Featured Post

Welcome to the Forensic Multimedia Analysis blog (formerly the Forensic Photoshop blog). With the latest developments in the analysis of m...

Thursday, February 14, 2008

A tale of two technology reviews

Like most of us, I get a ton of free trade magazines every month. This month, two articles from two separate mags really jumped out at me.

The first, "Why the French spot terrorists better than Americans" by Peter McKeelaw (Security Info Watch), made me want to throw the issue in the bin. The author starts with a confusion of the terms digital video and IP (no they are not the same, Mr. McKeelaw) and goes on from there. (Note to Mr. McKeelaw, it is possible to connect a legacy camera to an a/d converter and send the packets around a network via IP.) Here is one of my favourite quotes: "Not only are analog video surveillance images less clear than digital video, and therefore less useful, but analog cameras lack the functionality and intelligence of digital cameras. They see less detail, store less detail, and lack the ability to analyze or send alerts when a suspicious event is spotted."

Let's treat Mr. McKeelaw like an "opposing expert" and cross examine him and his testimony.

"Not only are analog video surveillance images less clear than digital video ... " Assumes facts not in evidence. Clarity is a function of many things; lense, camera body, transmission path, storage medium, and maintenance - to name a few. Mr. McKeelaw, is it true that video images, when stored digitally, can suffer from all of these same issues (bad lens choice, wrong camera body for the chosen view, lack of maintenance, and etc.) as video images stored on an analogue medium such as video tape? Yes, Mr. McKeelaw, it is true and we see it all the time. The problem gets compounded with the compression that is a function of the digital realm.

"... therefore less useful ..." Usefullness has many factors. A VHS tape can be placed in just about any VCR and it will play just fine. Some tapes are from multiplexed systems (multiple camera views on a single tape - time compressed) and require a hardware/software solution for proper viewing. But, most law enforcement agencies can handle multiplexed tapes - or know a neighbouring agency that can. Digital video, when saved as a proprietary file type with a proprietary codec, is often worse than useless. Proprietary file types cause severe delays in investigations as agencies are forced to research file extensions and manufacturers. Often times, the manufacturer is a foreign company with little or no documentation about their files. Proprietary codecs offer unique challenges as well. Compression = loss. Security companies determine, for their own reasons, what will be saved and what will be thrown away when designing their own codecs. When an 80x100 pixel face gets turned into 4 groups of beige pixels, I don't particularly find that useful.

" ... but analog cameras lack the functionality and intelligence of digital cameras ... " Camera bodies have many features; back light compensation, day/night, auto iris, and so forth. Many of these features do not depend on transmission method, meaning the video stream can be passed down an analogue or digital signal path. Functionality is software driven. Many of the intelligent/analytic functions often described as "digital" can be accomplished with an analogue signal coming down an analogue path - the signal gets digitized at the control centre. There are cameras on the market which house built in linux based servers and offer much control at the local level. These are rare at the moment and on the upper end of the price scale. More often than not, people confuse the camera with the control room when speaking of functionality.

" ... they see less detail ... " Mr. McKeelaw, how is it that analogue cameras "see less detail?" Is it possible that there are some analogue cameras that "see" more detail than digital cameras? Can you describe the difference between a "digital surveillance camera" and an "analogue surveillance camera" in terms of chips (CCD/CMOS) and capture mechanisms?

" ... lack the ability to analyze or send alerts when a suspicious event is spotted." Mr. McKeelaw, is it true that the analytics and alert functionality is facilitated by the control room's equipment and not the camera body itself - in the majority of cases.

I really hate to bang on Mr. McKeelaw so much, but I think I've made my point. Forensics is debate and oration. Mr. McKeelaw's oration, his article in Security Info Watch, attempts to present an "expert's" view of an emerging situation. Unfortunately for Mr. McKeelaw, Forensics is also debate. When subject to cross examination, Mr. McKeelaw's oration - his "expert opinion" - is shown for what it is. It's not an expert's opinion on a critical subject. It's an advertorial written by a marketing executive for a company with a financial stake in driving the market to a particular solution ... conveniently offered by his company.

