Short answer
Because most of what the camera records from them is not their own colour. A glossy or mirror-like surface returns a picture of the lights and the room; a transparent or translucent one passes on whatever is behind it, tinted by its thickness. Their apparent colour therefore changes with the angle, the surroundings and the backing far more than a matt object's does. Getting it right is a matter of controlling what the object reflects and what it sits in front of, and only then of camera settings.
The standard physical description of a painted, plastic or glazed surface splits what it returns into two parts. Body reflection is light that entered the material, was coloured by its pigment and scattered back out in all directions; this is the object's colour. Interface reflection bounces off the surface itself without entering, and on such materials it keeps the colour of the light source. The camera records their sum. On a matt object the interface part is spread thinly everywhere. On a glossy one it is concentrated into highlights that are the colour of the lamp, and any pixel under a highlight is paler and less saturated than the object. Where the highlight clips, the object's colour is gone entirely.
A clean glass or lacquered surface reflects about 4% of light arriving head-on, and much more as the angle becomes shallow: over a sixth at 70° and well over a third at 80°, as the table shows. Whatever lies in the mirror direction is therefore laid over the object, in its own colours. A green wall beside a glass bottle shows up as a green edge; a white ceiling shows as a milky veil across a dark glossy panel. The reflection is polarised, most completely near 56° for glass, where one polarisation is not reflected at all. That is why a polarising filter on the lens, rotated to block the reflected polarisation, removes much of the reflection at around that angle and almost none of it head-on.
Polarisers also cut reflections from foliage, water and wet surfaces, which is why colours under them look deeper: the pale surface reflection has gone and the body colour is left.
Bare metal has no body reflection to fall back on. Everything it shows is a reflection of its surroundings, tinted by the metal: gold reflects yellow and red light far more strongly than blue, so even its highlights are coloured, while a metal that reflects all visible wavelengths about equally returns its surroundings untinted. A polished spoon photographed in a room is a distorted picture of that room. Its colour in the photograph can only be controlled by deciding what it reflects, which is why metal is lit with large white surfaces for brightness and black cards for definition. A lens polariser on its own does little here, since it relies on the polarisation that reflection from non-metallic surfaces produces; polarising the lights as well, and crossing the lens filter against them, is the method used in conservation photography.
A transparent or translucent subject is coloured by absorption along the path the light takes through it. A tall glass of tea is darker and more saturated than a shallow one of the same tea, and the same drink looks different in a wide tumbler and a narrow flute. Instrument makers' guidance on measuring such samples makes the related point that the reading is affected by whatever lies behind, and recommends either a thickness great enough to stop light passing through or a fixed white backing. Fabric lit from behind is the same case: the camera sees light that has passed through the dyed fibres rather than bounced off them, often brighter and more saturated than the front-lit cloth, and quick to clip a channel.
Start with the surroundings, because they are part of the picture. Use a light source much larger than the object so that reflections become broad, even fields rather than hot spots, and remove or cover coloured things in the mirror direction. For glass, light the background and let it shine through, or place black cards just out of frame to draw dark edges. For liquids, fix the container, the fill level and the backing, and say what they were. For reproduction work on glossy or varnished surfaces, cross-polarise. Then set white balance from a grey card under the same light, and expose by the per-channel histogram, since a saturated transmitted colour or a tinted highlight clips one channel long before the picture looks bright.
| Angle from perpendicular | Glass, unpolarised | Glass, s-polarised | Glass, p-polarised | Water, unpolarised |
|---|---|---|---|---|
| 0° | 4.0% | 4.0% | 4.0% | 2.0% |
| 30° | 4.2% | 5.8% | 2.5% | 2.1% |
| 45° | 5.0% | 9.2% | 0.8% | 2.8% |
| 56.3° (Brewster's angle for glass) | 7.4% | 14.8% | 0.0% | 4.7% |
| 70° | 17.1% | 30.0% | 4.2% | 13.3% |
| 80° | 38.8% | 53.9% | 23.7% | 34.8% |
| 85° | 61.3% | 73.2% | 49.3% | 58.4% |
Why: A wall, garment or object of that colour sits in the mirror direction of that edge.
Fix: Move it or cover it with a white or black card, depending on whether you want a light or dark edge.
Why: The metal is reflecting a dim, featureless room.
Fix: Surround it with large white surfaces for brightness and add narrow black cards so the form has edges.
Why: Surface reflection of a bright ceiling or softbox is laid over the body colour.
Fix: Change the angle so the reflection misses the lens, flag the source, or cross-polarise.
Why: The glass shape, fill level or background changed, altering the path length and the light behind.
Fix: Standardise the vessel, the depth and a white backing, and light it the same way each time.
Why: Transmitted light is bright and saturated enough to clip one channel.
Fix: Expose for that channel using the per-channel histogram, and add front light to bring the rest up.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
FactStrong evidence
The dichromatic reflection model describes light from a surface as the sum of interface (specular) reflection, which forms the highlights, and body (diffuse) reflection, each with its own spectral distribution scaled by geometry.
Caveat: A model for materials such as paints, plastics and glazes; bare metals have no body term.
FactStrong evidence
Light reflected from a non-metallic surface is partially polarised and is completely polarised at Brewster's angle, where the reflection coefficient for light polarised parallel to the plane of incidence falls to zero.
Source: HyperPhysics (optics, atmospheric optics and vision pages); Polarizing filter (photography)
Colourwise analysisStrong evidence
For glass of refractive index 1.5, a single surface reflects 4.0% of unpolarised light at normal incidence, 17.1% at 70° and 38.8% at 80°, and Brewster's angle is 56.3°.
Based on: Computed by Colourwise from Fresnel's equations for one air–glass surface with n = 1.5, and Brewster's angle as arctan(n). See the table.
Caveat: One surface of an ideal smooth dielectric; a real pane has two surfaces and coatings change the figures.
Source: HyperPhysics (optics, atmospheric optics and vision pages); List of refractive indices
ConventionModerate evidence
In cross-polarised photography of artworks, polarising sheets on the lamps and a polariser on the lens rotated to 90 degrees extinguish glare, because specular reflection keeps the lamps' polarisation while diffusely scattered light does not.
FactStrong evidence
Colour measurements of semi-transparent samples are influenced by what lies behind them; recommended practice is to increase thickness until no light passes through or to place a white surface behind the sample.
Source: Precise Color Communication
FactModerate evidence
Gold reflects yellow and red light strongly, which is why its reflections, unlike those from a clear lacquer or glass, carry the metal's own colour.
Source: Gold
Colourwise interpretationModerate evidence
For a mirror-like or transparent object, the colours recorded in a photograph are set more by the lights, surroundings and backing than by the object, so controlling those comes before any camera setting.
Based on: Follows from the dichromatic model (interface reflection carries the source's colour), from the Fresnel figures in the table, and from instrument guidance that translucent samples take on their backing. Not a measured comparison of photographs.
Reviewed 1 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.