How Depth Cameras Measure Distance

An ordinary photo is flat.

A depth camera adds a second picture in which every pixel is a distance. Robots, scanners and motion games use it to tell a hand from the wall behind it. There are three main ways to measure that distance, and each has a weak spot.

Updated 19 October 2026 · 4 sources

Two eyes: stereo
Painted dots: structured light
Stopwatch: time of flight
Three ways to measure distance

Pick a scene

THE SHORT ANSWER

Three ways
to measure depth.

Depth cameras output a colour image and a depth image.
They differ in how they find the distance.

STRUCTURED LIGHT

Paint the scene with dots

A projector throws a known pattern, often in infrared you cannot see, and a camera works out depth from how the pattern lands on each surface.

Pick a scene

STEREO

Two eyes, one answer

Two calibrated cameras find the same point in both pictures and use triangulation. Blank white walls and repeated patterns make that matching hard.

Where each works

TIME OF FLIGHT

Time the light

The camera sends out near-infrared light and measures how long it takes to come back. Lidar on survey aircraft uses the same idea with laser pulses.

See the numbers

A small plain white survey aircraft flying low over a winding river and coastline in early morning light — AI-generated image, not a photograph of a real person or place.

MATCH THE METHOD

Each method has
a favourite room.

No depth camera is best everywhere. The lecture notes point to a simple pattern of strengths.

Stereo likes texture
Two-camera systems do well on busy, detailed scenes and badly on blank or repeating surfaces.
Projected light likes the dark
Structured light and time of flight bring their own illumination, so they work on blank walls and at night.
Sunshine is the rival
Outdoors, sunlight can be stronger than the camera's own light and wash out the pattern.
Corners can echo
For time of flight, light that bounces between surfaces can make a corner look further away than it is.

TWO NUMBERS WORTH KNOWING

Light is fast,
so the clock must be faster.

Time-of-flight cameras measure how long light takes.
Both figures follow from the exact speed of light.

≈ 30 cm

Distance light travels in one billionth of a second

The speed of light in a vacuum is exactly 299,792,458 metres per second, so in one nanosecond light covers just under 30 centimetres.

SPEED OF LIGHT · NIST

≈ 6.7 ns

Time for light to reach something 1 m away and return

Distance is half the speed of light times the round-trip time. For an object 1 metre away, the light is back in about 6.7 billionths of a second.

ROUND TRIP · 1 METRE

Many time-of-flight cameras do not time single pulses: they send out light whose brightness rises and falls like a wave and measure the phase shift of the returning light, which converts to distance.

A QUICK SCENE CHECK

Will the camera
find the depth?

Pick a scene and a type of camera. The bar shows how well the lecture notes say that method copes.

Based on TU Graz Robot Vision lecture notes on depth sensing (2024). Real cameras vary in how well they handle each case.

Scene check

Choose a scenePlain white wall, two-camera stereoRepeating tiles, two-camera stereoDark room, projected infraredBright sunshine, projected infraredCorner of a room, time of flight

Plain white wall, two-camera stereoStruggles

Passive stereo needs texture to match points between the two pictures. Untextured areas such as white walls are a known problem.

Repeating tiles, two-camera stereoEasily confused

Repeated structures make it hard to tell which tile in one picture matches which in the other: the correspondence problem.

Dark room, projected infraredWorks well

Active systems bring their own light, so structured light and time-of-flight cameras work on blank areas and in the dark.

Bright sunshine, projected infraredCan falter

Strong sunlight can swamp the projected light. A stereo camera with a dot projector then falls back to ordinary stereo matching.

Corner of a room, time of flightSome error

Light can bounce off more than one surface before returning, an error source called multi-path interference.

Source: Prof. Friedrich Fraundorfer, Robot Vision: Depth sensing, TU Graz (2024).

Two identical small camera lenses mounted side by side on a plain metal rail on a workbench, like a pair of eyes — AI-generated image, not a photograph of a real person or place.

FROM ROOMS TO COASTLINES

The same idea,
from the air.

Lidar, short for light detection and ranging, uses pulses of laser light to measure distances. NOAA flies it on aircraft to map shorelines and build elevation models.

Topographic lidar usually uses a near-infrared laser to map land, while bathymetric lidar uses green light that can pass through water to measure the seafloor and riverbeds. Millions of measured points form a 3D point cloud.

In conversation

Questions people ask

What is a depth camera?

A camera that outputs a depth image as well as an ordinary colour image, so that every pixel carries a distance. Depth cameras are especially important in mobile robotics.

Why can't you see the dot pattern?

Systems that project a pattern usually do it in the infrared, which people cannot see. That also lets them work in the dark.

Why do depth cameras struggle with white walls?

Passive stereo cameras find depth by matching the same detail in two pictures. A blank wall has no detail to match. Active systems avoid this by projecting their own pattern.

Do depth cameras work outside?

Active ones can struggle, because sunlight can be stronger than the light they project. Stereo cameras with a dot projector fall back to ordinary stereo matching in bright sun.

How is time-of-flight different from lidar?

Both time how long light takes to return. A time-of-flight camera measures a whole grid of pixels at once; airborne lidar scans laser pulses across the ground and builds a point cloud.

Sources

Each figure on this page comes from the sources below.

Seeing how far,
not just what.