Unfortunately, many of our supervisors read these magazines and get duped into thinking a certain way based on these advertorials. Digital = good. Analogue = bad. Digital = new/future. Analogue = old/past. How about looking at it this way, 1 full frame of uncompressed NTSC video = 1.3MB of data. 1 full frame of NTSC video compressed by the average DVR for sale in the US = .0035MB. Another question, why can't you frame average still frames of "digital video?" 

It is for these reasons that I've chosen to cross examine a portion of Mr. McKeelaw's statements today.

In contrast, take a look at Government Video's web site and an article by Wayne Cole entitled The impact of compact. Here's a quote, "Video resolution is related to the luminance bandwidth in the recorded signal. Smaller imagers capture less light and therefore less resolution. Lossy compression formats like DV, HDV, and XDCAM HD are based on throwing out high-frequency information that will be least noticed by the human eye. In other words, resolution is sacrificed in the hopes that viewers will not notice. HDV and XDCAM HD are worse yet because they use long-GOP MPEG-2 to add temporal compression to produce P and B-frames.
Many DV camcorders use 1/3-inch imaging chips followed by 10:1 DCT compression with 4:1:1 sampling (for NTSC). The reason this high compression rate "works" is that the smaller chips don't capture more resolution than the DV format can deliver. " 

WOW! But wait, there's more, "Like MPEG-2, there is a higher bit rate I-frame only profile for AVC that is not only more amenable to post-production, but can deliver higher resolution given 1/2-inch to 2/3-inch imaging chips. Despite being simpler to encode than AVCHD (there is no need to do the temporal compression to produce P-frames, B-frames, and long-GOPs), the only manufacturer to implement AVC I-frame only, Panasonic, is maintaining the cost of camcorders for this format in a range above the 4K capable Red One system (ed. note: Red One's body is $17,500)."

Now, one author is talking about surveillance systems and one is talking about camcorders ... but check out the language used? Check out the implications of what the latter is saying vs the former. "Lossy compression formats ... are based on throwing out high-frequency information that will be least noticed by the human eye." Compression = loss. What did we lose? Is it important? How will a person in a post production role (forensic image analyst) know what was lost and what was retained? How will a jury? If the only manufacturer in the camcorder market to impliment AVC I-frame sells cameras that cost over $17k, how much would a similarly featured surveillance package cost? Is there such a set up?

I am sure that there are some outstanding examples of digital video recorders and surveillance systems that employ digital from end to end. I've seen a few of them and have been impressed. I've seen a few, not many. The majority of the stuff I see falls under the DVR category at the local Costco. Notice the difference between "display resolution" and "recorded resolution" on these models. This is the type of stuff local store owners are buying to comply with the regs for selling alcohol or lotto, $4k for 16 cameras. 1TB of HDD and 1CIF images (352x240)? Wow! 

To conclude, always ask why. Who is Peter McKeelaw and why is he writing his advertorial? Who is Wayne Cole and why is he writing? Why did one choose Security Info Watch and the other Government Video? What is it about the two audiences that is different? Why is that important? As "experts," our clients and our employers look to us for sound and unbiased advice. Sometimes, agencies spend millions of dollars based on this advice. If you get it wrong, your job and your reputation is on the line. Be careful in your choice of source materials ... as this long post illustrates so vividly.

So, armed with your forensic skills, read this article and see what questions it brings to mind.

Enjoy.

1 comment:

Doktor Jon said...

Interesting points well made Jim,

The analogue / digital argument is manifestly daft, if we consider that the average installer 25 years ago, could produce more usable images with a 2/3" Vidicon B/W tubed camera, than many installers can manage today using a 540 line resolution Exview CCD model.

The important point about using video surveillance equipment effectively, is that it's as much about technique, as it is technology.

If IT departments believe they can operate security camera networks efficiently, I look forward to the day when 85% actually prove that they can.

At the moment, the industry is full of experts, many of which unfortunately don't really understand what they're talking about.

Such is life

Doktor Jon
CCTV Advisor