Earlier this fall at IBC in Amsterdam, Thomson Grass Valley introduced its new Infinity line of products intended to give broadcasters a new approach to high definition acquisition and storage.
With support for JPEG 2000 compression, Infinity broadens the range of approaches broadcasters must consider when weighing HD trade offs in terms of budgets, quality, infrastructure improvements and a host of other considerations.
At the heart of Infinity’s JPEG 2000 support is compression and decompression processors from Analog Devices. High Definition Technology Update talked to Bill Bucklen, the company’s Advanced TV Segment Director, for background on the compression format and how it might impact the unfolding HD options from which broadcasters will ultimately choose.
HDTU: JPEG 2000 has grabbed the attention of the broadcast and film industry lately. What’s Analog Devices’ history with the compression standard?
Bill Bucklen: Analog Devices has been doing wavelet-based compression for over 10 years. Our major focus was in the surveillance market where the attractiveness of wavelets was evident. These were closed systems, and surveillance brings out the need of efficient storage in a system.
JPEG 2000 was developed and was close to what we were doing, so we had an advantage. We were the first with a compliant chip. So, we have a great hammer, and we have been going around looking for nails, like surveillance and military. Increasingly, our technology was broadly adopted by the U.S. government, which embraced it for satellite and military image storage and by the National Archive, which is using it for storing documents and photos.
HDTU: Surveillance isn’t the only application where efficient storage is critical. Digital cinema and HDTV, with their significant file sizes must been another obvious “nail” for your JPEG 2000 “hammer.”
BB: JPEG 2000 had a big win last year when it was selected as the digital cinema standard compression technology for distribution of feature films. The movie studios wanted perfect quality, and JPEG 2000 was far and away the winner. It’s designed into many of the digital cinema standards.
The next step in the food chain is the broadcast industry. Thomson Grass Valley saw the value of JPEG 2000 in that application. We think that this is the beginning of a significant move through the entire chain — with the exclusion of terrestrial transmission. I am not going to talk to the ATSC about changing the terrestrial broadcast standard.
The scalability of JPEG 2000 is what attracted the broadcast space. For newsgathering, it is now possible with the Thomson Grass Valley Infinity to shoot HD content, and then transmit a lower resolution version of the content directly out of the storage media without compression or scaling — for instance, shooting under a tank with a cell phone strapped to your back. That can’t send HD live, but it can send a low resolution version.
When the news crew has access to more bandwidth and the HD arrives, it matches up frame for frame every time you show the same clip. It is a really neat scalable thing.
Thomson Grass Valley really stepped out and has taken the lead. Hopefully, other companies will recognize the advantages as well, and JPEG 2000 will grow in the broadcast industry.
HDTU: How do you see it specifically impacting broadcasters as they continue on their DTV transition and look to upgrade their facilities for HDTV?
BB: We believe JPEG 2000-based editing equipment can be less expensive than MPEG. It also can reduce the operating cost by streamlining how the editing process is done.
Editing systems are not happy working with full resolution HD images. To reduce the resolution of the image takes time. In the case of JPEG 2000, you just don’t have to do that. You simply pull off the resolution you want to work with from the core data.
The way JPEG 2000 assembles images is by beginning with little thumbnails and adding more and more detail, so you can pull off the pieces that you want.
HDTU: Could you explain the advantage JPEG 2000 might have over MPEG compression formats for HD?
BB: JPEG 2000 is an interframe compression approach. There’s no motion estimation, and for editors that is a dream. Each frame stands alone, and it’s robust in terms of transmission errors so errors won’t degrade video in as noticeable a way as with MPEG.
JPEG 2000 offers low latency — less than one frame — so two-ways can be spontaneous. That half second with MPEG can be very disturbing and destroy the spontaneity of the interview process.
The fact is that the compression output rate is highly deterministic. In MPEG, you determine the available bandwidth and run the algorithm. In JPEG 2000, say you want 5Mb/s, that’s what you get. If it suddenly drops to 2Mb/s, that’s what you get.
You can cut the number of pixels down easily. The way reduction of quality hits JPEG 2000 is different than the way it affects MPEG. In MPEG, you see lots of motion artifacts and you also get the blocking artifact. In JPEG 2000 you get a softening of the picture and a little noise. It looks more “analogie” than digital in terms of errors.
JPEG 2000 is very watchable with errors as opposed to be being very disturbing like MPEG with errors. There’s also bit error resiliency. A bit error in MPEG can affect one block, or one line of blocks across several frames. In JPEG 2000, a bit error typically adds a tiny bit of error. The visual impact of the error is much lower at any given error rate.
HDTU: How will the costs of JPEG 2000 compression versus MPEG compression compare?
BB: JPEG 2000 is a highly symmetrical compression technology. It takes the same horsepower to do a compression and a decompression and the same amount to do any quality of compression.
Compared to MPEG, it is very efficient. But to do MPEG requires a sophisticated motion estimation algorithm. MPEG is the perfect solution for asymmetrical applications like DVDs or broadcast distribution.
A DVD may spend weeks decompressing and compressing to optimize a specific scene and stop the instant it fits in the DVD. But if you are doing it in real time, you don’t get that choice. The only choice is to throw more hardware at it. Doing real-time MPEG compression is very expensive — $50,000 to $100,000 for real-time compression of MPEG.
JPEG 2000 compression requires two chips for HD and one for SD, so it is a very cost-effective solution.
HDTU: Are there consumer applications for JPEG 2000 compression?
BB: The low cost of compression, scalability, deterministic data rate and low latency are important characteristics of JPEG 2000. There’s no annoying delay when pressing the channel change button. If you’re playing a game on an HD wireless, and you move the mouse and wait a second for the character to turn around, that can be fatal. The consumer arena will drive the whole cost of JPEG 2000 applications down to consumer levels.
In the consumer space of DVRs and Tivos, one of the things you have to do is pick a quality level. The basic level is pretty bad even for SD. With JPEG 2000, you can pick the lowest quality you are ever able to accept and then record at the highest. Later, you can go back through the material and reduce quality overtime to balance image quality versus storage.
Tell us what you think!
HDTU invites response from our readers. Please submit your comments to firstname.lastname@example.org. We’ll follow up with your comments in an upcoming issue.