Skip to content
AI Camera SystemsGuides & tools
Hardware

Camera Resolution and Lens Choice: A Practical Guide

More megapixels do not automatically mean a more useful picture. What matters is how many pixels land on the thing you care about, and whether there is enough light to use them.

5 min read Updated 1 October 2026 General information, not legal advice

Resolution is the number on the box. Pixels per metre decides whether footage is evidence or just a record that something happened. The two are related, but a 4K camera on the wrong lens at the wrong height loses to a 2MP camera placed properly.

Pixels on target, not pixels on the sensor

An 8MP camera spreads 3,840 horizontal pixels across whatever the lens shows it. Pointed at a two-metre doorway, that is enormous detail. Pointed across a forty-metre yard, each metre gets fewer than a hundred pixels.

Pixels per metre is simply horizontal pixels divided by the width of the scene at the distance you care about. Commonly quoted targets are roughly 250 pixels per metre to identify a stranger, 125 to recognise someone you know, and 60 to 80 to observe behaviour.

That gives a theoretical maximum scene width for each resolution:

Camera Horizontal pixels Width for identification Width for recognition
2MP (1080p) 1,920 about 7.5 m about 15 m
4MP 2,560 about 10 m about 20 m
5MP 2,880 about 11.5 m about 23 m
8MP (4K) 3,840 about 15 m about 30 m

Read those as best cases: good light, a decent lens, a subject not moving fast. Lens softness, compression, motion blur and dim lighting all eat into them, so plan to sit comfortably inside the figure. The camera count tool applies the same logic area by area.

2MP, 4MP and 8MP in practice

2MP (1080p) is still sensible for close work: a doorway, a till, a small room. Its pixels are physically larger on most sensors, which usually means better low-light behaviour than a higher-resolution camera of the same size and price.

4MP is the common default for business sites. It gives meaningfully more width than 1080p without the storage and night-time penalties of 4K, and it suits aisles, car parks and general coverage. If you are unsure, this is usually where to start.

8MP (4K) earns its place on wide scenes needing detail across the width: a large yard, a long dock face, an open retail floor from a high mount. The costs are real — roughly double to triple the storage of 4MP, as the storage guide sets out, more bandwidth, and often weaker night performance because each pixel collects less light. A 4K camera in a dim yard can be noisier and less usable than a 4MP camera in the same spot.

Choosing the lens

Fixed lenses are cheaper and reliable. Varifocal lenses can be adjusted on site, which is valuable where the right framing is not obvious from a plan. As a rough guide on the sensor sizes used in business cameras:

  • 2.8mm — about 100 to 110 degrees. Doorways, small rooms, tight corners.
  • 4mm — about 80 to 90 degrees. The general-purpose choice.
  • 6mm — about 50 to 55 degrees. Longer aisles, driveways, dock approaches.
  • 8mm to 12mm — about 25 to 40 degrees. Gates, specific targets at distance.
  • Above 12mm — narrow, specialist views such as plate capture.

Wide lenses also distort. At 2.8mm, objects near the edge of frame are stretched and softer than those in the centre, so put what matters in the middle third.

Mounting height and angle

The most common installation mistake is mounting too high. A camera at four or five metres looking down at 45 degrees gives you shoulders, hats and the tops of heads. It is excellent for showing where people walked and useless for showing who they were.

Practical habits:

  • For identification, mount around 2.4 to 3 metres and keep the vertical angle under about 30 degrees.
  • Aim across the direction of travel. A person walking towards a camera presents a face for a frame or two; a person crossing it stays in frame.
  • Put identification cameras just inside doorways, where everyone passes at a known distance.
  • Save high mounting for context cameras, where the job is the sequence of events.

Low light is where systems disappoint

Most incidents happen when the site is dark, so night performance matters more than the daytime image.

What actually helps:

  • Site lighting. A modest amount of real light beats most camera upgrades, and lets cameras run in colour at night, which preserves clothing and vehicle colour.
  • Larger sensors and faster lenses. A bigger sensor with a lower f-number gathers more light. Low-light and colour-at-night models are built around this.
  • Sensible shutter speeds. Cameras brighten dark scenes by holding the shutter open longer, which blurs anything moving. Capping the slowest shutter trades brightness for usable detail.

Infrared or white light

Infrared is invisible, included on most cameras, and monochrome. It works, but the quoted range assumes a reflective subject, range drops in rain and fog, insects are drawn to the glow, and dark clothing can vanish. You also lose all colour, which matters when a description is the point.

White light keeps the image in colour, deters people who notice they are lit, and helps both your eyes and any detection software. It also annoys neighbours, can be unwelcome on a residential boundary, and draws attention to the camera. Many cameras offer both, with white light triggered by an event.

Where white light or good site lighting is practical, it is usually the better outcome. Colour at night is far more useful than monochrome in a report.

WDR for doorways and dock openings

A camera looking at a doorway in daylight is handling two scenes at once: a bright outside and a dim inside. Without help, you get either a blown-out white rectangle or a silhouette with no face.

Wide dynamic range combines differently exposed captures to hold detail in both. For any camera pointed at a door, roller door, window or dock opening, treat true WDR as a requirement rather than a feature. It still works best when the camera is not fighting direct sun, so aim across the opening rather than straight out of it.

Putting it together

Pick the purpose, then the pixels per metre, the scene width, the lens and the mounting point — checking lighting before any of it. For a plan built around your own site, ask for a quote. Terms used here are in the glossary, and the guide index covers the rest of the planning process.

Frequently asked questions

Is 4K worth it for business CCTV?

On wide scenes where you need detail across the width, yes. On a doorway or a till, a 4MP or even 2MP camera placed properly will do the same job with less storage and better night performance.

Why does my camera look sharp in the day and useless at night?

Almost always light. The camera slows its shutter and raises gain to compensate, producing blur and noise. Site lighting, or capping the slowest shutter speed, usually fixes more than a new camera would.

What lens should I use for a car park?

Often 4mm for general coverage plus a longer lens on the entry, because a single wide camera cannot both show the whole car park and read anything on a vehicle — see the plate camera guide.

Do more megapixels make AI detection better?

Up to a point. Detection needs enough pixels on the subject, so resolution helps at distance. Beyond that, lighting and lens quality matter more.

Should I use varifocal or fixed lens cameras?

Fixed lenses are fine where the view is obvious. Varifocal is worth the extra on gates, long aisles and anywhere you will want to adjust framing after seeing the real image.

Want this set up properly?

AI Camera Systems installs camera systems on the Gold Coast, in Brisbane and across South East Queensland, and supports the AI monitoring layer Australia-wide.