Data Compression Method (Information Networks) U.S. Patent 5,537,551

Document Type


Publication Date



A method for compressing and subsequently decompressing digital data communicated in an interactive computer network, the network designed to provide informational and transactional services to a very large population of users. The method features steps for compressing bytes of network data before transmission by substituting variable-length code words obtained from a fixed, look-up table, and, reconstituting the bytes using a fixed, decompression look-up table when the code words are received at the data reception site. In accordance with the invention, the compression and decompression look-up tables are statistically compiled by sampling the occurrence frequency of byte pairs in the network data stream, and where byte pairs are found to occur above a predetermined frequency, code words having lengths inversely related to the occurrence frequency are created for inclusion in the table so that a code word may be substituted for one byte of a pair when the other byte of the pair is found to precede it during compression, and the byte reconstituted from the code word using the decompression table when the code word is received at the reception site. Additionally, where a byte and its preceding byte constitute a pair not found within the pairs compression table, the method features steps for transmitting the byte compressed in accordance with a context-free encoding scheme, together with a suitable escape code word. Yet further, the method features steps for combining other compression and decompression procedures with the byte-pair compression and decompression to produce a compound compression scheme for the network data stream.


U.S. Patent 5,537,551 issued July, 1996.

Published Citation

Denenberg; Jeffrey N., Weinberger; Edward D. , Gordon; Michael L.. Data Compression Method (Information Networks) U.S. Patent 5,537,551 issued July, 1996