CCTV Lens Size for Face Recognition at 10 Metres (2026)

For a typical 4MP camera, a lens of roughly 4-6mm will cover a 10 m wide scene at identification density, which is about 250 pixels per metre on the commonly used DORI scale. An 8MP camera reaches the same density across a wider scene, so it can use a shorter lens. That only holds for a face-on subject in good light. The pixel-density target, the mounting height and the viewing angle matter as much as the lens number, and the formulas below show how to work it out for your own camera.

What lens size do I need to see a face at 10 metres?

Focal length equals sensor width × 10 m ÷ frame width, and frame width is image width in pixels ÷ your PPM target. Both formulas are set out on the DHI lens calculator page. Combined for a 10 m subject, they give: focal length (mm) = sensor width (mm) × 10 × PPM ÷ image width (px).

The three inputs

• Sensor width: read it from the camera datasheet. It is not the same as the "1/3-inch" style label.
• Image width: a 4MP camera typically outputs 2560×1440, so use 2560 px.
• PPM target: the pixels per metre of scene width you need at the subject.

For the PPM target, the DORI model sets one level per task: Detect, Observe, Recognise and Identify, as described in the Axis pixel-density white paper. Each tier needs more pixels per metre than the one before. Pick the tier first, take its figure from that paper or the standard, and drop it into the formula. A face-ID target at 10 m needs a much longer lens than simply noticing that someone is there.

DHI's own example shows the maths in use: a 5.60 mm sensor at 20 m with a 4.00 m frame needs 28 mm. A 1,920 px image across that frame gives 480 px/m.

Megapixels do not change the view

Field of view depends only on sensor size and focal length, as the cctvplanner FOV calculator explains. For example, a 4.0 mm lens on a 1/3" sensor covers 12.00 m of width at 10 m. Across 2560 px that is roughly 213 px/m.

More megapixels put more pixels inside the same view, which raises PPM. A longer lens narrows the view to the same effect. The tiers also assume a subject facing the camera in full daylight, per NtechLab's camera requirements, so allow margin for poor light or side-on angles.

How do I calculate the focal length for a face?

Multiply the sensor width by the distance, then divide by the frame width you need: f = (W × D) ÷ F. Three steps get you there.

1. Pick a pixel-density target (PPM). This is how many pixels per metre must land on the subject.
2. Find the frame width. F = image width in pixels ÷ PPM.
3. Apply the formula. f = focal length (mm), W = sensor width (mm), D = distance (m), F = frame width (m).

Worked example: take a 1920 px wide image and 480 px/m, the density used in the DHI calculator example. That gives F = 1920 ÷ 480 = 4.00 m. With the illustrative 1/3" sensor at about 4.8 mm wide and D = 10 m, f = 4.8 × 10 ÷ 4.00 = 12 mm.

The 4.8 mm width is a simplification. It is the figure implied by the cctvplanner example, where a 4 mm lens at 10 m covers 12.00 m. Real sensors vary, so take the width from the datasheet. If it isn't listed, back-calculate it from the stated horizontal field of view: W = 2 × f × tan(HFOV ÷ 2). The DHI tool does exactly this, turning 2.8 mm and 90° into 5.60 mm.

A calculator is only as good as the sensor width you give it. A wrong width shifts the answer by the same proportion.

The maths also assumes a pinhole lens and a subject far away compared with the focal length. Real lenses lose some detail towards the edges of the frame, so keep faces near the centre rather than planning them at the margins.

2.8mm vs 4mm vs 6mm vs varifocal: worked example at 10 m

At 10 m, a 4mm lens gives about 213 PPM on a 4MP camera, 6mm gives 320 PPM, and 2.8mm falls to around 149 PPM. These figures assume a 1/3" sensor 4.8mm wide and measure the centre of the frame, where the picture is sharpest.

Lens	Frame width at 10 m	4MP (2560 px)	8MP (3840 px)
2.8mm	~17.1 m	~149 PPM	~224 PPM
4mm	12.0 m	~213 PPM	320 PPM
6mm	8.0 m	320 PPM	480 PPM
12mm (varifocal long end)	4.0 m	640 PPM	960 PPM

The 4mm row matches the 12.00 m width the CCTV Planner FOV calculator gives for a 1/3" sensor at 10 m. The method is the one in the DHI lens calculator: frame width equals image width divided by PPM. Focal length then equals sensor width × distance ÷ frame width. Always take the real sensor width from the datasheet, as 1/3" sensors vary slightly.

