Analysis of Cascading Compression Algorithms

Abstract

The Huffman coding scheme is a character-based algorithm in which every leaf node represents a character only. In this paper, we study three variations of the Huffman coding scheme for compressing $16$-bit Chinese language. Although it is observed that $IDC$ can generate the shortest code length among the three variations, but its empirical compression ratio is below $1.8$, which is unsatisfactory. In order to achieve higher compression performance, i.e., compression ratio over $2$, word-based compression algorithms should be employed. A possible way to develop word-based algorithms is to use the technique of cascading. Two kinds of algorithms are chosen for cascading. They are $LZ$ algorithms and the Huffman coding scheme. $LZ$ algorithms are used for finding repeating phrases while the Huffman coding scheme is used for encoding the phrases instead of characters. The experimental results show that the cascading algorithm of $LZSSPDC$ outperforms a famous $UNIX$ cascading compressor $GZIP$ by $5\mathrm{\%}$ on average.