Dallas Airport: The Massive Cost of JPEG2000

By Carlton Purvis, Published May 01, 2013, 12:00am EDT (Info+)

The Love Field Modernization Program (LFMP) is an ongoing $519 million dollar project to upgrade Dallas Love Field Airport that has received significant mass media attention. Renovations include security upgrades that LFMP hopes will provide capability to screen passengers more thoroughly with less intrusive techniques. To do that the airport said it wanted a surveillance system that could read license plates from afar and zoom in to polo shirt logos [link no longer available]. In this note, we review the airport’s $3 million surveillance contract, an example how the cost of a contract can increase substantially when using JPEG2000.

Project Overview

**** ******** ***** ** *** ** **** **** contract, ********* ********* *****/***** ************ *********************** *********, ******** $*.** ******* ***. The ******* ******** ***+ ** ******* of ***** ** ***** **** *** the **** ***** ******* * *** 5MP. *** **** ******* *** *** the ******** ******** *****************.

[******: *** *** *** ********************* ****** **** ************ ** ** March **, ****.]

** ******** ** *** *******, ***************** **** **** ** ******* **** were ********* *** $***,***.** ** **** as 17 ******************/ **** ******* ****** ** **** each *** $*.** *******. *** ******** were ** ********** ~$**,*** *** ******** to ********* **** *** ** ****** control ****** **** ******* ~$**,***.

Storage **** ********These metrics show how large the server / storage component and cost were:

  • ~$*,*** *** ***** *** ****** **** for ******/*******
  • ~*** *** ******* *** **** *** camera
  • *** ***** ** *******, ~***** ** storage *** ****
  • **% ** *** ***** ************ ********* cost **** ** *******/*******
  • ~** ** ******* *** ****** (~*** IP ******* *** ** *******)
  • *,***+ *** **** **** *** *** ~500 ******* (**** * ** ******* + ***** * ** ******* ******)

**** ******* ***** **** **** *** the *** ** ********. **** *.*** *******, either********'***** ****** **(*** ***) ******* (***** ******** *** been *******) ** ***** ***** *******, we ****** *** ***** ** *** project ** **** ** $* ******* or ****. **** ***** **********'********* ******* ****** * ******* ********* ******** ** 10x **** ******* **********/*** *.*** *******. ** *** other ****, *** *** *******, **** the (**) **** *******, ***** ******* JPEG2000.

*** ********, ** * *****, *** 'standardized' ** *.***, *** **** ******** shows *** **** **** ** ******** less *** **** ******************* ** ******** ** *****.

** ********* ******* ********* *** ********** information ***** *** *** **** ***** JPEG2000, *** **** ******** ******, “*** the ****** *** ******** ** *** airport, ********* *** *********, ** ** not ******* ** ******** ******** ***** utilized ** ****** **** *****." 

****: *********** ****** ** ****** ******** to **** *** ****** ** ****** and ********.

** ******* *** ************* *******, *** ***** *** ****** were *** ********.

Comments (60)

This is a very one dimensional way of judging the cost of an Jpeg2000 installation.
Does the statement: With H.264 cameras, either Avigilon's more recent HD (non PRO) cameras (where JPEG2000 has been dropped) or third party cameras, we expect the price of the project to drop by $1 million or more” take into account the extra cost:
- for additional H.264 camera and installation to replace the high megapixel cameras. The market does not provide H.264 cameras with 35 mm chip.
- The extra cost of high performance pc clients. As Jpeg2000 is a scalable codec it has huge advantages when searching in footage. If the end-user requires smooth and quick ways to search footage, it requires extreme pc power on the client side to keep up with a scalable codec search.

IPVM could do better than this populist punch lines.

Thanks for the feedback, Michael. You could easily buy (3) 5MP cameras for far less than the price of a single 16MP Avigilon camera plus get the benefits of a more bitrate efficient CODEC, better angle the individual cameras for optimal FoVs / line of sight and higher frame rate (12fps vs 3fps).

As Avigilon itself acknowledges, the alternative to scalable CODECs is multi-streaming, which I know you aware of is how Avigilon itself has redefined what it means for HDSM as applied to its 2nd and 3rd generation non JPEG2000 cameras.

What do you mean regarding JPEG2000 is a scalable codec. Isn't H.264 also scalable?

We may not be as mature as an airport for video surveillance, however we look at each camera, and will pick codecs that fit the need.

In my experience, H.264 doesn't require "extreme pc power". My 5 year old laptop can handle at least one H.264 stream at full 1080p. A modern middle of the line desktop that we purchased, was tested for 16 1080p streams on the display at one time. And, with 64 streams, the system could still show an acceptable frame rate for monitoring activity.

H.264 is not a scalable codec, at least not the version that essentially everyone uses. Here is our tutorial section on scalable codecs.

