Invasion Of The Data Shrinkers

The biggest obstacle to the vaunted multimedia revolution is digital obesity. That's the bloat that occurs when pictures, sound, and video are converted from their natural analog form into computer language for manipulation or transmission. Analog minnows become digital whales, because it takes scads of ones and zeroes to represent a single graceful analog wave. Other things being equal, a broadcast of Oprah Winfrey that fits into one analog channel would require 45 channels if sent in digital language.

Luckily, other things aren't equal. Compression, a rapidly developing branch of mathematics, is putting digital on a diet. Computerized compression techniques find the fat in a signal--that is, redundant or needless information--and get rid of it. It works so well that digital TV signals can actually be made to take up less room than analog TV signals--much less. By summer, for example, GM Hughes Electronics Corp. and Hubbard Broadcasting Inc. hope to be operating two satellites capable of broadcasting more than 150 channels of compressed digital television--in a space that could handle just 32 uncompressed analog ones.

SMALL WONDER. To make a long story short, compression is expanding--for everything from videoconferences and compact-disk encyclopedias to realistic-looking video games to Mariah Carey recordings on Sony Corp.'s pint-size new MiniDisc. Its popularity is rooted in economics: Compression lowers the cost of storage and transmission by packing data into a smaller space. Many new electronic products and services simply couldn't exist without it. An example: Playing an uncompressed two-hour movie on your computer's CD-ROM would require you to change disks 360 times. Compressed, two disks might suffice.

The potential cost savings from compression are mind-boggling, even if you count only the data-storage devices and telephone switches that won't have to be bought. If the capacity of copper wires can be significantly expanded through compression--to carry high-quality video, for example--the Information Superhighway will effectively extend to people's doorsteps without the estimated $150 billion expense of laying optical fibers to every home. Researchers connected with Stevens Institute of Technology recently claimed to have done exactly that, inventing a system that could send 10 channels two miles over existing wires.

HOT HYPE. Compression even has ramifications for scientific research. NASA estimates that by the year 2000, earth-viewing satellites will collect data amounting to a stack of magnetic tapes as high as the Washington Monument--every day. Saving and examining all that is no lark: It turns out that satellite data on the hole in the earth's protective ozone layer had been in U.S. archives for years but went unnoticed until British researchers discovered the hole independently in 1985. Triada Ltd. in Ann Arbor, Mich., is developing a supercomputer-powered system that would make such patterns easier to spot amid the enormous swamps of data, while also compressing the data drastically.

No wonder, then, that compression has become an obsession at some of the world's leading electronics companies. Last month, Philips Electronics and Compression Labs wielded their compression expertise to help win a deal potentially worth hundreds of millions of dollars to supply Bell Atlantic with set-top boxes for digital video-on-demand over fiber or copper wires. Bill Kennedy, senior vice-president of Philips Digital Videocommunications Systems, calls digital compression for video on demand "the most significant breakthrough in television since color." If that sounds like hype, he isn't alone. Says Larry Thorpe, vice-president of Sony Advanced Systems, which makes professional video gear: "Compression is absolutely the enabling technology for our future."

Today, the leading standards in visual compression are JPEG (jay-peg), for still images, and its younger sibling, MPEG (em-peg), for moving images. They're based on the work of Baron Jean-Baptiste Joseph Fourier, a French mathematician, who found in 1822 that spatial patterns can be represented numerically through a mathematical transformation. After a Fourier transform, it's easier to locate parts of an image--such as strands of hair--that can be severely squeezed without noticeable distortion.

JPEG looks for ways to compress a single image. MPEG does that, then goes on to look for redundancies between neighboring images in a stream of video. Rather than digitally record each frame in its entirety, it records just a few frames in detail--and then describes others in terms of how they differ from the detailed ones. Both standards are named after the bodies that created them: the Joint Photographic Experts Group and the Motion Picture Experts Group. And both are "lossy" techniques, meaning that they permanently discard some information during compression. "Lossless" techniques, used for computer data, achieve less compression but preserve every bit of the original input.

MULTIMEDIA APPLESAUCE. Claude E. Shannon of AT&T Bell Laboratories helped advance compression in 1948 with his theory of how to maximize the use of communication pipelines. Still, the field was quiet until around five years ago, when prices fell and effectiveness rose thanks to faster digital signal-processing chips and more efficient mathematical formulas. Take LEAD Technologies Inc. of Charlotte, N.C. In early 1992, the company introduced a special circuit board for still-image compression and decompression that cost $1,995. Today, thanks to more efficient programming, LEAD Technologies has created a software-only version that has more features and sells for $99.

There's a price to be paid for such pell-mell progress: a lack of uniformity. That's especially apparent in the field of still-image compression, where the main standard is JPEG. Software companies have rushed more than 30 "flavors" of JPEG onto the market, and unless the decoder in your computer recognizes the version that was used to compress a particular image, the result on your computer screen will be multimedia applesauce. Indeed, that is frequently the case.

There's some fear that the same thing could happen with MPEG. Here, the tiff comes down to a disagreement between video purveyors over bidirectional frames, or B frames (illustration). B frames efficiently preserve image detail by interpolating from detail contained in frames before and after them. But they add at least $40 to the costs of digital TV set-top boxes.

Tele-Communications Inc., the nation's largest cable company, would like to limit the use of B frames in its digital-transmission system, which promises to deliver 500 channels to the home. TCI hopes to lower the cost of set-top boxes without hurting video quality. But other cable companies, including Time Warner Inc., are planning to handle B frames in their set-top boxes. If the schism isn't healed, some observers fear that TVs and VCRs with built-in decoders will require extra chips to work on all cable systems. TCI has delayed deployment of 1 million digital set-top boxes until the end of 1994, citing uncertainty over the MPEG standard.

Fights over compression standards have even reached court. Stac Electronics, the leader in data compression for personal computers, is charging that Microsoft Corp.'s DoubleSpace technology infringes on Stac's patents. Microsoft has countersued. The trial, which is under way, is the first patent-infringement case against Microsoft ever to reach court--perhaps a measure of how hotly contested compression has become.

FRACTAL FERNS. The biggest battles, though, are in the research labs, where ever more advanced compression schemes are being developed. Two of the hottest areas are fractal compression and wavelet compression. Fractal compression schemes take advantage of the concept that things in the real world appear similar at different scales of magnification. For example, the spikes on a fern look like tiny copies of the fern itself. Instead of recording each spike separately, it's possible to store just one and then use it over and over to re-create the whole picture--say, for display on a computer screen. The insight of Iterated Systems Inc. of Norcross, Ga., a leader in the field, was that fractal techniques can be applied even to images where self-similarity is not as evident.

Equally intriguing are wavelets, which compress an image as a whole, rather than block by block as with JPEG and MPEG. Aware Inc. of Cambridge, Mass., is a leader in wavelet compression for varied uses requiring high fidelity such as transmitting medical images, storing music, and video editing.

Still other compression concepts are on the horizon. Meanwhile, however, less venturesome souls are happily adopting widely recognized formats, such as JPEG and MPEG. With standards more or less in place, prices falling, and the appetite for information in all forms seemingly insatiable, compression looks to have a limitless future.

Before it's here, it's on the Bloomberg Terminal.