Over the last 10 years, the CCTV industry has seen a major shift from analog to digital networking solutions for surveillance applications. In the next few years, the industry will see the adoption of High-Definition (HD) IP Video into mainstream CCTV systems, as in consumer electronics. For HD IP cameras to be adopted by mainstream video security applications, the issues associated with high data rates and storage need to be solved. This is now being addressed with the recent launch of HD IP cameras with low data rates.
Advantages of HD IP
The advantages of HD IP cameras have been well documented, but there are three areas where they can provide real benefits:
General Surveillance — A single HD megapixel camera can replace several standard 4SIF cameras, thereby reducing costs. An HD megapixel camera can see more detail in the same field of view or view a wider field of view at the same level of detail.
Forensic Detail — Many existing analog CCTV systems simply do not provide enough resolution or quality for forensic evidence. Megapixel cameras solve many of these quality/resolution issues. They are ideal for applications where the system wants to identify and record faces, vehicle license plates or objects.
Digital PTZ — HD megapixel cameras can digitally zoom quicker and with greater detail than analog cameras, while still recording the whole picture for later analysis. This provides superior performance and is more reliable than mechanical PTZ mechanisms.
Surveillance can definitely benefit from HD technology. Currently, typical applications for HD IP cameras include retail point of sale, banks, casinos, car parks, building entrances, military installations and city center monitoring.
In order for HD megapixel cameras to be adopted for mainstream use, the following technology hurdles have to be overcome:
Lens — Megapixel cameras require a higher resolution lens than ordinary CCTV cameras to maximize the picture quality. These lenses are readily available but are expensive in CCTV mounts — although this will change with the higher volumes from mainstream adoption.
Sensor — Megapixel cameras use the same CMOS image sensors as used in still digital cameras, whereas analog cameras typically use CCD sensors. This is likely to change with the adoption of sensors from the HD TV/Video industry. A higher density of pixels on the same-sized sensor means there is less light falling on each pixel. Each pixel, therefore, has less sensitivity and needs more light — and the “noise” in the sensor has a larger impact because it is a higher percentage of the signal. This is why first-generation HD IP cameras typically had a poorer low-light characteristic than analog cameras; however, sensor technology is improving quickly.
Video Compression — Arguably the most important factor, due to its impact on network bandwidth and storage requirements. HD megapixel cameras are unlikely to be adopted for mainstream use until low-bandwidth camera designs are readily available. This is now starting to happen with the launch of HD IP cameras.
Video Compression: The Scale of the Problem
H.264 is the latest video codec (compressor and decompressor) standard, which follows on from the highly successful MPEG-2 and MPEG-4 video standards. The codec offers improvements in both video quality and compression. Many of the current 1- and 2-megapixel HD cameras use MPEG-4 compression, resulting in higher video data rates. For HD to become usable in mainstream CCTV applications, H.264 compression technology needs to be deployed in the camera, to provide the lowest possible data rates. However, not all implementations of the H.264 standard deliver the same quality of compression.
The huge disparity in camera performance makes a significant difference in the cost of an HD CCTV solution. Using cameras with data rates of less then 1 Mpbs means that HD IP cameras can use standard networks and storage and be cost-effective for everyday CCTV applications.
It is therefore very important for system designers and end-users to know exactly the data rates and storage requirements for particular HD IP cameras in order to fully understand performance and costs. However, some of the actual data rates are so high that it is not surprising that these figures are often hidden and difficult to determine. Take a look at a typical datasheet for a 1.3-megapixel camera from a mainstream manufacturer and you will see the camera has a frame rate of up to 30fps. However, nowhere is there a mention of how good the compression is — i.e. what the typical data rate is and how much storage is required to record a stream from that camera.
Some manufacturers are forced into using local storage because their HD IP camera bandwidth cannot be reasonably streamed live across the network. This somewhat negates the distributed and scalable benefits of IP Video. By removing the high-bandwidth problem, designers are free to choose a truly distributed architecture, placing storage wherever the best system design dictates it should be — whether that is in a central location, distributed close to the camera or a fault-tolerant redundant configuration mixing the two.
Designing a Low-Bandwidth HD IP Camera
The key to the adoption of HD CCTV into mainstream surveillance is the ability to develop low-bandwidth HD IP cameras, which consist of three main elements: a true IP camera solution; implementation of H.264; and dedicated hardware architecture.
True IP Cameras: A true IP camera completely eliminates any analog signal by connecting the digital signal processor — present in all analog cameras — directly to the compressor chip. This ensures no additional signal noise is introduced.
H.264 Compression: There are three common compression standards used in current HD IP Cameras, MJPEG, MPEG-4 and H.264.
Video is compressed using two types of frames: I Frame — also known as the Index or Key Frame and contains the whole image; and P Frame — which only contains the information that is different from the previous frame.
MJPEG only uses I Frames, whereas MPEG-4 and H.264 use a combination of both I and P Frames and consequently use considerably less bandwidth than MJPEG. H.264 will require up to 50 percent less bandwidth than MPEG-4 to transmit the same quality image, therefore it is the chosen compression standard for the highest performance IP cameras.
The H.264 standard specifies a set of optional tools which can be used to compress video. A compliant decoder must implement every tool, whereas a compliant encoder can choose which tools to use. This means that there can be a big difference between encoders from different suppliers — some compress well, some do not.
To determine what information is transmitted in a P Frame, the image has to be searched for motion in each frame. The quality of the compression depends on how far and how well the search is completed on each frame. The limitation to this searching is the available processing power in the camera, even more so with HD resolutions at full frame rate.
A poor encoder design could result in a higher bandwidth for good quality video; increased bandwidth during high motion; dropped frames; and blocky or blurry video.
Hardware Architecture: Due to the huge processing demands of a low-bandwidth HD IP camera using H.264, it is essential that the compression engine is implemented in dedicated hardware such as Field Programmable Gate Arrays (FPGA). With this type of design, low-bandwidth HD compression can be achieved with a guarantee of no dropped frames.
Oliver Vellacott founded IndigoVision in 1994. He was previously a product manager with a background in intelligent camera products. Oliver studied piano at the Guildhall School of Music before gaining his first degree in Software Engineering from Imperial College London and then a Ph.D. in Electrical Engineering from Edinburgh University.