MPEG: a video compression standard for multimedia applications

Reviewer: Bruce Jay Schachter

Digital video compression is used for a variety of telecommunications applications: teleconferencing, digital broadcast codec, and video telephony. Military and space systems have used image compression hardware for many years and will continue to do so. New uses include digital recording and playback in consumer electronics. Standardization of video compression techniques will reduce the cost of future hardware and will allow interoperability of equipment. In 1988, the Moving Picture Expert Group (MPEG) began developing a standard for compressing motion video plus audio, with 1.5 Mbits/sec as the goal rate. Seventeen groups submitted proposals to the committee. Each proposal consisted of algorithms, system claims, video, and computer files. The important considerations in choosing the MPEG standard were random access, fast forward and reverse, reverse playback, audiovisual synchronization, robustness to errors, decoding delay, editability, format flexibility, and cost tradeoffs. The goal was not to choose a single winner, but rather to combine good ideas and techniques into a final solution. A draft standard was accepted in September 1990. Only minor changes will be accepted until the standard is finalized. The final standard will take what the author calls a tool kit approach, meaning that not all its features are meant to be used all the time for all applications. The MPEG video compression algorithm relies on two techniques: block-based motion compensation for reduction of temporal redundancy and discrete cosine transform–based compression for reduction of spatial redundancy. The quality of video compression with the MPEG algorithm at 1.2 Mbits/sec is comparable to that of VHS recording. Proponents of the MPEG standard expect it to open up a broad range of applications loosely termed multimedia. The next activity of the MPEG committee is to study video compression at rates of up to 10 Mbits/sec. All image compression techniques are designed to work for typical imagery. They generally work well for complex natural scenes in which the nai¨ve observer will not know where to look for degraded information. Papers on image compression would be useful if they included examples of classes of imagery for which the compression technique works poorly.