Thanks. Somehow, I missed that...

Given that today's PC hardware a) can transcode very fast (6x faster than real-time for a mid-high end PC), and b) gets faster, is not having to transcode that important these days, and in the future? Especially compared to the cost and heat of the disk storage required for JPEG2000?

Michael, most VMSes do not transcode, at least not 'out of the box'. The typical approach is to record multiple streams, say one at 1080p and another at CIF or 4CIF. When the camera is displayed is a 3x3 or 4x4 matrix, the low resolution stream is displayed (to save computing resources when the size of the screen used is small) and then switches to high resolution stream when the camera is displayed by itself.

in side by side video comparisons of h.264 vs jpeg2000 which will provide better overall image clarity...

At the same bit rate, H.264 will deliver far better image quality than JPEG200, it's not even close. Even at dramatically higher bit rates, H.264 will deliver equal image quality to JPEG2000. There's some argument that if you massively increase the bit rate for JPEG2000 vs H.264, the image quality may be better for JPEG2000. On the other hand, you could simply decrease the compression level on H.264 for a similar effect (see our compression/quantization tutorial).

Readers may be interested in our test of H.264 vs MJPEG (JPEG2000's sibling). In any fair test, a modern H.264 camera (like Avigilon H3, Axis, Sony, etc.) will match the image quality of any JPEG2000 camera at the same pixel count / fps but with far lower bandwidth consumption.

i look forward to reading those two articles you posted... one thing that may have been a deciding factor in jpeg2000 was that h.264 had not been fully offered by avigilon at that point in time or at least prior to an rfp or bid opening going out...

as always i appreciate the knowledge offered...

Keefe, Avigilon's H.264 cameras (at least 1MP and 2MP) were certainly available in 2011. Indeed, we tested them before the SOW was submitted. Correction: Avigilon's first H.264 cameras were announced ~2 months before the SOW was submitted (September 2010). Certainly, they were known at that time and, by the time the deployment started (months later), they were clearly available and could have resulted in substantial storage reductions.

Certainly, there may be good reasons the airport choose these specific Avigilon models. It would be interesting to know given the massive storage cost involved. Unfortunately, I doubt we will ever get their perspective. The airport replied to our request for comment saying, "For the safety and security of airport personnel and travelers, the airport does not comment on security measure being utilized for Dallas Love Field."

yes they were available pre 2011 but the proposal that you have listed from R&D is from (november) 2010 not 2011... as with any project especially one that size they began planning months/years before.... my thought is that specs may have been written only when a jpeg2000 camera was available commercially and avigilon was the specified product/manufacturer thus leaving only the choice of jpeg2000 model cameras... it could have been to the point where it was so far along they didn't want to muddy the waters by changing the specs as well... who knows...

it is kind of like buying a car and then a couple months later finding out the manufacturer changed the body style on the new model...

Keefe, yes, I updated with that correction above. I agree, as a logistical matter, it's tricky to switch within the process. On the other hand, CODEC choice has a big impact on total cost. They clearly did not take delivery of the products until later in 2011 at the earliest.

Equally interesting, what was the cost / feature tradeoffs of using that vs. the many H.264 IP cameras available then?

Beyond that, the bigger issue relevant today is the use of MJPEG or JPEG2000 only cameras in projects today. There are still a small minority who advocate that. I think they need to address the full cost differential and the hidden increase in storage costs.

i agree that it did drive the storage costs up on the project and there were (are) other h.264 options available at that time so my only though (without having read the spec) was that avigilon was the specified manufacturer...

i have seen compliance letters for surveillance resolutions that specified only jpeg2000 (avigilon) cameras as well back in 2011 as well... now that you raise the question I am curious if two years later if they are still specifying jpeg2000 or if they have changed it to include h.264...

the storage costs between the two are night and day and typically price will follow that as you have mentioned...


Sorry, I have to disagree with this article. Your past dealings with Avigilon make it appear as if you are sniping with little valid reason. I'm sure there are many systems installed in recent years that, for one reason or another, use MJPEG for some or all of their cameras.

I don't think it is really a big deal which codec LFMP chose. It appears a bit disengenuous for you to intimate that LFMP chose poorly because in your opinion h.264 is better suited for their needs.

Carl, thanks for the feedback. A few points:

  • I estimate less than 5% of new surveillance systems installed in 2011 used MJPEG or JPEG2000. So while there are 'many systems' that do, they are overwhelmingly the minority.
  • H.264 is not simply 'better suited for their needs'. This is not an issue of whether a minidome is better than a box form factor. H.264 is simply better, across the board, for storage than MJPEG / JPEG2000. And the revealed choice of people in 2011 and today confirm that. Storing video in MJPEG or JPEG2000 is a real disservice to any customer.
  • It is a big deal which codec one chooses because as the contract shows the amount spent storage, both in total and in percentage terms was massively abnormal. Choosing a different codec would easily result in hundreds of thousands of savings, if not a million. That's not chump change even to an airport.

