Colour is not just in the pixel: how our eyes construct it

An unchanged square in a file looks cooler against a different background. A white sheet still appears fairly white under warm lighting, although its photograph turns orange. These differences are not merely settings errors: colour is a perception built from light, comparisons and context. Understanding that process makes photographic choices and editing more precise.

Kolorshift editorial6 chapters

Remember this

Cones provide signals rather than ready-made red, green and blue labels. The visual system compares and interprets them. Surroundings can change the appearance of an unchanged patch, while adaptation only partly compensates for illumination changes. A photograph should also be judged in its presentation conditions.

A colour is not simply a wavelength

Light reaching the eye can contain many wavelengths. For a surface, the mixture depends in part on illumination and on what the material absorbs or reflects. Perceived colour is not a label permanently attached to an object. Paint, a photographic print and a display also produce their outgoing light in different ways.

Monochromatic light can be described by one wavelength, but many photographic colours correspond to mixtures. Looking for the single wavelength of pink does not explain its appearance. The visual system receives a pattern of stimulation and interprets it in relation to the rest of the scene rather than simply reading a colour name from one incoming ray.

1. Light source · 2. A surface modifies the returning mixture · 3. Light stimulus reaching the observerA colour is not simply a wavelength — Relationship diagram, not a geometrical ray trace or anatomical drawing.Enlarge diagram

The first practical consequence is that a source change can alter the photograph without altering the object. A pleasing outdoor colour may separate differently under an interior lamp. Before moving several sliders, examine the actual illumination and the image’s overall tonal relationships instead of assigning all meaning to an isolated pixel.

[1] National Eye Institute [3] CIE e-ILV

Three cone families, not three pots of paint

In typical trichromatic vision, three cone families are called S, M and L, reflecting relative sensitivity to short, medium and long wavelengths. Their sensitivity ranges overlap. Calling them simply blue, green and red cones is convenient but can misleadingly suggest narrow detectors, each reserved for one pure colour.

The signals do not directly become the three channels of a photograph. Comparisons and opponent interactions begin in retinal circuits, with further transformations along visual pathways. Circuits described as red–green or blue–yellow should not be treated as a perfectly simple, universal wiring diagram for every named colour experience. Current research reveals substantial complexity.

1. S family · 2. M family · 3. L family: broad, overlapping sensitivities rather than a simple match to three paintsThree cone families, not three pots of paint — Unmeasured qualitative curves, arbitrarily normalised; axes have no spectral scale.Enlarge diagram

The drawing shows broad overlapping sensitivities without measured values or a particular eye’s anatomy. One stimulus can engage several cone families. An RGB editing control is therefore not a switch independently commanding three cone classes: it changes an image representation that will be reproduced by a particular device in particular viewing conditions.

[1] National Eye Institute [2] Kim, Packer & Dacey

The same square can change without changing

The two central squares in this example have identical encoded values: RGB 160, 160, 160. Only their surroundings change. Nevertheless, they may look different. Simultaneous contrast reminds us that colour is not judged in isolation: a neighbouring area, a contour and the spatial organisation of the scene influence appearance.

Observe the patches, then use the experiment’s button to make the backgrounds uniform. The effect may be subtle or different depending on your display, surroundings and perception. It is neither a vision test nor display calibration. What is controlled is the identity of the patch values, not a perceptual response guaranteed for every reader.

The same square. Two contexts.

Blue background

Orange background

Both central squares use exactly the same sRGB colour: #A0A0A0 (160, 160, 160).

The impression can vary with the display, surroundings and observer. This illustration is not a vision test or calibration.

1. Central patch RGB 160, 160, 160 · 2. Strictly identical central patch, different surround · 3. Encoded identity does not guarantee identical appearanceThe same square can change without changing — Patches share encoded values. The effect depends on display, surroundings and observer; this is neither a vision test nor calibration.Enlarge diagram

This distinction matters in editing. Skin can appear less warm after a background change even if its values remain untouched. Before correcting a face or sky locally, compare the complete photograph with a more isolated detail view. Otherwise you may alter the subject to compensate for its surroundings, then need to undo that correction elsewhere.

[3] CIE e-ILV [6] Kanematsu & Koida

Adaptation seeks stability, not a perfect white

Colour constancy is the tendency for object appearance to remain somewhat stable despite illumination changes. Chromatic adaptation contributes alongside other scene cues. Compensation is approximate: the visual system does not magically remove all illumination colour or recover perfect physical reflectance from any possible scene. Stability is an achievement with limits.

A familiar sheet beneath a warm lamp can look more neutral to an observer than in an uncorrected photograph. The still image does not contain every condition of the original observation. It will be viewed under different light and among different surfaces. Photographers decide whether to preserve a reference neutrality, the atmosphere or an expressive compromise.

1. Warm or cool lighting illustrated · 2. A surface shown in two contexts · 3. Adaptation may partially compensate for the differencesAdaptation seeks stability, not a perfect white — Illustrated contexts, not a colorimetric prediction of real lighting or perfect compensation.Enlarge diagram

Compare several plausibly neutral areas rather than assuming a single white patch reveals the complete truth. To separate perceptual adaptation from camera settings, read about illumination and white balance. A technical reference and an aesthetic intention can coexist when we know what each describes; one does not automatically invalidate the other.

[4] CIE e-ILV [5] Brainard, Brunt & Speigle

Hue, lightness and saturation: name the change

Hue locates a colour among families such as red, yellow, green and blue and their neighbours. Perceived brightness concerns how much light an area appears to emit or reflect. Lightness, in colour-appearance terminology, relates that impression to a similarly illuminated white reference. These concepts are not interchangeable with the value of one image channel.

Saturation concerns the colourfulness of an area relative to its perceived brightness; chroma uses a different reference. Software sliders named saturation, vibrance or brightness use different models. Their numbers are not universal measurements of sensation. Different hues with equal HSL lightness L do not necessarily have equal perceptual lightness. A software adjustment also need not produce a physically identical change in every photograph.

1. Hue variation · 2. Dark-to-light variation · 3. Colour-intensity variation; samples constructed in a software modelHue, lightness and saturation: name the change — Demonstration scales are not perceptually uniform and are not CIE measurements.Enlarge diagram

Before intensifying a colour, describe the problem: too orange, too light, too dominant compared with the rest? This vocabulary makes tests easier to interpret. Change one dimension at a time and examine transitions as well as patches. A vivid garment can retain detail; increasing saturation should not replace observing its folds and highlights.

[7] CIE e-ILV [8] CIE e-ILV [9] CIE e-ILV [12] CIE e-ILV [13] W3C

An image should remain readable beyond colour alone

Observers do not all distinguish colours in the same way. Colour vision deficiencies can affect particular distinctions without necessarily turning the world black and white. A screen simulation does not reproduce every person’s experience. It cannot diagnose a condition or promise a universal reconstruction of vision for everyone using the same settings.

For captions, controls or diagrams accompanying a photograph, combine colour with text, shape or a clear position. W3C guidance advises against making information depend on colour alone. This also helps with limited displays, different printed reproductions and real environments in which neighbouring hues become difficult to tell apart.

1. Colour alone distinguishes two states · 2. Colour and distinct symbols provide redundant informationAn image should remain readable beyond colour alone — Example of redundant cues; no colour vision deficiency is simulated.Enlarge diagram

In the photograph itself, also consider tonal separation, boundaries and the direction of attention. An attractive palette does not replace readable composition. Display assistance may help particular tasks without treating a vision condition. Understanding perceptual differences encourages several useful cues, rather than imposing a single colour code and assuming it communicates identically to every viewer.

[10] National Eye Institute [11] W3C WAI