Improving Compression via Substring Enumeration by Explicit Phase Awareness

Compression by Substring Enumeration (CSE) is a recent and promising lossless compression scheme. The first experiments on CSE showed that it yields compression ratios that favorably compare to other lossless compression techniques. However, the experiments also showed that it tends to incur a performance loss on non-textual, byte-oriented sources and it was conjectured that CSE´s phase unawareness was responsible for this loss of performance. Subsequent work confirmed the conjecture by obtaining improved compression ratios when synchronization codes get inserted in the data source, indirectly solving the phase-unawareness problem. This indirect solution does not give an absolute measure of the loss incurred by the phase unawareness problem. This paper presents a modified CSE algorithm that is made explicitly phase aware. It compares the synchronization-code approach to the explicitly phase-aware approach and shows that, in the end, the approach based on synchronization codes is almost as good as the phase-aware approach.