The airport certainly may have its reasons, but the contract demonstrates the financial penalty of using JPEG2000.

Well stated Carl. I think many recent posts on this site definitely appear to be disingenuous and arrogant. There are many reasons end users choose specific compression types including legal (non interframe compression, no interpolated frames, etc), quality (higher quality throughout entire frame in every frame when examining finer details), consistency in compression and many more other than simply bitrate. I think having this article brings no value to the site considering there are already great articles about compression types and the advantages and disadvantages of each on this site. Seems IPVM is spending a lot of time with biased speculation on things with little to no value to US the customers and a disproportionate amount of it toward one company recently. If you want to provide good info simply put a storage calculator together with your compression article and that would be infinitely more helpful than this post to IPVM users. If you would rather assert your own admittedly ignorant opinion with regards to full details, requirements and intent of various projects then maybe your priorities are shifting from non biased reviews and reporting... for this site that would be completely unacceptable.

The value of this article is that it shows the actual cost of a real installation using a legacy CODEC (like JPEG2000 or MJPEG). Rather than do hypothetical estimates, we have obtained an executed contract that demonstrates the true price paid.

I appreciate that you noted potential reasons to pay far more for a surveillance system. I do see, on occasion, legal requirements, though, from a technical perspective, that's more fear than grounded in real risk. As for higher quality, I challenge anyone to prove that JPEG2000 delivers higher quality than a professional H.264 camera (like the Avigilon H3 or similar models from Axis, Sony, etc.). If there are many other benefits, feel free to share.

The point does remain, using JPEG2000 (or MJPEG) clearly increases cost massively. Feel free to cite those potential benefits and how they outweight the substantial cost increase.

what i find interesting in the sow is r&d's equipment cost that they list vs. the sale price there is virtually no markup... those prices don't come close to msrp... looks like what the airport spent in storage they saved in cameras and equipment...

Keefe, interesting observation. Here's a screencap excerpt for others to review:

It appears that each item is being marked up 5% (i.e., the second to last column multiplied by 1.05% returns the last column). One of the header's say's "R&D COST" which we are both assuming is the price they buy from Avigilon.

The price does seem below MSRP, though that is commonplace for any manufacturer when a job is this big. Volume discounts, etc.

Just wanted to point out one thing: "You could easily buy (3) 5MP cameras for far less than the price of a single 16MP Avigilon camera plus get the benefits of..."

While this may be true, now you're talking about 144 cameras instead of the original 48 16MP units, or 96 additional cameras on top of the ~500 already spec'd. You may save on storage (or maybe not, depending on the relative stream configurations), but now you're also looking at needing 96 additional VMS channels, 96 additional cable runs, 96 additional switch ports (with requisite additional PoE capacity... or 96 more power supply channels, depending on the design).

It may not seem like much beside 500+ cameras, but it IS a 20% increase in overall infrastructure to at least partially offset any savings in storage.

Besides all that, do we know that there wasn't some additional reason that played into the choice of these cameras is specific areas? I mean, sure you can reason that each 16MP camera could be replaced with three 5MP cameras... but by that logic, each could also be replaced with eight 2MP/1080p cameras, providing even more options on angles... or even 48 analog cameras recording at D1.

To be a bit tongue-in-cheek about it, it's kind of a reverse version of the standard MP marketing line that IPVM has often called Avigilon on: "One of our XX megapixel cameras can replace YYYY analog cameras!"

Point is, you can't really state categorically that three separate cameras would be a SUITABLE replacement for one 16MP camera, without knowing exactly what that camera's intent is - I skimmed the contract and didn't see any drawings specifying what each camera was intended to cover.

Not that the airport is necessarily looking to do exactly THIS, but just by way of example: there's no way in heck this rig could be replaced by... well, by just about any combination of other cameras:

Anyway... just wanted to throw that out there.

Matt, thanks for the detailed feedback. One thing is clear - you would absolutely save on storage, simply because the higher resolution camera (16MP) is JPEG2000 only and the lower resolution ones (5MP) are H.264. There's no reasonable 'stream configuration' that would make (3) 5MP H.264 cameras require more storage than a single 16MP camera.

Beyond that, going by the price on the spec, the 16MP cameras are ~$6,000 + $1,000-$2,000 for lens. By contrast, a 5MP H.264 camera is $1,500 or less (with lens). That's like $3,000 savings per 16MP camera to apply to the cabling, ports, and VMS licensing, etc. Plus, with lower bandwidth requirement, you can reduce the number of servers needed and the amount of storage.

