Colour in photography begins with illumination, surface reflectance, lens transmission, sensor spectral response, RGB colour filters, white balance, camera profiles, colour temperature, colour space and display calibration. Search for “colour photography,” “white balance photography,” “colour temperature,” “why colours look different in photos,” “camera colour science,” or “RAW colour,” and most explanations teach photographers to make whites look neutral and avoid unwanted colour casts. The deeper mechanism is that a camera does not directly record an object’s colour name. It records wavelength-dependent light after that light has been shaped by the illuminant, the object and the imaging system.
A white shirt under tungsten light sends a different spectrum toward the camera than the same shirt under blue sky. Two materials that look identical under one lamp can diverge under another. Different camera sensors can convert the same incoming spectrum into different RGB values because their filter responses differ. RAW software can render the same sensor data with different profiles. A calibrated monitor can show a colour differently from an uncalibrated phone. Colour is therefore a chain of measurement and interpretation, not a pigment label attached permanently to the object.
This is why white balance, colour temperature, RGB, RAW, camera profiles, colour management, calibration, metamerism, colour constancy and lighting belong to one mechanism. The photographer’s job is not merely to remove casts. It is to decide which colour relationship should be preserved, which illuminant should be treated as neutral, how much scene atmosphere should remain, and whether the output needs emotional plausibility, product accuracy or scientific repeatability.
Central proposition: a camera measures coloured light. “The colour of the object” is an inference produced from illumination, reflectance, sensor response and rendering.
1. Objects Do Not Emit Their Everyday Colour Labels
Most ordinary objects reflect part of the illumination falling on them. Change the illuminant spectrum and the reflected spectrum changes. The object may remain physically unchanged while the camera measurement shifts.
2. Illumination Is Half the Colour Problem
Daylight, tungsten, fluorescent and LED sources have different spectral distributions. Two sources can look similarly white to human vision while rendering coloured materials differently.
3. White Balance Chooses a Neutral Reference
White balance adjusts channel relationships so a selected neutral surface renders neutral under a chosen interpretation of the light. It does not physically remove the illuminant from the scene.
4. Auto White Balance Is an Inference
The camera analyses the scene and estimates which colour bias belongs to illumination rather than subject matter. A frame filled with genuinely warm objects can therefore confuse automation.
5. Kelvin Is Useful but Incomplete
Colour temperature describes the chromatic appearance of blackbody-like illumination along one major axis. Real fluorescent and LED sources can deviate in green-magenta directions that a single Kelvin number does not capture.
6. Tint Handles Another Axis
RAW processors often pair temperature with tint. Temperature broadly moves blue-yellow; tint compensates green-magenta deviations. Together they provide a more practical neutralisation control.
7. A Grey Card Gives the Workflow a Known Neutral
A calibrated neutral target photographed under the subject light provides a reference for white balance. The card is valuable because its expected neutrality is known rather than guessed from scene content.
8. A Colour Chart Provides More Than Neutrality
Known coloured patches allow software or operators to compare camera rendering against reference values and build or select profiles. This is important for products, reproduction and multi-camera matching.
9. RGB Is a Camera Encoding, Not the Spectrum Itself
Most cameras reduce complex incoming spectra to three broad colour-filter responses. Many different spectra can produce similar RGB triplets. Information is compressed.
10. Metamerism Is Colour Ambiguity Made Physical
Two spectral distributions can produce the same perceived or measured colour under one observer and condition while differing under another. Such matches are metamers.
This is why two fabrics can match in a shop and diverge outdoors, or why a camera and human eye can disagree about two nominally identical colours.
11. Cameras and Human Eyes Are Different Observers
Human cone sensitivities and camera colour-filter sensitivities are not identical. A spectrum that produces one perceptual relationship for a person can map differently through a camera sensor.
12. Camera Models Have Different Spectral Responses
Manufacturers design filters, microlenses and processing differently. Two cameras photographing the same target under the same light can therefore require different colour matrices or profiles to match.
13. RAW Does Not Have One Finished Colour
RAW files contain sensor measurements and metadata. Demosaicing, white balance, camera profiling, tone mapping and colour-space conversion are needed before ordinary colour appears on screen.
Different RAW converters can therefore begin with different renderings while using the same underlying capture.
14. JPEG Commits Earlier
A camera JPEG already contains white balance, colour style, tone curve and compression decisions. This can deliver excellent consistent colour quickly but leaves less freedom to reinterpret the capture later.
15. Picture Styles Are Creative Colour Decisions
Vivid, portrait, landscape, monochrome and manufacturer-specific looks alter saturation, hue relationships and tone. They are renderings, not corrections toward one universal truth.
