M-correlated dynamic Huffman code

The compression results of traditional variable length coding (VLC) are limited by the global entropy bound calculated from the statistics of entire source. Many entropy coder have been proposed for the purpose of approaching this compression bound as close as possible. In this paper, the local entropy (LE) coding principle is given to reduce the coding bound by partitioning the source into parts while it still can be exactly recovered. In addition, the LE theorem is also developed for the purpose of realizing the above principle for practical applications. The M-correlated DHC (MDHC), which is a realization of the LE theorem, is proposed for image compression. The compression result of MDHC shows higher compression ratios than adaptive Huffman code (HC) in every case of our experiment.<>