Btw, while your example of a 6.2 km setup is interesting, it's not relevant here because the contract clearly discloses the lenses and length used for the 16MP cameras:

Btw, while your example of a 6.2 km setup is interesting, it's not relevant here because the contract clearly discloses the lenses and length used for the 16MP cameras

I did add the disclaimer, that it's unlikely the airport is doing anything like this... the point remains, without knowing why that specific configuration was chosen, and exactly where those cameras are aiming, it's impossible to simply say the cheaper option would have been "better", or a "suitable" option.

One thing I can say with some certainty: those lenses are going to kick ass over just about anything you'll find on the CCTV market when it comes to sharpness, clarity, and just all-around optical quality. That's the kind of glass most photographers drool over.

Edit: curious about this statement, though: "you would absolutely save on storage, simply because the higher resolution camera (16MP) is JPEG2000 only and the lower resolution ones (5MP) are H.264. There's no reasonable 'stream configuration' that would make (3) 5MP H.264 cameras require more storage than a single 16MP camera."

Aren't the 16MP cameras limited to something like 4fps? If you figure H.264 gives about a 10x space savings over JPEG2000, and adding the three up gives you ABOUT the same pixel count, you're looking at being able to do a combined 40fps with those three cameras, which should be quite easy to match or exceed - figure even if each one maxes out at 15fps.

But that aside, and accepting that you're probably going to see a fair if not substantial savings in storage space... there are probably still valid reasons that the single big camera is preferable to a bunch of smaller ones. And as others have pointed out, there are some (though few) instances where JPEG2000 provides benefits of its own.

Matt, as a professional, based on what we do know, it is not hard to imagine realistic scenarios where the other option would be cheaper and better in many ways. While it is not guaranteed either way, it has far more than enough evidence to be worth seriously considering.

I am not sure how to quantify your comment about photographers 'drooling over' lenses. We'd be happy to buy those lenses and do a shootout against traditional surveillance ones to measure the difference. If you think that's worth doing, let's start another discussion to address that.

Btw, Matt, two other known facts disclosed in the contract help identify where these cameras are likely being used. The contract says these 16MP cameras are color only and use f/2.8 lenses (as noted in capture above). Avigilon, in its own documentation, notes that f/2.8 lenses are not recommended in low light. Plus, as the lens spec indicates, the FoV is fairly wide (14mm on a 16MP camera is 104 degrees, etc.). These facts point to the camera's use indoors (or a heavily lit wide outdoor area). Either of which should work well, if not better, with multiple lower resolution cameras.

Yes, ultimately, unless Dallas explains all the details to us, we can not be certain but we can carefully analyze the significant documentation before us to develop reasonable estimates and theories.

Therein lies the arrogance. Dallas does not need to explain anything to you. You need to accept the fact that sometimes others make valid choices that do not align with your view. Also if you want to provide value to US, your customers, then as stated previously you need to go back to your unbiased approach. I am honestly getting sick of your ranting and publishing of articles disproportionatley stirring up critism of a single company recently... It really seems like you are on some personal vendetta here. If you want to report on 2 year old projects then get all the details first. If you are ignorant to the details then don't try to pass your speculation on as beneficial to your customers. Look at the volume of recent articles you have published and recent time you have spent - it's obvious, ridiculous and unprofessional. If you can't see it then we are in for a whirlwind of disguised tirades and less value from IPVM. Again - if you want to help your customers then storage calc along with compression article and maybe a cost estimate tool - an unbiased approach with valuable info and tools with great, beneficial, real world application.

Let's look at the recent volume of articles and perhaps realize that Carlton is doing a series of investigations into RFPs and contracts. For instance:

He is also working on reports covering a Tyco city contract, an OKC RFP, the SFO airport, etc.

You need to accept the fact that we raise questions and/or criticize numerous manufacturers - sometimes it will be your business partner, statistically most times it will not.

If you have any analysis or insights on the contract provided or the technologies used therein, please feel free to share.


I think we all understand your points about codec efficiency and the associated storage requirements but the question remains: would you have made such a big deal about this if the airport had chosen, say, Genetec VMS and used the MJPEG streams from whatever cameras they specified?

You must admit that they could have valid reasons for choosing a non-interframe codec, whether that was JPEG2000, MJPEG or even h.264 with a GOP size of "1". Without knowing their reasoning, it appears you are just using this as an excuse to take a poke at Avigilon.

Now I'm no Avigilon fan. As you know, they failed our Phase 1 testing miserably. But that failure was based on concrete evidence such as their encoders' ridiculous latency (>500ms), failure to properly control Pelco Spectra PTZs and operational shortcomings of the GUI, not any bias against the company.

Have you ever noticed pixelation changes in the background of H.264 video. Specifically when things tend to be similar color?