16. Colour Constancy Is Why Humans Often See Stable Objects
Human vision adapts to illumination and tends to perceive familiar surfaces as relatively stable in colour across changing light. Cameras can mimic part of this through white balance but do not possess the same perceptual context.
17. Correcting Every Cast Can Destroy Atmosphere
A candlelit room is genuinely warm in appearance. Neutralising a white napkin perfectly can make the scene feel colder than it was experienced. Technical neutrality and perceptual fidelity are not always the same job.
18. Mixed Lighting Has No Single White Balance
A face lit by window daylight and tungsten lamp receives different spectra across different regions. One global white-balance setting cannot make both illuminants neutral simultaneously.
The repair is lighting control, local correction or acceptance of the mixture.
19. LED Lighting Can Create Colour Problems Beyond Kelvin
LED spectra can contain narrow peaks and gaps. Two lamps with similar correlated colour temperature can render skin, fabric and products differently because their spectral power distributions differ.
20. Fluorescent Light Can Produce Similar Difficulties
Discontinuous spectra can create colour relationships that simple temperature correction cannot fully repair. Colour rendering quality belongs to the source, not merely the camera.
21. Skin Tone Is Not One RGB Target
Human skin varies widely in reflectance, undertone and melanin. Lighting, makeup, exposure and profile all change the recorded values. Good portrait colour preserves believable relationships rather than forcing every face to one numerical recipe.
22. Product Colour Requires a Defined Viewing Condition
A catalogue cannot promise that one RGB value will make a fabric look identical under every display and room light. Colour-critical workflows define illuminants, profiles, displays and print conditions to reduce uncertainty.
23. Display Calibration Is Part of Photography
A well-captured file can look wrong on an uncalibrated monitor. White point, brightness, gamut and tone response influence what the editor sees.
The camera is only the first colour device in the chain.
24. Colour Spaces Define Numerical Meaning
An RGB triplet such as 200, 80, 60 does not specify a unique visible colour without knowing the colour space. sRGB, Adobe RGB and larger working spaces define different primaries and transformations.
25. Gamut Is a Boundary
A camera may capture colour relationships that a particular display or printer cannot reproduce. Conversion must compress, clip or remap out-of-gamut colours.
26. Soft Proofing Predicts Another Device
Editing software can simulate how an image may reproduce through a printer-paper profile. The preview is useful because output colour is a device-dependent transformation.
27. Print Adds Illumination Again
A monitor emits light. A print reflects room light. The same print can appear different under daylight, warm domestic lamps and gallery lighting.
The colour chain ends by returning to illumination.
28. Underwater Photography Loses Colour Spectrally
Water absorbs wavelengths unevenly, removing reds rapidly with distance. White balance can amplify remaining red-channel information but cannot recreate wavelengths that never reached the sensor with useful signal.
29. Sunset Colour Is Illumination, Atmosphere and Adaptation Together
Warm low-angle sunlight travels through more atmosphere and changes the spectral balance of illumination. Auto white balance may partially neutralise that warmth even when the warmth is the subject.
30. Stage Lighting Is Deliberately Non-Neutral
Concert lights may be saturated blue, magenta or red by design. Neutralising performers toward daylight colour can destroy the lighting designer’s intent.
Colour accuracy must be defined relative to the job.
31. White Balance Can Change Histogram Warnings
Rendered previews apply channel gains. Change white balance and one channel can move closer to clipping in the preview even though the RAW exposure is unchanged.
32. Channel Clipping Can Destroy Colour Before Luminance Looks Extreme
A saturated red flower can clip red-channel detail while overall brightness appears acceptable. RGB histograms reveal failures hidden by a combined luminance graph.
33. Infrared Contamination Shows That Sensors See Beyond Human Colour
Digital sensors can respond outside the visible range, so cameras use filters to suppress infrared and ultraviolet contamination. Remove or alter those filters and familiar materials can render in unexpected tones.
34. Colour Photography Is Always a Translation
The scene spectrum is filtered into sensor channels, transformed through profiles, mapped into a colour space and emitted or reflected by an output device. No stage simply hands the original spectrum unchanged to the viewer.
35. Scientific Colour Requires Controlled Conditions
If colour change is being measured over time, illumination, camera settings, target calibration and processing must remain controlled. Auto white balance can hide the very change being studied.
36. Documentary Colour Needs Context
Correcting a strong ambient cast can make a scene easier to read but can also alter how the environment actually appeared. Documentary workflows should distinguish clarification from reinterpretation.
37. Memory Colour Influences What Viewers Accept
People carry expectations about blue sky, green foliage and skin. A technically measured rendering can feel wrong if it conflicts with those expectations, while a slightly adjusted rendering can feel more natural.
38. Colour Contrast Changes Perceived Colour
The same