Quick Read. A digital colour photograph is not captured as a ready-made grid of complete red-green-blue values. In common camera designs, neighbouring photosites measure light through different colour filters. The camera or RAW processor reconstructs full-colour pixels from that incomplete sampling, then applies white balance, colour transforms, tone curves and output profiles. The colour you see is therefore the result of measurement, inference and rendering.
One-sentence answer: Photographic colour works by measuring parts of the incoming spectrum, estimating missing colour information and translating those measurements into a colour system a human display or print can reproduce.
The Red Apple Is Not Sending “Red” Into the Camera
Place a red apple in daylight, then under a warm household lamp, then beneath greenish fluorescent light. The apple has not changed identity, but the spectrum reaching the camera changes because illumination and surface reflectance interact. A camera does not receive a semantic label saying RED APPLE. It receives wavelength-dependent light.
This is why colour photography is inseparable from illumination. The same surface can generate different sensor responses under different light sources. Human vision performs powerful adaptation, so we often experience a white sheet of paper as “white” across many environments. Cameras need their own strategy for deciding how neutral objects should be rendered.
Most Sensors Do Not Measure Full RGB at Every Photosite
A common digital sensor places a colour-filter array over its photosites. The Bayer pattern is the best-known example: some sites receive red-biased light, some green-biased, and some blue-biased. There are typically more green samples than red or blue because luminance detail and human visual sensitivity are strongly tied to the middle of the visible spectrum.
Each filtered photosite therefore measures only part of the spectral information arriving at that location. To produce a normal image in which every output pixel has red, green and blue components, the camera estimates missing values from neighbouring measurements. This reconstruction process is called demosaicing.
Demosaicing Is an Educated Reconstruction
Suppose one photosite measures a green-filtered signal. The final RGB pixel still needs red and blue values. A demosaicing algorithm looks at surrounding samples and estimates them while trying to preserve edges and avoid false patterns. Sophisticated methods can do this remarkably well, but difficult repeating textures can expose the limits through moiré or false colour.
This is an important conceptual shift. A digital colour photograph is not simply a mosaic of tiny perfect colour readings. It is a reconstructed image based on sampled measurements. The output can be highly accurate and useful while still being computationally assembled.
White Balance: What Counts as Neutral?
A tungsten-lit room may send relatively more long-wavelength energy toward the camera than open shade. If the camera rendered sensor values without compensation, neutral surfaces might appear strongly orange or blue. White balance adjusts channel relationships so that a chosen or inferred neutral surface is rendered neutral—or so the overall colour resembles the intended viewing experience.
Auto white balance attempts to infer the illuminant from the scene. Sometimes it succeeds beautifully. Sometimes it sees a predominantly warm sunset and “corrects” away the warmth you wanted. The camera can estimate statistics. It cannot know your emotional contract with the evening.
Colour Temperature Is Useful, but Not the Whole Story
Photographers often describe light using correlated colour temperature in kelvin. Lower values are commonly associated with warmer-looking light and higher values with cooler-looking daylight or shade. The apparent inversion makes sense once we remember that colour temperature comes from the spectrum of an ideal thermal radiator, not from the everyday language of warm and cool colours.
Real light sources may not lie neatly on the blackbody locus. LEDs, fluorescent lamps and mixed illumination can have spectral characteristics that require an additional green-magenta tint correction. Two sources with similar correlated colour temperature can still render colours differently because their spectra differ.
The Camera Has Its Own Colour Vision
Sensor spectral sensitivities do not perfectly match human cone responses. Manufacturers therefore characterise cameras and apply colour matrices or profiles that map sensor responses toward standard colour representations. Different cameras can produce subtly different colour from the same scene even before stylistic picture profiles are applied.
This is one reason “camera colour science” is not pure marketing language. There are real decisions about spectral response, transforms, tone, saturation and preferred rendering. But claims that one brand possesses universally “true” colour should be treated carefully. Accuracy depends on illumination, calibration, workflow, display and the definition of truth being used.
RAW Is Upstream, Not Colourless Truth
A RAW file generally contains minimally processed sensor measurements plus metadata. It gives the photographer more freedom over white balance, demosaicing, noise reduction, tone and colour rendering. But RAW data still came through a lens, filters and a sensor with particular spectral sensitivities. It is not a philosophical escape hatch from mediation.