H.264 is actually very complex with tons of settings. You can raise bit rates, which will eliminate some posterization but not all but if you raise it then you are not getting that much storage savings. With H.264, you get less tones. Banding of continuous tone images is very common in H.264.

Not saying H.264 is bad, but it has certain limitations. In fact, it should be used most of the time. Heck, I shrink by Blueray movies down with H.264 before they go into my NAS so I save on storage and cost.

But if I was a major airport, I would probably want a few cameras that were not H.264. Take the picture above, from the new airport, and you will see many of the same tones. What would it look like if someone blended into that background with similar tones? A good chance it would be blocky. In this day of age...how to you take a chance?

What I don't understand is why designers, in this day of age, use similar tones. Designers should be working with security companies to come up with good designs that will help surveillance. A good design would dramatically help H.264 cameras.

Million $ idea?

  • Come up with a carpet pattern that will prevent H.264 blocking.
  • Come up with a paint pattern that will prevent H.264 blocking.


We'd be happy to report about a Genetec system that used MJPEG only, just as we had no qualms criticizing their partner video that ambulance chased the Newtown shootings or their weak criticisms of Milestone.

Any system in the last 5 years that standardizes on MJPEG or JPEG2000 is highly questionable and worth debating, especially given the contract's clear demonstration of the massive storage cost it required.

Finally, as to not 'knowing their reasoning', we regularly give manufacturers and users an opportunity to respond. That's their choice. We are not going to stop reporting on topics simply because those involved will not comment. If we took your implied approach, we would become a press release service / trade magazine. Even absent official comment, we have our own expertise, extensive test results and obtained documentation that allows us to analyze and share with the community. We will continue this series covering various projects and numerous manufacturers.

John - I am stating a fact that over the past few weeks there is clearly a disproportionate amount of time and material from IPVM associated with a single mfg. It really does appear as though you guys are taking on the very attitude you criticize and most of your users despise by your recent posted material... and that is my concern. I enjoy reading your reviews and articles and thinking about the questions and projects you present and I look forward to the new material. I just can't shake the feeling that your recent disproportionate focus is because you were offended and it really concerns me that this may be your approach moving forward with any mfg. It is an unnacceptable approach and tarnishes the image of what I thought this site was about. Whether you recongize it or like it that is my observation and concern as your customer.

Your concern is noted as I've read and responded to your numerous comments over the last few weeks. That said, it is a fact that we have had just 2 posts in our last 50 on Avigilon (see our chronological list of articles). Perhaps you can review our other 48 posts on other topics.

I can't "shake the feeling" that you are upset because this is your business partner. I don't see you objecting to the same level of coverage / criticism over the same time period against BRS Labs? (i.e., BRS Labs Ambulance Chases Mass Stabbing, BRS Labs $25 Million Investment investigated)

Ultimately, look at the facts of this post. The 'criticism' is on JPEG2000, which does have real competitive weaknesses. Indeed, as you know, Avigilon discontinued JPEG2000 HD cameras (i.e., the models of ~470 cameras deployed at the airport) on March 28th. So, put aside your biases and your perception of my biases, and let's look at the facts of the matter of JPEG2000 and its cost vs its value.

John - it's timing. As you know I've commented on many other topics and we are a proponent H.264 and thoroughly understand various compressions including JPEG2000. I wanted you to know how it appears as least to myself (and somewhat even to Carl) based on the timing and recent comments. That's all. I do not want this to become a site that posts articles for any other reason than the benefit of it's users. In dealing with you I am sure you understand this - thanks John!

The 16MP and the pro lens will be very good for FR and VA application.


You said "I shrink by (sic) Blueray movies down with H.264 before they go into my NAS so I save on storage and cost." I hope you realize that if you are indeed shrinking the bitrate of your Blu Ray rips, it isn't because h.264 is a more efficient codec (Blu Ray uses either h.264 AVC or VC-1 codec) but because you are essentially capping the video stream at a lower bitrate, and thus at a lower video quality. You may not notice much difference, especially on smaller screens and portable devices but it will be there in the form of artifacts on high motion scenes, etc.

Ssounds good and have a nice weekend. Just so you know (and can emotionally prepare) we are going to do a post on Avigilon's Q1 financial results next Thursday (May 9th). We cover every one of them, including the most recent one for Q4 2012 / year end (you did comment on that but surprisingly did not call my ethics into question).

@ITECW Library, I do not think a 16MP plus Pro lens would be a great choice for facial recognition nor video analytics (nor, for that matter, does the contract indicate that either is being used).

For video analytics (and given this is an airport, typically one is talking about perimeter violation), the two big issues are (1) eliminating common causes of nuisance alerts (like shadows, headlights, blown debris, etc.) and (2) low light performance. To that end, thermal is a better bet. Fine details of a person are not important for detecting people at a perimeter using video analytics). Thermal eliminates many/most nuisance alert causes, requires no lighting and can 'see' long distances.

