The Three Dials of Colour

Three sentences come up constantly: “the colour is too heavy,” “the colour looks odd,” “the colour looks muddy.”

They describe three entirely different problems and need three different fixes. If you cannot say which one you mean, all that is left is dragging sliders and hoping. So this chapter is practical in its aim — break colour into three dials you can turn separately, so that in front of a scene you can say which one you are reaching for.

• Hue — which colour it is. Red, orange, yellow, green, cyan, blue, violet, around a circle.
• Saturation — how pure that colour is. Drop it to zero and you have grey, whatever the hue was.
• Luminance — how bright it is. Pink is a high-luminance red, burgundy a low-luminance red, and the hue is the same in both.

Map those back onto the three complaints: “too heavy” is saturation, “looks odd” is a hue shift, “looks muddy” is luminances that were never separated.

Why exactly three

This is not a convention, and it is not because colour intrinsically has three aspects. It is because the instrument reading it has three channels.

Colour vision is carried by cone cells, and there are only three kinds, each most sensitive to long, medium or short wavelengths. Any beam of light entering the eye — a rainbow, a sodium lamp, the reflection off a leaf, however complicated its spectrum — is reduced to three numbers. Colour is three-dimensional because the reader is three-dimensional.

That has one strange and enormously useful consequence. Two beams with completely different spectra will look identical as long as they produce the same three numbers. This is called metamerism. Your screen mixes yellow out of red, green and blue dots; physically it has nothing in common with single-wavelength yellow light, and you cannot tell them apart. Colour photography, colour printing and colour displays are all built on that loophole. So is your camera: the filter array over the sensor is an attempt to imitate the three cone sensitivity curves, and the closer the imitation, the closer the colours to what you saw.

Metamerism also fails, and photographers meet the failure often. Two fabrics that match perfectly in daylight can visibly disagree under an LED panel, because a match is only ever a match for one illuminant and one observer. Change the spectrum of the light and the three numbers stop agreeing. When a client insists the jacket was not that colour, this is frequently what happened — not a white balance error, but two objects that were only ever the same colour under the light you first saw them in. A camera with a slightly different filter array counts as a different observer, which is why two bodies photographing the same scene can disagree about one particular colour while agreeing about everything else.

James Clerk Maxwell turned this into engineering in 1861, projecting three filtered exposures of a tartan ribbon — taken for him by the photographer Thomas Sutton — back through red, green and blue lenses to make the first colour photograph, demonstrating that three channels suffice. It arguably should not have worked: the emulsions of the day were blind to red, and later reconstructions concluded the red exposure recorded ultraviolet that the dye happened to reflect and the filter happened to pass.

Then a second layer complicates the picture in a way worth knowing about. Three cone types is only the front end; the retina immediately re-encodes those three signals into opponent pairs — red against green, blue against yellow, and light against dark. Both accounts of colour vision that spent the nineteenth century arguing with each other turn out to be right, one stage apart. This matters here because the next chapter’s complementary pairs are not an artistic convention; they are the axes of that second encoding.

One more thing, because it explains why the colour wheel closes at all. The spectrum is a line, running from violet at 380 nm to red at 700 nm, and the two ends do not meet. Yet the wheel you have seen is a circle, with magenta bridging red and violet. That magenta does not exist in the spectrum — no single wavelength is magenta. It is what the brain returns when the long-wave channel and the short-wave channel are both stimulated and the middle one is not. The wheel closes only because the visual system invented a colour to close it with.

Dial	What it is physiologically	How people describe it going wrong
Hue	The ratio between the three channels	“The colour looks odd”, “the whites aren’t white”
Saturation	The spread between the channels; no spread is grey	“Too heavy”, “too lurid”, “like a paint box”
Luminance	The weighted sum of the channels	“Muddy”, “grey”, “everything runs together”

While we are here: luminance is not exposure. Luminance is how bright a given colour is in itself — pure yellow is simply brighter than pure blue, and no exposure decision changes that. Open up two stops and blue is still the darkest colour in the frame.

The number of colours matters more than their intensity

Nine times out of ten a beginner’s colour problem is too many, not too few. A street with a red sign, a blue umbrella, a green tree, a yellow taxi and white road markings has five things shouting at once, so nothing is heard.

