With more and more video surging across the Internet not just to computers but also televisions and handheld devices, something that could relieve the congestion or improve the quality would be a huge breakthrough.
That is inspiring multiple efforts to chase a long-sought — and often unachieved — goal of better compression, the trick of shrinking a movie, song, picture or other large file so that it can be whisked over the Internet without anyone noticing a difference.
Some compression companies are making some eyebrow-raising promises, such as the ability to squeeze a video file tight enough to facilitate high-definition television over the Internet. Others say their compression schemes will enhance existing video applications such as medical imaging.
“We stand a tremendous chance of becoming a de facto standard,” boasts Daniel Kilbank, who founded Bethesda, Md.-based compression-technology vendor Qbit LLC with former Apple Computer Inc. CEO John Sculley in 2003.
Qbit’s compression system is considered “loss-less,” meaning it can shrink a file without losing a detail from the original file. It uses pattern-recognition algorithms to spot recurring aspects of a scene rather than sending them multiple times.
In contrast, today’s top method of video compression, MPEG-4, is considered “lossy” because some quality is sacrificed in shrinking the file. If you see fuzzy or pixelated images online, the faults could lie in the compression.
Consequently, even if Qbit merely matches MPEG-4’s ability to shrink files but can do so losslessly, Kilbank argues that his technology “really creates a business argument” for new kinds of portable devices and Internet services with high-definition content.
Qbit figures to find its best success in “professional imaging fields” that generate huge files, such as medicine, defense, movie production and oil exploration, In-Stat analyst Gerry Kaufhold believes. Eventually, however, Qbit expects to be able to shrink files that already have been compressed with MPEG — potentially a boon to cable, satellite and phone-companies pumping out video content.
A different compression approach is in the works at Euclid Discoveries LLC, based in this proudly historic Revolutionary War town.
Euclid recently announced that it can dramatically improve on MPEG-4’s ability to render faces, while also shrinking files up to 10 times more efficiently. By focusing on faces first, Euclid hopes to spur “talking head” services, such as news broadcasts or videoconferences, on portable devices.
The compression is not lossless, but Euclid founder Richard Wingard argues that mathematical tricks will make the loss essentially unnoticeable.
For now, it appears Euclid still has far to go. In an April demonstration for The Associated Press, the system presented clearer facial images than files compressed with MPEG-4. But the edges weren’t smoothly integrated into the picture, as if the face had been glued on the background for a collage.
Wingard said those problems will soon be fixed. And he said rendering faces is one of the hardest aspects of image manipulation, so having already solved that puts Euclid close to essentially removing bandwidth constraints for high-quality video. Euclid pledges to eventually get a two-hour movie down to 50 megabytes — small enough to fit 20 on a portable USB drive, for example.
Indeed, In-Stat’s Kaufhold believes that Euclid has “something valuable at its core” that might lead to delivering HDTV over the Internet.
But other observers are far more skeptical. For one thing, innumerable compression claims have been made over the years only to fizzle.
And while there have been notable compression advancements over the years from such players as Divx Inc. and On2 Technologies Inc., those technologies power niche applications.
One reason is that existing, lossy compression — such as the MPEG-2 format that encodes DVDs — is usually good enough for entertainment purposes, said Predrag Filipovic, an analyst with The Diffusion Group.
He also notes that a startup compression scheme always faces an uphill fight. It must attract the interest of equipment makers who are reluctant to stray from industry standards, and it has to overcome rival offerings from full-service providers such as Microsoft Corp.
“Is there a need for yet-better compression? Yes, you can argue that there is,” Filipovic said. “The question is how much better somebody has to be in order to justify the huge cost of marketing this thing and the huge cost to the provider. Those equations usually do not end up in a revolution.”