For facial recognition, higher resolution could be interesting but the low frame rate (3fps) of the 16MP is a problem. Getting a good face shot is critical and, key to that, is having enough different shots to find the one that is most head on. People naturally tilt their head slightly up and down, left and right as they walk. At 3fps, you are much more likely to miss the split second where the head is best positioned. Now, one could argue that the wider field of views higher resolution affords allows the person to stay in frame longer. However, wider area facial recognition is problematic. Even if you have the pixel density, if people are even modestly askew to the camera, accuracy will drop significantly. This is why facial recognition is typically done at choke points and why so many vendors tout 2d to 3d conversion as the holy grail of the application.


I look forward to your coverage of the Q1 financial results. :-)

In this entire discussion I've noticed that not one person mentioned what I believe the single dumbest mistake Dallas made by selecting the Avigilon JPG2000 cameras (And likely the main reason Avigilon pushed them towards those cameras)... Software.

Their h.264 cameras are standards compliant and as such can be used with most major software. The airport could switch to Milestone or Genetec or even some other solution yet unknown. Since Avigilon refuses to ever share the SDK for their jpg2k cams, Dallas Love field is forever stuck using Avigilon software or ripping out all those cameras and replacing them with cameras of open architecture. New cameras, even if they choose h.264 cams of a different mfr, they're still stuck buying Avigilon storage servers and Avigilon software licenses.

You can argue the quality of their cameras all day but choosing hardware that locks your organization into having zero future software options makes as much sense as platforming an organization on Lotus Notes to me.


You are right in a way. I should have said I reencode. The H.264 Profile that comes on a Blu-Ray is High Profile (HiP). When I reecode, I will set my application up to look at the Blu-Ray frame, make calculations on that frame and allows it to encode that picture in the most efficient way. Many times for action movies, the bit rate is higher but for animations, the bit rate is a lot lower. Anyway, if the average person watched my videos, they would not notice as the video is still in Hi-Def. I just happen to notice in dark scenes. Some times it is blocking from h.264 or it is from bit rate.

This fall, turn on a local NFL game on your TV. Go to the Hi-def cable channel. Picture looks good but have you ever looked at the Football field and the field looks kinds of funny? Now change your TV to anntenea and watch the Hi-def local feed. You will see a lot better picture with little blocking or postorization.

With all the Profiles, Levels, and etc.. with H.264, there is a lot you can do. I am surprised that we do not have secondary reencoding for archived video past 3 days old, 10 days old or 30 days old. You will save major money on storage or would be able to store for longer.

Just by a carpet change, floor color change, and wall color change...you could truly record awesome HD video at lower bit rate an diiferent h.264 settings and save on major storage. You could also reencode that video later for better storage savings.


I get it. It mirrors the argument for lossless versus lossy audio codecs. Audiophiles prefer ALAC (Apple Lossless Audio Codec) or FLAC (Free Lossless Audio Codec) versus MP3. My experiments with MP3 on a relatively high-end system have shown me there is little audible difference between the original and the compressed music once you set MP3 to at least 256k. I tend to use 320k to allow a bit of extra headroom and because storage and download bandwidth are far less of a concern than they once were.

I haven't transcoded HD video material for storage (technically illegal if you rip protected material) but I set my Hauppauge WinPVR to its maximum bitrate (13.5Mbps VBR) for HD capture. Of course, it is incapable of capturing 1080p so that lowers bitrate requirements by at least 50%.

Carl, the audio analogy is actually a good example of a problem with lossy codecs: it's not just a matter of what information may go missing, but of what may be added. Great example was the Styx song "Renegade" - in the intro is just a vocal, and a kick-drum beat. I used to have a mix CD that was made from MP3s (so the music was crushed down to MP3, then "re-expanded" to CD format), and every time that kick hit with nothing else around, there was a high-pitched metallic grating sound on top of it.

With video too, artifacts generated in decompression could appear as false objects, or features on objects.

You were probably using 128k or less. I have a number of songs that were converted to, or downloaded at, 128k and the sound reminds me of analog AM radio. I rarely listen to them. But at 256k or higher, MP3 has excellent quality if the source was clean.

Alas, many modern recordings are so over-compressed to make them sound louder on portable devices that there isn't much you can do to make them sound worse. Even some older recordings that were well-mastered have suffered at the hands of over-enthusiastic remastering.

Yep... good example of the trade-offs between compression level and quality.

The thing that still gets me about this whole "debate" is that it's all about JPEG2000 vs. H.264, with a nod to being locked into the Avigilon platform... but most keep avoiding (or tossing aside) the fact that there simply are no other cameras on the market like the Avigilon Pro cameras, and that they may actually have been chosen for their specific strengths, with JPEG2000 and the Avigilon platform being accepted as a necessary trade-off to utilizing those strengths.