Be careful with one misdiagnosis here: colour failing is not the same as the photograph being badly made.

That market frame is here deliberately as the counter-example: nothing about it is technically wrong and the colour still does not work. The fault is not that saturation is too high. It is that no colour is more important than any other. Colour carries information by difference, and when every colour is present, difference is gone.

Three things to do about it, easiest first:

1. Move closer. Exclude the colours you do not need. Fastest and most effective, because it fixes composition at the same time.
2. Wait. For the wrongly coloured person to walk out, or the red car to drive off.
3. Desaturate or convert. Switch the variable off entirely — chapter ten is about when that trade is worth making.

Saturation is not free

Why the banding? Because each hue has a ceiling on how saturated it can get, and the ceilings differ. As saturation rises, regions already near their ceiling hit it first — they cannot go further, so neighbouring colours that used to differ subtly are crushed to one value.

It shows first in reds and oranges, not because the gamut is narrow there but because the red channel runs out of headroom first: a red or orange is already a near-full red channel with the other two low, and there is nowhere further to push. Sunsets, petals and skin are the three subjects most likely to break, and they happen to be three of the most photographed.

Among Fujifilm’s film simulations, Velvia is the saturated one. It is lovely on landscape and flushes faces on portraits — not because it is badly designed. Two things combine: skin already sits where the red channel has least headroom, and skin is one of the memory colours from later in this chapter, so a shift of a few percent that nobody would notice on stone or foliage is immediately visible on a face.

There is a side effect people rarely mention: adding saturation makes colours darker. In RGB terms, pushing a colour towards purity means pushing the other two channels down — to make red redder you lower green and blue, and those channels were carrying brightness. So a heavily saturated photograph usually goes dark and heavy at the same time, the photographer then raises brightness to compensate, the two adjustments fight, and the result is bright, lurid and flat.

The three sliders that look alike are doing three different jobs:

Slider	What it moves	When to use it
Saturation	Everything in proportion, including what is already saturated	When bringing things down; upward, be very conservative
Vibrance	Only the less saturated regions, protecting saturated colours and skin	When pale colours need to come up but faces must not move
HSL per-hue chroma	Only the one hue you name	When a single colour is doing the shouting

Order matters too: settle the number and the area of the colours before touching saturation. Doing it the other way round is tuning parameters on a structure that is wrong.

The colour in front of you is not the colour in the file

Almost everyone has had this experience: the colour was so good on site that you raised the camera, and at home the file looks ordinary. The camera did not fail. Three physiological facts did the work.

One: your eyes move, the photograph does not. Standing there, your gaze roams, and at each stop the pupil and the retinal gain readjust. You are effectively seeing a dozen separately optimised exposures and assembling an impression in which everything was legible. A photograph is one exposure on a flat surface and the viewer takes it in at once — that ease has to be paid for in dynamic range (chapter nine).

Two: the visual system subtracts the colour of the light. This is colour constancy: white paper stays white under a tungsten lamp because the brain knows the yellow belongs to the lamp, not the paper. A camera has no such mechanism and can only guess, which is what white balance is. So interiors that felt perfectly neutral come back alarmingly yellow — chapter thirteen is about that.

Three: three colours have stored expectations. Sky blue, foliage green and skin tone are called memory colours. Deviate slightly and viewers notice; they usually cannot name what is wrong and simply say the colour looks off. The inverse is just as true — brick, clothing and paint can drift several hundred kelvin and nobody notices, because there is no stored value to violate.

The practical conclusion is blunt: when the colour looks wonderful in front of you, ask whether the wonder comes from your being there. If it does, no file will recover it, and you should photograph something else. If the colour genuinely lives in one part of the frame, walk towards it until the area it occupies in the picture resembles the area it occupies in your head.

Colours carry their own brightness

This is the most consequential thing beginners overlook: hue comes with luminance attached.

Measured as sRGB relative luminance, the gaps between pure hues are startling:

Pure hue	Relative luminance	Roughly as bright as
Yellow	0.93	Almost white
Cyan	0.79	Light grey
Green	0.72	Mid-light grey
Orange	0.48	Mid grey
Magenta	0.28	Dark grey
Red	0.21	Dark grey
Blue	0.07	Almost black

Pure yellow is more than twelve times as luminous as pure blue.

