The list of numbers produced by the approximation step typically contains many zeros, interspersed with a relatively small percentage of non-zero numbers. The researchers at UCLA have identified a new way to efficiently represent this sequence of numbers using a substantially lower number of binary ones and zeros than was previously required. Such a technique, which is referred to as stack-run coding, is one-half to three times more efficient than the widely used JPEG standard. An improvement of 50% means that an image of same quality can be transmitted using this technique in half the time that JPEG requires, or equivalently, it could be stored using half as much disk space.

This invention will allow images to be transmitted or stored using far fewer resources than are currently required and could speed image compression for applications ranging from the Internet to direct broadcast satellite television. This procedure could also boost the information-carrying capacity of electronic communication links allowing, for example, a substantial increase in the traffic that could be carried over a link to an image database.

While there are other existing techniques that are known to provide improvements over JPEG, such techniques are typically far too complex to implement in the low-cost hardware that dominates the multimedia compression market. These techniques also fall short of the performance offered by the UCLA algorithm. Finally, this coding technique will be presented at the next round of the standard revisions.

Reference: UCLA Case No. 1996-590

US Patent Number: 5,818,877