Bullet camera mounted on a wall covering a building entrance
CCTV & Surveillance

How Many Pixels Do You Need? Choosing CCTV Resolution by Pixel Density

Pixel density in pixels per metre tells you whether a camera can identify a face or only observe a scene. A short method with a worked example for 2, 4 and 8 megapixel cameras.

Shruti IT Service Engineering Team

Survey, design and maintenance engineers, Nashik

Published On

16 Nov 2026

Read Time

6 min read
Share Guide:
Bullet camera mounted on a wall covering a building entrance

Megapixels sell cameras, but they do not tell you what the camera will show. A 4 megapixel camera pointed at a 40 metre wide yard cannot identify anyone in it, and an 8 megapixel camera on a corridor wall is wasted. The measure that links the camera to the job is pixel density: how many pixels fall on each metre of the scene.

Key Engineering Takeaways

  • Pixel density is horizontal resolution divided by the width of the scene at the target distance.
  • IEC 62676-4 describes common targets: about 25 pixels per metre to detect, 62 to observe, 125 to recognise and 250 to identify a person.
  • Lens angle and distance decide the scene width, so the same camera gives very different density at 5 m and 15 m.
  • Pixel density sets a ceiling. Light, shutter speed and compression decide whether you reach it.

The four jobs a camera can do

IEC 62676-4, the video surveillance system design standard, groups tasks by how much detail you need. Detection means noticing that something is there, around 25 pixels per metre. Observation means following what a person is doing, around 62. Recognition means knowing a familiar person, around 125. Identification means being able to tell who someone is with confidence, around 250.

Before you choose a camera, decide which of these each location needs. A perimeter fence needs detection. A shop floor may need observation. A reception desk, a cash counter and a gate need identification.

The calculation

Pixel density in pixels per metre equals the camera's horizontal pixel count divided by the width of the scene in metres. A 4 megapixel camera with 2688 horizontal pixels covering a 10.75 metre wide scene gives 2688 / 10.75 = 250 pixels per metre.

Scene width comes from the lens. For a camera with a horizontal field of view of 90 degrees, the width at distance d is 2 x d x tan(45 degrees), which equals 2 x d. At 10 metres the scene is 20 metres wide. The same 4 megapixel camera then gives 2688 / 20 = 134 pixels per metre. That is enough to recognise a familiar person, not to identify a stranger.

A quick reference for three resolutions

The widest scene that still meets each target is the camera's horizontal pixel count divided by the target density.

  • 2 megapixel (1920 pixels wide): 7.7 m for identification, 15.4 m for recognition, 31 m for observation.
  • 4 megapixel (2688 pixels wide): 10.8 m for identification, 21.5 m for recognition, 43 m for observation.
  • 8 megapixel (3840 pixels wide): 15.4 m for identification, 30.7 m for recognition, 62 m for observation.
Use these as the widest scene, not the nearest. If a person may stand anywhere across a 12 metre gate, a 4 megapixel camera cannot identify them at the far edges. Use two cameras, a narrower lens or a higher resolution.

Why the number is a ceiling, not a promise

Pixel density assumes a sharp, well lit picture. In practice three things pull the result down. Low light forces a slower shutter and higher gain, which blurs movement and adds noise. Strong backlight, such as a bright window or a sunlit gate, hides faces unless wide dynamic range is set up correctly. Heavy compression smooths away fine detail even when the pixel count is high.

That is why a design should be checked on site. Mount a camera temporarily, view it at the real distance in the worst light the location will see, and confirm that a face or a plate is readable before cabling and mounting are finished.

Where dedicated cameras are the better answer

Reading number plates at a gate is a different job from identifying people. Plate reading needs the right lens, a fast shutter, controlled infrared lighting and a narrow lane, and it is usually done by a dedicated camera with analytics rather than by a general purpose camera. If plate capture matters, plan it as its own camera position.

Conclusion & Next Steps

Decide the task for each camera, work out the scene width, and check the density before you buy. A one page camera schedule that lists location, task, distance, lens angle, scene width and resulting pixels per metre is a good sign of a well designed system, and you are entitled to ask for it.

Certified Engineering Partner

Need a site survey or project quote?

Our engineers provide comprehensive BOQs, system designs, and turn-key execution across Maharashtra and neighboring states.

Related services

Want this done at your site? These pages explain the scope, the process and the standards we work to.