What RAW gives you is more latitude and a position earlier in the pipeline. That is enormously useful. It is not the same as saying the file contains the scene exactly as a human saw it.
sRGB, Adobe RGB and Other Colour Spaces
Once colour has been reconstructed, the image needs a defined coordinate system. A colour space describes how numeric values relate to colours. sRGB is widely used because it is supported across the web and consumer devices. Wider-gamut spaces can represent a larger range of saturated colours, which can be useful in editing and printing workflows.
Numbers without a profile are ambiguous. RGB 200, 50, 50 does not identify one universal physical colour unless the colour space and transfer behaviour are known. Colour management exists so numbers can travel between devices without silently changing meaning.
Your Display Is Another Camera-Like Translation in Reverse
The camera translated light into data. The display translates data back into emitted light. Its primaries, brightness, white point, calibration and viewing environment influence what you see. A perfectly prepared file on a poorly adjusted screen can look wrong. Two phones can show the same image differently.
Printing adds another transformation. Ink and paper reflect ambient light rather than emitting it. Their achievable gamut differs from a bright display. A deep luminous screen blue may not reproduce with the same intensity on matte paper. Good colour workflow is therefore not about freezing colour unchanged. It is about managing translations deliberately.
Why Skin Colour Is So Demanding
Humans are extremely sensitive to faces. Small colour shifts in skin can feel immediately wrong even when other objects remain believable. Mixed lighting, reflected wall colour, makeup, underexposure, saturation and local tone mapping can all alter skin rendering.
Good portrait colour therefore requires more than moving a temperature slider until the wall looks neutral. The photographer has to observe whether the person still looks alive, dimensional and plausible under the intended light.
Colour Can Be Accurate and Still Be Bad
Scientific reproduction, product photography, fine-art documentation and medical imaging may demand disciplined colour accuracy. A fashion campaign, cinema still or personal photograph may deliberately depart from neutral colour to create mood. The same processing move can be an error in one job and a creative choice in another.
This is why “correct colour” is an incomplete phrase. Correct for what? Measurement? Memory? Print matching? Brand identity? Atmosphere? The job determines the tolerance.
Three Experiments
- Neutral object under three lights. Photograph a white or grey object in daylight, shade and indoor light using fixed white balance, then auto white balance. Compare what the camera chooses to neutralise.
- RAW white-balance test. If possible, shoot RAW and change white balance afterward. Observe how much freedom remains compared with a strongly processed JPEG.
- Screen comparison. View one image on several displays at similar brightness. Notice how device rendering changes your judgement of saturation and warmth.
Common Misconceptions
- “Each pixel records full colour.” In common Bayer systems, neighbouring photosites sample different colour bands and full RGB values are reconstructed.
- “White balance fixes bad lighting.” It can correct overall channel balance, but it cannot restore spectral information that the light source never provided.
- “RAW has no processing.” It is less processed output, not unmediated reality.
- “Kelvin describes all colour casts.” Many real illuminants also require tint correction and have complex spectra.
- “If it looks right on my screen, colour is finished.” Other displays and print conditions may reproduce it differently.
From Beginner to Advanced
At beginner level: light has colour, cameras need white balance, and screens differ. At intermediate level: understand Bayer sampling, RAW, demosaicing, colour temperature, tint and colour spaces. At advanced level: study spectral sensitivity, ICC profiles, camera characterisation, CIE colourimetry, metamerism, display calibration, rendering intents, chromatic adaptation transforms and multi-spectral imaging.
The deeper you go, the less colour looks like a paint label and the more it looks like a negotiated relationship among light, material, sensor, mathematics and observer.
For Parents and Teachers
Colour photography is an excellent bridge between physics and perception. Ask students to photograph the same object under different lights, then predict which change came from the object and which came from illumination. Let them discover that seeing is not passive recording.
The larger lesson is useful everywhere: measurements always come through an instrument, and instruments have response characteristics. Understanding the instrument helps us interpret the result.
Further Reading
Cambridge in Colour provides accessible technical guides to digital camera sensors and white balance. For deeper study, look toward colour-science texts, camera-characterisation literature and ICC colour-management documentation.
The Final Idea
The world does not arrive in the camera as finished colour. Light is filtered, sampled, compared, reconstructed, transformed and displayed. Photography becomes more interesting—not less—when we understand this. The colour photograph is not a bucket of colour collected from reality. It is a carefully engineered translation of spectral evidence into something our eyes can read.