Who this suits: Businesses that need useful facial detail at entrances or other defined points.
Direct answer: resolution alone cannot answer it
A camera may identify a face only when the complete recorded image provides enough useful detail under the actual conditions. Distance, field of view, face angle, light, movement, exposure, focus and compression all matter. The fact that a camera is 4K does not establish an identification distance. Digital zoom enlarges recorded pixels; it does not recover facial detail that was never captured.
Start by defining a target point, likely direction of travel and the evidence you need. Verifying a known employee is different from trying to establish an unknown person's identity. This guide explains a planning calculation and acceptance approach. It does not promise an investigative outcome or replace a site-specific design.
Understand pixel density at the target
Pixel density expresses how many image pixels represent a real-world distance at the location being assessed. A simple horizontal estimate is image width in pixels divided by scene width in metres. A person near the camera and a person far away occupy different parts of the geometry, so one density figure cannot describe every depth in a wide scene.
Use the recorded image dimensions and the actual width at the target plane. Do not use a low-resolution live-view tile or assume the recorder saves the camera's maximum available resolution. If the design uses cropping, a panoramic view or dewarping, have the designer explain the usable detail in the view staff will actually retrieve.
Current standards terminology needs care
Older DORI material based on IEC 62676-4:2014 often labels 250 pixels per metre as identification. Axis's September 2026 explanation of IEC 62676-4:2025 uses a different model: 250 pixels per metre is Characterize, 500 is Validate and 1,500 is Scrutinize. Do not combine an old distance table with a newer label and assume the meaning is unchanged.
These are planning categories for human interpretation, not universal pass-fail guarantees or face-recognition analytics specifications. Ask a supplier which model, edition and task its stated distance uses. A proposal should explain the required result in plain language and verify it through actual recordings. A standards label alone is insufficient evidence that an unknown face will be identifiable in your particular incident.
Worked geometry with explicit assumptions
Assume a camera records 3,840 horizontal pixels and the scene is eight metres wide at the entrance line. The simple density is 3,840 divided by eight, or 480 pixels per metre. If a face is assumed to be 0.16 metres wide, it occupies roughly 77 pixels across at that plane. The assumed face width is an illustrative planning approximation, not a measurement of every person.
If the same image covers sixteen metres at the relevant plane, density falls to 240 pixels per metre and the assumed face width to about 38 pixels. The camera's resolution has not changed; the allocation of detail has. Neither calculation accounts for blur, view angle, shadow or compression. It is a reason to review the lens and position, not a promised identification result.
Position can matter more than a resolution upgrade
A camera mounted very high may capture the top of a head rather than a useful face. A broad yard view can show movement while leaving every person small. Consider a defined capture point such as an entrance or pedestrian route, where a suitable angle and narrower view are practical. Keep a separate overview if the wider sequence remains important.
Avoid forcing people into an unsuitable or unsafe path merely to improve the image. Physical design, access needs and privacy still matter. If subjects can approach from several directions, a single view may not capture a useful face each time. Document the likely directions and residual limitations rather than promising complete identification coverage.
Motion and lighting can defeat the calculation
A mathematically dense image can still be unusable when a person moves during a long exposure. Frame rate is the number of images produced per second, while exposure is how long each image gathers light. Increasing frame rate alone does not necessarily remove motion blur. Ask the installer to set and test the whole exposure approach for the task.
Check night-time lighting, backlit doors and reflections. Extra illumination can help, but it needs appropriate design and installation. A bright colour still of an empty scene is not proof. Test a walking subject in normal clothing, at normal speed, with the intended lighting state. Use the exported recorded stream rather than a more flattering live stream with different settings.
Define an honest acceptance test
Agree the task, route, target point, day and night conditions, recorded settings and viewing method before purchase. Ask an authorised reviewer to inspect footage without relying on the installer narrating who the subject is. For a known-person verification task, use a suitable controlled test and record exactly what was established.
Keep the approved sample securely and note limitations. Do not publish customer or staff footage as a marketing example without the required authority. A test can show that the proposed setup meets a practical agreed task under tested conditions. It cannot prove the same result for every disguise, angle, weather condition or future incident.
Ask for the assumptions behind a distance claim
If a supplier quotes an identification distance, ask for the image dimensions, lens setting, target width, pixel-density model, subject angle and light assumptions supporting it. Then ask which of those conditions will exist at your site. A distance calculated from a catalogue lens value is different from a verified night-time walking test at the proposed mounting position.
Record how the footage will be reviewed. A small compressed phone preview is not equivalent to an original export on a suitable display, and a full-screen display cannot invent absent samples. Preserve original recordings when an incident matters and avoid repeatedly converting files before appropriate review. The buying decision should establish a practical capture and retrieval chain, rather than a single impressive-looking number detached from the scene.
Frequently asked questions
Will 8MP identify someone farther away than 4MP? It can provide more samples over the same scene width, but the improvement depends on the actual horizontal pixel counts, lens and image quality. Doubling total pixels does not double horizontal detail. A narrower 4MP view may provide more face pixels than a very wide 8MP view.
Can software sharpen an unreadable face later? Processing can sometimes help interpretation, but do not buy a system on the assumption that missing detail will be reconstructed reliably. Enlarged or generated-looking detail is not equivalent to information captured by the original camera. Preserve original exports and use appropriate expert advice for evidentiary handling.
A better buying brief
Write a requirement such as: capture usable facial detail for a person walking through this entrance, at this target line, during ordinary daytime and after-hours lighting. Ask for the proposed geometry, recording settings and test. Require the supplier to state which conditions remain outside scope. That is more reviewable than a headline identification distance.
Complete the camera planning questionnaire and review office security planning or shop security planning as relevant. You can request equipment options against the resulting brief. BoxFreeIT and Security Wholesalers share ownership, so any related commercial suggestion is not an independent market award. Prefer demonstrated recorded detail at your required point over a specification label or an optimistic zoom demonstration.
Sources and review notes
Official references used for this guide. Product features and terms can change. Examples are illustrative unless explicitly described otherwise.