Working backwards from a target

For 250 PPM, the density commonly used as an identification target, a 4MP camera needs about 4.7mm and an 8MP about 3.1mm. For 125 PPM, a recognition-level target, a 4MP needs about 2.3mm and an 8MP about 1.6mm. So an 8MP camera can hold identification detail with a wider lens than a 4MP one.

The cost of zooming in

A longer lens narrows the view. A 6mm lens gives an 8 m wide frame at 10 m, which suits a doorway or till but covers an open shop floor poorly. A fixed-lens static camera locks in whichever compromise you pick at purchase.

Where varifocal fits

A motorised varifocal turret, typically 2.8–12mm, spans every row of the table. Your installer sets zoom and focus on site and can re-frame later without swapping the camera. Before buying, check the datasheet for the IP and IK ratings, the IR range and whether it is PoE-powered. The ratings and IR range matter for an outdoor mount, and PoE affects your cabling and switch budget.

Does mounting height and angle change the lens you need?

Only slightly: height barely changes the slant distance, but a steep downward angle spoils face detail far more than lens choice does. Slant distance is √(height² + horizontal distance²), so a camera 3 m up covering a subject 10 m away is working at about 10.4 m. That rarely justifies a longer focal length.

The real penalty is geometry. From a high mount, a subject walking towards you shows the camera a forehead or cap peak instead of eyes and nose. Lowering the mount, or aiming along the approach path rather than across it, will do more for recognisable faces than a lens change.

Typical mounting heights by site

General installer guidance, not manufacturer data:

Location	Typical mount height	Distance vs tilt compromise	Usual camera style
Entrance door	Roughly 2.4–3 m	Lower end keeps the angle shallow; aim along the approach so faces turn towards the lens	Fixed turret or bullet
Shop till	Roughly 2.4–3 m	Subjects are close, so tilt is the issue, not reach	Fixed turret
Shop floor corner	Higher, near ceiling	Covers the whole room, but views shoppers from above and across	Wide fixed turret
Car park pole or wall	Higher still	Long reach pulls towards a narrower lens, while height steepens the tilt	Varifocal bullet

For a CCTV camera for a retail shop, that usually means a wide camera in the corner for overview and a separate camera at the door or till for faces.

Why wide lenses suit corners, not faces

A 2.8mm turret spreads its pixels over a huge view. The DHI lens calculator uses a 2.8mm lens with a 90° horizontal field of view on a 1920 px wide image, while cctvplanner lists 120° for 2.8mm. Check your own camera's datasheet, since the figure depends on sensor size.

At 90°, the frame is about 20 m wide at 10 m, which works out to roughly 96 px/m. That is fine for seeing someone enter, but thin for a face. A varifocal lens lets you tune the view after mounting, which helps when the real viewing distance differs from your estimate.

What else stops a face from being identifiable?

Poor light, motion blur, heavy compression and a low-resolution recording stream can ruin a face as surely as the wrong lens.

Light and shutter

A walking subject smears if the shutter is slow, and a bright doorway behind them turns the face into a silhouette. NtechLab's camera requirements stress even lighting, sensible exposure and WDR (wide dynamic range) for this reason. For dark scenes, Hikvision's manual describes a Face Enhancement option that helps face-capture devices produce clearer pictures at night.

Human viewer versus software

Identifying someone by eye and matching them with software are different jobs. NtechLab quotes 500 px/m, about 80 px of face width, for face recognition. Treat that as the software target and don't assume a lens chosen for a human viewer will meet it.

Recording settings

A 4K camera recorded through a low-resolution substream throws away the detail you paid for. Set the NVR to record the main stream and don't starve it of bitrate. The NVR and 4MP vs 8MP guides on this site cover that in detail.

Plates and privacy

CCTV ANPR cameras are a separate job with their own lens and shutter setup, so don't expect a face camera to read plates. On the legal side, UK GDPR and the ICO's CCTV guidance can apply when cameras capture people beyond your own private space, so plan signage and a stated purpose. Software that matches faces is a bigger step than recording, and worth getting advice on first.