Not too long ago, I was speaking at an International Security Conference on the subject of forensic video analysis.
Coincidentally, the vast majority of the more than 500 vendors with booths at the conference were also interested in video security systems, but from the perspective of selling them. Digital CCTV in particular seemed to be the hot new item to offer customers.
I spoke to representatives of many of the vendors, and requested written information from all of them. The big selling point they were making to prospective customers was the compression ratio. Although the idea of compressing the image from a security or surveillance system may sound good when you are only interested in bandwidth and storage, it may prove to be the element that makes your system useless to you should you ever need it as evidence. This may prove to be a serious problem for those users of digital video surveillance or security systems that intended the systems to be more than an expensive alternative to deterrent devices such as mock cameras or camera use signage.
If this sounds as if it is a condemnation of the technology, it is not. It is a condemnation of the fact that many buyers are not fully aware of the capabilities and limitations of the technology, due to either their own lack of research or the well meaning, but commission-oriented facts set forth by some vendor representatives.
Let me start at the beginning to explain this. Digital video is a definite boon to the entertainment industry, and that success has pushed it into other areas, such as security. For entertainment, the compression of an image file will reduce the amount of fine detail in the picture, but that makes little or no difference to the audience, who are there for the entertainment value.
Compression is the key element in this argument. Compression is the conversion of a digital format so that it takes less bandwidth for transmission and less storage space. When dealing with image files, compression is considered to be "lossy." Meaning the image that is compressed, transmitted and/or stored is only an approximation of the original image. The higher the compression ratio, the more loss of data there will be within the image. In this process, details of an image may be lost.
A single frame of analog video using the North American standard format (National Television Standards Committee, or NTSC) contains 349,920 pixels, or small picture elements. It also contains two separate images called fields. In this format, there are 29.97 frames per second, resulting in 10,487,102.4 pixels, and 59.94 images per second. This is an incredible amount of information.
In fact in order to convert a single color video signal into a digital format requires an encoding of 8 bits of brightness, and 24 bits of color for each individual pixel, or in other words, 335,587,276.8 bits per second. It is easy to see why digital video requires compression. However, the very best known "lossless" compression methods can save most of the important details at a compression ratio of 2:1. Many of the digital CCTV surveillance systems are using compression of 500:1 or greater.
One of the big selling points I noticed was that each vendor had a monitor where they were showing the live images captured by the camera. In most cases the images were very good, and very detailed. However, I also noticed that customers were not told that this was live feed from the cameras, and had not been through the compression process.
Some systems are using one of the MPEG standards to capture the image and then compress it further for transmission and storage. The MPEG compression uses four different techniques. Pre-processing is the first and is used to filter out "non-essential" visual information. The second is motion compensation, which identifies motion in the image and predicts it for the next frame, so that it does not have to transmit the actual frame information. The third is temporal prediction, which like motion compensation, sends only updates to approximate the actual information. Finally there is quantization coding, which rounds off all coefficient values of the algorithms that fall within a certain range and set them to the same value. As you can see, even the initial encoding will lose information, and then there will be the loss that must come from further compression for transmission.
One of the promising digital technologies is the wavelet, which operates as a Fourier transform. Most simply put, it captures both large and small information values and fills in between them. The most promising technology I encountered didn't use wavelets, but increased the number of pixels while decreasing their size and encoding. This method retains more of the information that would be available with an analog format. However the beta for this technology will not be released for several months.
Obviously digital video is headed in the right direction, but it isn't there yet. Most digital surveillance images are so compressed, they have the same pixilated appearance that is used on television to protect an individual's identity.
When you have to rely upon the images from a surveillance camera as evidence for you when you go to court, then the most minute details are vital. The basis for forensic video analysis or the process that is used to identify and isolate evidence on video tape or CCTV is in the details.
Forensic video analysis is both the science and the process of examining videotape or images on digital media. In this process, the examiner compares images of people, property and clothing that have been captured by the camera, with those that are known. This comparison requires the use of all available details, such as buttons or patches on a jacket, the unique small pattern variations that you see along the seams in a pair of jeans, the drape of material across an individual's shoulders and back, full or partial tattoos or skin imperfections, or even unique habits of movement. In many cases, there will also be audio material to examine. In most cases, the details will determine whether the person in court is the one whose image was captured by the camera, or just one of several people fitting a general description that may or may not be the one in the image. As you see, the entire case could fail if the video image does not have sufficient detail for a unique identification to be made.
For those that are considering trading in your old analog system and making the purchase of a digital CCTV, take time to determine your expectations of the system. If you expect the images captured, transmitted and stored by the system to be a vital element in your case against an intruder or individual who failed to uphold the law, you might want to keep the old analog system until digital video meets the requirements of security and business professionals, not just the needs of the entertainment industry. The money may be better spent on new videotapes.
Theresa Lang, CISSP, CCNA is a security and training consultant, who also teaches information systems protection at the George Washington University. She may be contacted on firstname.lastname@example.org.