How we see colour. Context changes appearance.

Rods are especially sensitive in dim conditions; cones underpin ordinary daylight colour vision. Between these conditions lies a mesopic range in which both systems contribute. There is no simple “rods at night, cones by day” switch. Evaluating video on a display normally assumes controlled viewing conditions, rather than complete darkness or arbitrary bright light around the display.

The three cone classes have overlapping spectral sensitivity curves. S cones are biased toward shorter wavelengths, M toward medium wavelengths and L toward longer wavelengths, but each class responds across a broad range. Renaming S/M/L literally as B/G/R is therefore misleading.

For a large class of colour stimuli, visual matching can be represented using three independent quantities. This makes tristimulus colorimetry, RGB systems and CIE XYZ possible. Importantly, mathematical tristimulus coordinates are a perceptual matching model, not direct readings from the three cone classes.

Different spectra can match in tristimulus values for a selected observer. This lets a display reproduce a colour without copying the scene’s spectrum. For reflecting samples, changing the illumination changes the outgoing spectra themselves, so a match may disappear. Observer differences also matter for emissive displays. Metamerism does not mean that any two identical RGB files are a pair of different physical spectra.

Neural processing builds opponent and luminance relationships: roughly red–green, blue–yellow and light–dark components. Perceptual hue, brightness and colourfulness therefore cannot be read directly from three cone responses. This is one reason perceptual colour models are more complex than plain RGB.

Vision partially adapts to illumination; a familiar white sheet can retain a white appearance in different conditions. Colour constancy is not absolute, and the outcome depends on the surround and the observer’s task. A chromatic-adaptation transform, for example between reference whites, models a limited part of this behaviour. It does not “reprogram the brain” or repair every lighting error.

The same image patch can appear lighter, darker, warmer or cooler depending on neighbouring regions and overall adaptation. A reference monitor alone therefore does not solve the whole problem; the viewing environment is part of the image-evaluation system.

Real people differ slightly in colour sensitivity. To make measurements reproducible, the CIE standardised colour-matching functions. The CIE 1931 2° Standard Colorimetric Observer is used for smaller visual fields, while the supplementary CIE 1964 10° observer represents larger fields. This is not a 'perfect human'; it is an agreed measurement model.

Scopes provide objective information about the signal, while vision judges the perceptual result. But eyes adapt, fatigue and depend on the surround. Serious grading therefore combines a calibrated display, controlled viewing environment, scopes, reference images and regular breaks rather than relying on any one of them alone.