Universal data compression algorithms by using full tree models
S.-Y. Chang, John J. Metzner · 2002
Summary form only given. A brief description and results of analysis of splitting and merging for constructing and updating full trees is provided. Since all the leaves are assumed to have equal probability, using these two tree-updating procedures the variable-to-fixed length codes are realized. Two algorithms are suggested. One is devised by modifying the Ziv-Lempel algorithm. The number of leaves varies but is always a power of the size of an encoded output alphabet. In the second algorithm trees always have a fixed number of leaves which is the power of the size of the encoded output alphabet. The two proposed algorithms are evaluated by the normalized average code length per source output symbol, required memory space to store information about trees, and encoding complexity.>