If the integrator really just wanted to lock the client into the Avigilon platform, he could have done it with a half-dozen or so Pro cameras; it wouldn't take four dozen of them. If he wanted to just spend stupid amounts of money when something lesser would suffice, he could have spec'd all 500+ being the Pro cameras. Since neither is the case, and assuming this integrator actually knows what he's doing, is it not a reasonable assumption that those four dozen cameras were selected strictly for operational reasons?

We don't (as far as I can tell) have any specifics on WHY these cameras were chosen over something else, or where they're purposed... so really, everything else so far on why they were used is nothing more than assumptions either.

Dallas has multiple Avigilon JPG2K cameras in 1.3mp resolution. I would love to hear a great pitch from anyone about why those cameras possess so much more Canadian-fried Awesome compared to a half dozen open architecture h264 cameras available at same resolution (or greater) at comparable price, even back when this project was spec'd.

I like their cameras. Especially their domes and their h264 lines. The new micro hdot they just announced (at first glance) looks like a winner too. Solid hardware engineering, robust designs, terrific image quality. To me, a case can be made for locking in a software for decades about their 16mp and 29mp cameras. They provide something above what a 1.3mp j2k offers. But even with those, the idea of removing the ability to change software platform in the future seems like too big of a trade off. As long as those cameras are there, that software is running your airport surveillance. You are forever limited to its features and level of innovation/evolution. It really is a shame they don't open the jpg2k SDK for other platforms to integrate. Ironic that a company so focused on sales hasn't figured out they'd probably exponentially increase the sales of those cameras. Unless ultimately they're more interested in selling the Dell servers and software bundled.

Dear John,

we work with Avigilon for a long time and we won many tenders with their equipment, even in spite of the fact with large storage as you mentioned. We made sure many times that expenses for labors, materials, VMS channels, ports on switch ALWAYS higher than storage costs, therefore we are very sad that JPEG2000 cameras with resolution up to 5 megapixel will be discontinued. Even many our customers are sad because they see the difference in image quality in comparison with H.264

Having large experience with JPEG2000 we made sure many times in better image quality with JPEG2000 and if the system is big than we can decrease overall quantity of workstations for images displaying (as minimum for the control) always less than for H.264.

We think that we can't compare any solutions when we don't know these solutions we can't compare them and this article isn't correct as minimum

Dear Ethan, to get maximum details you can use monochrome cameras with 0.01 lux sensitivity and we must remember that Avigilon has 35 mm full sensor but 10 MP cameras from Arecont or Stardot have 1/2 inch matrix and dynamic range and sensitivity is incomparable. We've made some tests and even in good light 3rd party 10 MP cameras provide less details, and in dark conditions image has a lot of noise and there is no sense to use these images for further investigations. In addition H.264 cuts details. To discuss H.264 we should read about it: H.264/MPEG-4 AVC is a block-oriented motion-compensation-based codec standard (Advanced Video Coding - Wikipedia) and if you go into details: (Inter frame - Wikipedia) inter frame can fail and we can lose details. And what we use in security system? Codec H.264 is perhaps best known as being one of the codec standards for Blu-ray Discs. I.e. save money on storage is much important than details in the image? Of course people have their own opinion to use H.264 because it's widespread and many manufacturers have it. But to use H.264 that loses details in system where we need to save details - it's at least senseless

Viachaslau, Thank you for the thoughtful response. Can you share specific image samples and setting specifics for the details that H.264 loses compared to JPEG2000? e.g., the same scene, same pixel count, etc.

You say it's 'senseless' to 'lose details' but it's not clear what details are lost and how valuable those details are. For instance, I've yet to see Avigilon make a clear technical case for that.

Also, can you expand on this, "made sure many times that expenses for labors, materials, VMS channels, ports on switch ALWAYS higher than storage costs." I assume you are comparing H.264 and JPEG2000. Can you break this down or use this airport as an example? It's pretty clear switching to H.264 would save about $1 million USD (given the $1.7 million spent on servers/storage). How would you make this up elsewhere?

Here's what I don't get about the 16/29MP cameras, though: what exactly do you do with 2 or 3 FPS with a color only camera? If you're looking for detail 24/7, the fact that it's color only is a knock it would cut it, nor does the framerate seem high enough to be suitable for detailed observation.

So then, ok, maybe you're looking for situational awareness. If that's the case, it seems like you could do it with 5 or 10 MP H.264 cameras. Two 10MP cameras will give you more horizontal PPF than a single 29MP, be thousands upon thousands less cost, and not lock you into a VMS.

