Abstract:
A classic area of application for variable-length codes is data compression. This paper looks at a class of variable-length codes, the T-codes, that have been noted for their self-synchronisation properties. While the coding efficiency of T-codes is less than or equal to that of Huffman codes, no simple algorithm for the construction of T-codes from a given set of source symbol probabilities has been developed so far. To date, the only approach to finding the most efficient T-code set for a given source seems to be an exhaustive search, which to date has been too complex to be of much practical value. This paper describes some shortcuts for this search and presents recent improvements that yield a much lower computational complexity.
