LZP: A New Data Compression Algorithm

Introduction A new compression algorithm is presented which may either be considered an improvement to dictionary coding or an improvement to context coding. This algorithm works by reducing the set of available window position, for an LZ77 encoder to match from; thus fewer bits are required to indicate the match-from location, and higher compression is attained because the average match length does not decrease as much as the output code size does. This algorithm is shown to be the best solution currently available in all situations, including archivers, distribution, and on-line compression such as disk compression or network datagram compression. Four variants are presented; LZP1 is faster than LZRW and gets higher compression; LZP2 is fast enough to be used in on-line applications (faster than the commonly used Stac LZS algorithm) and achieves compression on the order of LZSS; LZP3 achieves compression comparable to that of the most advanced hybrid LZ77+Huffman methods (such as the Zip algorithm) while operating at three times the speed of PkZip; LZP4 attains an average on the Calgary Corpus of 2.29 bits per byte, comparable to that of the state-of-the-art PPMD+ algorithm, while operating much faster and using less memory.