The authors present the adaptive Lempel-Ziv-Welch general-purpose algorithm and its implementation suitable for packet radio, telephone transmission, and archival storage. While the statistical variable-length Huffman technique compresses text by 20%, the LZW technique can compress data (text, numeric, mixed, and bit-mapped images) by 40 to 60%. The adaptive LZW algorithm has very simple logic, leading to inexpensive and fast implementations. Good LZW implementations use 9- to 16-bit codes, handling most applications. A 12-bit code is suitable for medium-size files. Efficiency improves with larger codes. A tight coding of the algorithm can compress 75 Kbytes in a second on a 1-MIPS machine. >