Cameras analyse images in ever-shorter times, video management systems consolidate numerous data sources, and control centres are expected to respond to events as promptly as possible. However, between capture and perception lies a technical chain comprising image processing, transmission, decoding and display. For operators of security-critical video systems, this brings a factor into sharper focus that has often remained in the background until now: end-to-end latency.

Modern video surveillance promises speed. Cameras detect movement, classify objects or report defined events. In the control centre, however, this only becomes actionable information once the relevant image is actually visible.

There are several processing steps between the camera sensor and the screen. The image is captured and processed; in the case of IP video, it is usually compressed and packaged into packets, then transmitted, buffered and decoded at the receiving end, and finally displayed on the monitor. Each of these steps takes time.

In technical terminology, the time span between image capture and display is referred to as end-to-end or sensor-to-screen latency. In the context of network-based video surveillance, Axis divides this into three main areas: processing within the camera, transmission over the network, and processing at the receiving end. Latency is therefore less a characteristic of individual devices than a feature of the entire video chain.

From the sensor to the screen

Delay occurs even within the camera itself. Exposure is followed by image processing and encoding; internal buffers may also be involved. Transmission then begins.
In IP-based systems, factors such as bit rate, available bandwidth, network infrastructure and transmission protocol influence how quickly the video data reaches its destination. Switches, routers and other network components also lie along the signal path. In a suitably sized local network, this component can remain minimal. If transmission paths become longer or the required bandwidth is not reliably available, network latency can become a more significant factor.

The signal path is not yet complete upon the arrival of the data. On the client side, streams must be received, cached if necessary, decoded and output via the graphics hardware. Larger receive buffers can stabilise smooth playback, but at the same time increase the time between capture and display.

For operators, this implies a key relationship: the performance of a camera alone says little about how up-to-date the image appearing before the operator is.

Real-time is achieved across the entire system

Cameras, networks, video management systems, clients and displays form a coherent processing chain. Accordingly, their responsiveness cannot be determined solely by resolution, frame rate or codec.

A high frame rate can reduce certain components of latency; high-performance networks can keep transmission times low; client hardware and software influence decoding and display. It is only this interplay that determines the actual sensor-to-screen time.
For operators, this is particularly relevant where the live image is directly integrated into an operational process: when verifying an alarm, in perimeter surveillance, in traffic and emergency control centres, or wherever decisions must be made at short notice based on the video image.
The frequently used term ‘real-time’ thus takes on a concrete technical dimension. What matters is not merely how fast an individual component operates, but the delay introduced by the entire system under real operating conditions.

Direct signal paths with SDI

Alongside IP-based architectures, direct digital video interfaces continue to be used in professional video applications. These include the Serial Digital Interface, or SDI for short. Originally widespread in the broadcast sector, the technology is also used in specialised surveillance applications.

AG Neovo offers monitors with SDI interfaces for this purpose. Depending on the model, the devices support SD-SDI, HD-SDI and 3G-SDI. The HX-2402, for example, processes 3G-SDI in accordance with SMPTE 424M and HD-SDI in accordance with SMPTE 292M, and supports 1080p at 50 or 60 frames per second via 3G-SDI. A BNC output allows the signal to be routed to other components.

The structural difference compared to a traditional IP video chain lies in the transmission path. A direct SDI connection does not require the video signal to be IP-packetised; furthermore, the decoding typically required for compressed IP video streams on the receiving end is not necessary in this section.
AG Neovo describes the SDI transmission of its surveillance monitors as uncompressed and low-latency. For 3G-SDI, the manufacturer specifies distances of up to 100 metres using RG6 coaxial cables; for HD-SDI, up to 150 metres; and for SD-SDI, up to 300 metres.
However, this does not allow for a general comparison of speeds between SDI and IP systems. The achievable end-to-end latency depends on the overall architecture in question. IP-based systems can also achieve low latency if properly configured. Rather, SDI offers an alternative architecture with a comparatively direct signal path between the source and the display.

When functions are integrated into the display

At the receiving end, the way in which information is consolidated in control centres is also changing. CCTV and NVR outputs are combined with situational awareness displays, analytics interfaces, process data and other operational applications.
Traditionally, these sources are distributed across multiple screens or organised via external switchers, matrix systems and multiviewers. However, with every additional component, the hardware and signal structure also becomes more complex.
Multi-input displays shift some of these functions directly into the display itself. With the PBP concept described by AG Neovo, multiple sources can be connected directly to a single display and shown in parallel. Depending on the device, picture-by-picture and picture-in-picture layouts are available; professional models can handle up to four input signals. [RE]

Infobox: Video latency at a glance

– End-to-end latency refers to the time between image capture and visible display in the control centre. It occurs throughout the entire video chain – from the camera and transmission to the client and display.
– For operators, therefore, it is not only the performance of individual components that counts, but the measured delay of the entire system.
– SDI transmits digital video signals directly and uncompressed. This eliminates the need for IP packetisation and the decoding typical of compressed IP streams at this stage of transmission. Whilst this can facilitate a shorter signal path, the actual end-to-end latency still depends on the overall system architecture.
– PBP/PIP and input failover can improve the availability of information at the workstation, but they are no substitute for a comprehensive redundancy strategy.
– Practical question: How much time is permissible between an event occurring and it being displayed?