Simplifying the search for digital recorders
Nov 1, 1999 12:00 PM
There is a burgeoning number of digital video recorders on the market. The bold claims, the hype, the specsmanship and the double-speak can cause confusion and make the purchasing decision a real challenge. This article will attempt to clarify some of the common misconceptions and to answer more frequently asked questions. But first, let's emphasize one aspect of overriding importance.
Compared to VCRs, digital systems offer a wealth of features and functionality and not always at a higher cost - in fact, most commonly with a major cost saving. The absolute, most critical thing you must do before choosing a system is to understand how the security system will be used. What does each camera do? What does it look for? What is needed to achieve it? So many people insist on 30 images per second as non-negotiable until they hear that 15 ips looks very similar for half the price. Maybe some cameras could just record when there is a trigger (pre-and post-event, of course) thereby reducing the cost dramatically. On-the-fly reprogramming (e.g. switch from 2 ips to 15 ips on activity or in daylight hours) makes the system design crucial if you don't want to throw your cash away.
Here are some common misconceptions.
* Everything is SVHS quality. Common usage is diverging rapidly from strict definitions of resolution and, needless to say, each manufacturer chooses its own way to measure resolution The analog measure of TV lines (i.e. how many black and white line-pairs on the screen) has a long history but beware: Some companies quote each line pair as 2 lines. As a basis for comparison 400 TV line-pairs of resolution approximates SVHS quality.
The digital market works in pixels and lines: 640 by 240 (approx SVHS) at the top, 320 by 240 (nearly VHS) typical and 160 by 120 for fast lower quality (primarily used in video-conferencing). Digital companies, quite legitimately, replicate lines or pixels to fill the screen, but then, questionably, try to claim a higher resolution (e.g., 640 by 480 when in truth it is 320 by 240). The key measure is to ask: When the image is digitized, how many pixels are there per line? No amount of subsequent replication, interpolation, convolution or any other math technique can restore the information you have lost at this stage.
* Frames are better than fields Very few companies capture frames and for good reason. Frames are made up by interlacing two fields to make a moving image more pleasing to the eye. This works well with movies. However, CCTV security is not a movie and when you need to examine an incident it is normally done by studying each image individually. Frames cause fast-moving objects to appear blurred when held frozen. Fields are sharper, and 30 images per second are more than enough for almost every application.
* Digital images are all equal and unproven in law. Well, they are all unproven in U.S. law, but certainly not equal. The key criteria seem to be:
a) It must be a good quality image;
b) It must have a good degree of protection against tampering;
c) Any losses or modifications must be explainable and reasonable.
Both JPEG and Wavelets apply some fairly simple math to each individual image. So, providing it is not over-compressed and the image is encrypted or watermarked, most experts agree that there is a good chance of the courts accepting it.
MPEG or H263 compression is much more doubtful. MPEG is a clever algorithm for compressing moving images; but security images are not movies. MPEG was not designed to stop the picture and step through it frame by frame. It works by taking a JPEG reference frame then looks for changes in subsequent frames; however, it makes estimations and approximations compounding error upon error in successive frames and then further adjusts the image to make sure it doesn't slip too far away from the next reference frame. In a movie you would rarely see this degradation, but if your legal case depended on the integrity of a specific frame, I'd say you are taking a big gamble.
I know many expert witnesses who could defend JPEG and Wavelet techniques, but defending MPEG (and H263) against a switched-on lawyer is not a challenge I would entertain.
* JPEG is better than Wavelets. At the moment JPEG probably is better because it is an ISO standard and Wavelet is not, but from a pure technical standpoint, I think Wavelet has a superior algorithm. There are numerous implementations of Wavelet, some good and some bad. In December the JPEG committee will publish a schedule for incorporating Wavelets into the ISO standard before the end of 2000.
At low compression JPEG is probably better (but it is subjective), whereas at high a compression Wavelet is definitely better. However, we don't know how high compression the courts will allow and which implementation of Wavelets will be the standard. In the meantime, just take a good hard look at the image quality.
* Split screen images are better Quad, 9 way and 16 way images are extremely useful to the security guard. Just remember that all images can look good when they are small. Digital systems should offer some form of split screen, but ensure you see a recorded image full screen size before buying.
Now Let's Answer Some Frequently Asked Questions * What are Algorithms? They are math techniques for manipulating data; e.g., compression, motion detection, zooming, etc.
* How big is the disk? This question is a little like attending an automobile show and asking how big is the gas tank without knowing the engine size, whether it's a car or a truck, or if it's propane gas or diesel. There is an answer, maybe lots of answers - but what would they get you? You have to specify the problem first.
* Can it zoom? If the stored image is bigger than the display (e.g., high detail maps) then zoom is very useful, but this is seldom the case in CCTV. Zooming adds no further information, it merely replicates pixels or interpolates or uses any other arbitrary math technique. In most cases, it is a gimmick and serves no purpose.
Given some known criteria, like a car licence plate, an imaging expert can enhance the contrast, sharpen the edges and zoom the image to read the plate; but these techniques don't generally work with faces and natural features. Best to leave it to the experts, and forget the gimmicks.
* How much is a basic recorder? If it's a multi-camera system, then you didn't read the opening paragraph.
* Can it record and replay simultaneously? Most digital recorders can, but do you need it? If so, make sure it's easy to operate.
* Does it have encryption? A good question. JPEG and Wavelets themselves encode the image, but not securely. Each image should have a password that locks an audit trail into the image and detects tampering. Many standards have been proposed, but none is generally accepted. None has yet been tested by the courts. It's a complex area where a consultant might help you ask the right questions, but even he probably couldn't guarantee the security. Nevertheless, you must ask the question.
* Can I export a clip? Digital images don't have an "original" (the camera is copied to the computer memory and overwritten, the memory to the disk, the disk to the tape, etc). Unlike analog copies (including photocopies) digital copies are 100% identical, so the legal requirement for "best evidence" is easily fulfilled. Exporting an image for use as evidence is important, but it is essential that its authentication and audit trail always go with it and do so securely.