Two direct consequences. First, “a yellow object on a blue background” is a tonal contrast as well as a hue contrast, which is why the pairing is so striking — the next chapter builds on it. Second, red and green are only about three times apart in luminance against yellow-and-blue’s twelvefold. Resist the over-reading here: pure red and pure green do separate in black and white — 0.21 and 0.72 convert to greys around 127 and 220, nearly a hundred levels apart. What actually collapses together is a red coat against green leaves, because neither is a pure hue; once each is mixed with grey and with the colour of the ambient light, their luminances land very close. The “subject disappears into the background” symptom in chapter ten is about the second case, not the first.

Six common symptoms and their causes

Always in the form: the picture shows X → the usual cause is Y → change Z.

1. Colour reads as noisy, like a paint box → too much frame area held by saturated colours at similar brightness, with no hierarchy → reduce the number of colours first (step closer, change angle), not the saturation.
2. Desaturating left the picture lifeless → you lowered global saturation, so already-pale regions lost what little they had → drop only the offending hue with per-hue controls, or lower that hue’s luminance instead: a darker red is quieter than a paler one.
3. Colour looks muddy, several patches run together → mid-saturation colours at similar luminance, so the boundaries are unreadable → this is a lighting problem, not a colour problem: find a side-lit angle that creates a brightness difference, or convert to black and white (chapter ten).
4. Faces look red, almost sunburnt → a saturated rendering has run the red channel out of headroom on skin, which as a memory colour has the least tolerance of any subject → switch to a less saturated film simulation, or shift white balance a step or two towards cyan (chapter thirteen).
5. The same jacket photographs orange indoors and red outdoors → this is a hue shift, not a saturation problem, and the cause is the light source → white balance, chapter thirteen.
6. Beautiful on screen, dark and dull in print → print gamut is narrower than screen gamut, and saturated reds, blues and violets fall outside it first → leave those colours some headroom at capture rather than pushing saturation to the limit.

Who invented these three dials

At the end of the nineteenth century, colours were described by name: rose, dove grey, azure. This is useless for teaching — two people using one name mean two different colours.

Albert Munsell (1858–1918) was a Boston painter and art teacher, and it irritated him enough to fix it. In 1905 he proposed describing colour by three mutually independent coordinates instead of names: hue, value and chroma — precisely the three dials in this chapter.

What kept his system alive was how he calibrated it. He did not compute the spacing; he adjusted it by eye until the difference between any two neighbouring steps looked the same size as any other. It was the first perceptually uniform colour system.

And what he ended up with is not a sphere but a lopsided tree, because hues differ both in the maximum chroma they can reach and in the lightness at which they reach it. Yellow attains its most intense chroma high up, near white; blue only does so far down, near black. The tree is squat on the yellow side and tall and narrow on the blue-violet side.

Why it leans is something you already saw in the luminance table above. That is not a coincidence but the same fact stated twice: hue carries luminance, so colour space cannot be symmetrical.

Every HSL and HSB slider in every imaging application is a simplification of Munsell’s three axes, and the United States Department of Agriculture still describes soil with Munsell colour charts. A system built so that children could be taught to paint became the coordinate frame of an industry — and the reason you can say “lower the saturation, leave the hue” and be understood at all.

A table for the field

What you want to change	The dial it actually lives on	On location	Afterwards
“Too busy”	The number of colours — no dial at all	Move closer, change angle, wait	Crop; if hopeless, convert to black and white
“Too lurid”	Saturation	Choose a less saturated film simulation	Bring down only the loudest hue
“The whites aren’t white”	Hue	White balance (chapter thirteen)	Free to change in RAW
“Subject sticks to the background”	Luminance	Change the direction of the light to build a brightness gap	Dodge and burn locally; do not reach for saturation
“No colour leads”	Proportion of area	Let one colour hold a small patch (next chapter)	Crop to change the ratio
“It fell apart in black and white”	Luminance distance between hues	Shoot with the monochrome preview on	Channel mixing or filters (chapter ten)