A device - screen, camera, printer - cannot match all of it. Each device has its own gamut, the set of chromaticities it can actually produce. For an additive three-primary display, the gamut is a triangle with vertices at the device's three primary chromaticities. Inside the triangle: producible. Outside: out of gamut.

Heat any object until it glows. At low temperatures it emits a dim, deep red; raise the temperature and it brightens through orange, yellow, white, and eventually blue-white. The trajectory in chromaticity space is the Planckian locus - the chromaticities of an idealized blackbody at each temperature.

For light sources whose chromaticity is near (but not exactly on) the Planckian locus, the correlated color temperature (CCT) is the Kelvin value of the nearest blackbody. CCT collapses the chromaticity to a single intuitive number: 1900 K = candle warm, 3000 K = incandescent, 5000 K = print viewing, 6500 K = daylight, 10000 K = cool sky. Every "warm white" / "cool white" lamp label is a rounded CCT.

Take any chromaticity inside the horseshoe and draw a straight line from the white point through it, extending until it meets the boundary. Where the line crosses the spectral locus, that wavelength is the chromaticity's dominant wavelength - the spectral hue it most resembles. Where the same line extended in the opposite direction crosses the boundary, the wavelength (or purple line position) is the complementary.

Dominant wavelength is the closest the CIE system comes to "naming the hue" of a color. It also gives a clean definition of excitation purity - how far along the line from white to spectral the color sits, on a 0-1 scale. Excitation purity is the colorimetric analog of saturation.

The xy diagram is geometrically convenient but perceptually skewed. The green portion is stretched out, the blue portion crushed into a corner. Equal numerical distances in xy correspond to wildly unequal perceptual differences. The CIE addressed this in 1976 with the u′v′ uniform chromaticity scale - a projective transform that flattens the distortion.

The u′v′ diagram looks broadly similar to xy but with the regions stretched more uniformly. Distances between chromaticities in u′v′ correlate much better with perceived color differences. CIE recommends u′v′ for color difference work on self-luminous displays and for color rendering metrics like Δuv (the distance from the Planckian locus).

A few mistakes show up over and over in articles, marketing copy, and casual explanations of chromaticity diagrams. Learn to spot them.

Six questions on the chromaticity diagram, gamuts, and white points. Wrong answers come with brief explanations.