FEATURE

Video Compression: The Big Squeeze

Delivery of raw video means big-bucks bandwidth-a steady 90 Mbps for simply sampled pulse-code modulation. Nowhere near that amount of bandwidth is cost-effectively available on most networks or media storage devices, so compression is crucial to successful delivery.

MPEG-2 is the standard for Digital Satellite Video, which operates in bandwidths from 4 Mbps to 20 Mbps-this means compressing video at the ratio of 23-to-1. CD drives produce steady bitstreams of not much more than 1.5 Mbps. Material to be stored on a CD must be compressed by a factor of 60-to-1 using MPEG-1. The greatest compression, more than 1,400-to-1, is required for H.261 standard videoconferencing operating over communication links that run at speeds less than 128 Kbps.

All of today's popular video-compression systems use a common strategy for encoding digital picture information. The encoder looks for frame-to-frame similarities inherent in typical video. In many sequences, for example, background areas do not change, and motion of distinct objects may track predictably through the scene. The MPEG and H.26x protocols achieve compression by transmitting a complete picture, called an "I" frame, only two to three times per second. Between these I frames, predictive "P" frames and bidirectional "B" frames may be transmitted. These are highly compacted frames that are coded to describe only the parts of the scene that have changed.

H.26x protocols, associated primarily with live two-way videoc onferencing at speeds between 56 and 1,936 Kbps, encode 352x288 pixels and use combinations of I and P frames. MPEG-1, designed for optimum efficiency at the bandwidths available on T1 networks or CD-ROM drives (about 1.5 Mbps), encodes 360x243 pixels using reordered I, P and B frames. The B frames are transmitted in a different order than the one they were generated in so they can be combined at the decoder with nearby P frames to fill in only what has changed in a picture. MPEG-2, the successor to MPEG-1, is optimized for the digital compression of television broadcast material for transmission at 1.5 to 6 Mbps. MPEG-2 encodes 720x486 pixels, with little visible degradation from NTSC-quality video. The National Television Standards Committee (NTSC) defines the 525-frame, 60-frame-per-second broadcast and VCR standard used in North America.

MPEG-2 is expandable. The protocol permits specification of several levels of decoders, as well as different types of video sources. It is flexible enough to accommodate high-definition television (HDTV).

MPEG, named after the Motion Picture Experts Group, also defines compression of audio information. While MPEG-1 handles only left/right stereo audio channels, MPEG-2 has five compressed audio channels (left, center, right, right surround rear and left surround rear), as well as a special channel for low-frequency effects. The MPEGs are highly efficient for digital audio and video transmissions, but they are not easy to edit.

Another compression standard, specified by the Joint Photographic Experts Group (JPEG), takes advantage of the characteristics of the human eye and removes picture information that is not very visible to viewers. Motion-JPEG stores I frames sequentially. It is not as efficient as MPEG, but the video content is easy to edit. You may find many JPEG files on systems used for digital video production.

Since desktop workstations are natural devices for viewing video that comes over a network or on CD-ROMs, two major encoding techniques have emerged f rom the PC industry-Apple's QuickTime (stored as *.MOV files) and Intel's Indeo (stored as *.AVI files). These were born in environments of narrow disk or CD-channel bandwidth and have limited processing power for picture regeneration. It's no surprise that they show up best in limited screen windows (320x240 pixels) at 10 to 20 frames per second (see our test results in Low-End Intranet Video.)