String dictionary structure for Markov arithmetic encoding

Data from statistically unknown sources is generally difficult to compress with significant ratios. The compression scheme should have the ability to gather context information of the source in an adaptive manner. Arithmetic coding has that capability and can be made to outperform many other compression schemes known. The sliding dictionary, on the other hand, is a compression algorithm which is best suitable for hardware implementation on systolic-array parallel processors, but due to the statistical properties of its information context, it results in inferior compression ratios. A string dictionary structure is proposed that combines the advantages of both algorithms, offering the possibility of hardware implementation on VLSI, avoidance of encoding almost random information, and encoding with arithmetic codes.<>