How many megapixels are enough in video surveillance?

One of the reasons behind the shift from analogue to IP was the opportunity to increase resolution. Early in the shift, there was a race among vendors competing to have the highest megapixel camera at the industry’s top tradeshows. The industry has changed again and is now taking a much more holistic approach to image and video quality. Standards such as HDTV at 720 or 1080, Quad HD and 4K have become much more important when talking about megapixels. Lastly, there are still some applications where very high resolution – over 10 or even 20 megapixels – makes sense, such as venues and stadiums where you don’t want to intimidate the crowd with too many cameras, but still want to see the details.

Customers look for more megapixels in the hope of getting better clarity at a distance. They want to be able to zoom into a point some distance away from the camera and see enough pixels on target to detect human beings or recognise licence plates. In perfect lighting, higher megapixels generally deliver, assuming the resolving power of the camera and lensing is indeed delivering to specification. However, when challenging lighting conditions are the norm, you need other capabilities as well. Users need great low-light performance, and they need true Wide Dynamic Range (WDR) to draw out detail from both bright and dark areas. Low-light and WDR performance has lagged every transition in resolution. For example, increasing resolution typically means reducing pixel size on sensors which reduces light-gathering capability and adversely impacts low-light performance. Without adequate low-light and WDR performance. you just have high numbers – of empty megapixels.

From my experience, the answer is: “As many as you need for the application.” Put another way, each end user must determine how many pixels on target they need for each camera. The industry commonly holds that 20 pixels/foot is enough for general surveillance, 40 pixels/foot is the minimum for facial recognition and license plate identification, and 80 pixels/foot is used for higher detail like reading logos, names embroidered on a shirt, etc. So, if it’s a camera in a small lobby and you only need basic contextual information, standard definition is probably enough resolution. For super large scenes, such as airport runways, large parking lots and sea ports, it’s more a matter of divide-and-conquer. Wide-angle cameras provide contextual information, and cameras with more resolution cover key points. The question of “How many megapixels is enough?” doesn’t have a simple answer, because it depends on what the end user needs.

For the vast majority of applications, HD to 5-megapixel resolution is more than good enough. The challenge is that commercial network infrastructure and storage continue to be limiting factors. Most of our commercial customers have sub-1 megabit bandwidth Internet connections to remote locations, which is not fast enough to stream a single 2-megapixel camera. A connection fast enough to support a good viewing and playback experience may cost an extra hundred dollars a month per location. For companies with hundreds of locations, this is a significant cost. We currently have many customers that run their cameras at a lower resolution than they are capable of. We promote that every customer should have a camera or two at a high-enough resolution to obtain clear, mugshot-quality images of all people who enter their facilities, and to be able to see bill denominations in areas where cash is handled.

Required video resolution depends entirely on the application, which considers the placement of the cameras and the information required from them. In most indoor applications with cameras mounted on ceilings that are considered at normal height – where general surveillance is required – the commonly available 1MP to 2MP resolutions are good enough. If precise identification is required, such as licence plate numbers or facial recognition, then the resolutions may have to be higher (e.g., 3MP to 5MP). For outdoor applications, with cameras placed at significant heights, or with those that are expected to cover large areas, then higher resolutions – from 5MP to 4K – may be required. Omnidirectional cameras – both multi-sensor and fisheye – that provide large coverage areas typically have higher resolutions; even up to around 50MP. The requirement for greater resolution needs to be balanced with the cost of the supporting infrastructure, namely the networking equipment, storage and the display monitors.

Most applications can be satisfied using 2 MP (1920 x 1080) video resolution. The cost versus performance makes sense as the price for true day/night Full HD IP cameras continues to drop and will satisfy the video surveillance pixel-per-foot (PFF) requirements for most applications (i.e. observation 20 to 40 PPF, forensic review 40 to 80 PPF, recognition 80 PPF and up). Also, the low light capabilities are still better with the 2 MP cameras over the higher MP cameras. However, for applications that require a wider field of view, you go with the 5 MP and higher MP cameras as required to properly satisfy the requirement. In summary, use the right camera for the application to support the environment, lighting conditions, etc. and do the math as most of the time the 1920x1080 resolution is more than enough to satisfy the requirement.