John, this article was started from the point 'high storage costs' but not compression standards comparisons. And it's always not clear till any certain moment when we didn't get any number plate, or we didn't get any detail of the person. In investigation any detail is helpful and we work together with our local law-enforcement agencies where they already experienced this comparison by theirself. We can talk here with many arguments why we choose JPEG2000 but not any Arecont or Axis with H.264 but you will stay with your opinion and we will stay with ours, anyway we support the airport and they made the right choise

Viachaslau, you made two very strong contentions but now you are backing out of supporting them with any evidence.

Your premise is faulty: "In investigation any detail is helpful." Often additional detail is immaterial and even if it helps every so often, it might not justify the cost. That's why I am pushing to see if you (or anyone) has detailed evidence that would quantify the benefits of JPEG2000.

We have demonstrated what the cost of JPEG2000 is for this airport. One may certainly make the case that the additional cost provides commensurate value, but no one has provided any specific evidence to that end yet.

It Is All About the Video...But It Is Really About the Data...Make Sense?

I aplogise but I have not had a chance to read all of the Posts, incase I repeat something

Firstly John, how did you get hold of a security Document for an Airport, that document also seems to be for internal use of the company? I feel this is a security risk that you can get hold of a document listing what has been installed on site.

Also there is no comparison between a 5mp H.264 camera and a 35mm CCD 16 mega pixel camera.

The second question I would ask is what is the operational requirement for the system? And has the system that is being installed met that requirement?

We can all say put a lesser camera in to save money on Storage, but what is the requirement for the system?

If the system has been designed with a requirement for certain resolutions at certain distances, then I am sure the Avigilon system has been designed with that in mind.

Paul, under US law, most government contracts are publicly available (e.g., the Freedom of Information Act). Agencies always have the ability to refuse or redact information that is deemed sensitive. However, we requested through official, proper channels and was provided to us by the City of Dallas Public Information Office.

Why is there 'no comparison'? There are certainly pros and cons of using (3) 5MP H.264 cameras vs a 35mm CCD 16 MP 3ips JPEG2000 one.

As for meeting the operational requirements, this is a point the government has deemed to be 'security sensitive' and has declined any comment.

Hi John, thanks for your Response

Thanks for the clarification; US Law differs somewhat to UK Law.

16 mp is 3pps, and can achieve 4.5 pps if cropped.

Agreed there are some benefits to the H.264 algorithm above JPEG2000 and vica versa but that would be another discussion! Also the Increased frame rate etc.

But it boils down to what was the requirement? I completely understand that the Government has deemed this to be sensitive, as so it should be.

To put this into perspective,

They was a requirement at the 2012 Olympics to Monitor All traffic into the venue.

The requirement was for 7 pps, and a resolution of 80 pixels per foot across the scene.

They wanted to have an audit trail and required less cameras for ease, so were looking for megapixel solutions.

5 mp cameras were trialled by consultants and the police and they were not chosen, it then went out to higher megapixel cameras manufacturers, Avigilon being one of them.

Avigilon Beat the competitors with ease and only achieved 4.5 pps. But the Image quality was un dismissible and won over frame rate.

When the system was installed, most of the cameras were not cropped and the frame rate set for 3pps. All I can say is that the police were blown away with the quality and ease of use.

The system met and surpassed the required operational requirement.

As we do not know the requirement for Dallas, I few it is unfair to criticise.

Paul, thanks again for the feedback. I corrected the 2ips reference above.

The main observation of the original post is that JPEG2000 increases storage costs significantly. While people talk about this abstractly, this contract shows it in real life, in detail.

I agree that it is not fair to criticize Dallas for their choice, but it is certainly worth discussing the pros and cons of selecting JPEG2000 (or MJPEG) vs H.264. My 'criticism' is that spending on the storage for JPEG2000 should require detailed justification given its massive increase in cost.

"My 'criticism' is that spending on the storage for JPEG2000 should require detailed justification given its massive increase in cost."

Given that there are a lot of specifics on the "operational requirements" not included in the report, is it not a reasonable assumption that the "detailed justification" is among those specifics that Dallas felt it inappropriate to share publicly?

In short: you can't assume the justification isn't there, just because you aren't told what it is.

I am not assuming a justification is not there. There should be something but who knows how weak or strong that is. Let's put that part to the side for a second.

How many surveillance experts have commented on this post? Ten? More? I think amongst those experts, there should be someone to make a thoughtful, detailed, quantitative case on why spending the extra ~$1 million was worth it. There's been some general comments like 'better quality', 'reduced workstations' but does that amount to $1 million in greater value?

Login to read this IPVM report.
Why do I need to log in?
IPVM conducts reporting, tutorials and software funded by subscriber's payments enabling us to offer the most independent, accurate and in-depth information.
Loading Related Reports