Chapter 5 - Adaptive Huffman coding

Abstract

In static Huffman coding, the probability distribution remains unchanged during the process of encoding and decoding. The amount of the loss in compression quality depends on how much the probability distribution of the source differs from the estimated probability distribution. Adaptive Huffman coding algorithms improve the compression ratio by applying to the model the statistics based on the source content seen from the immediate past. An alphabet and its frequency table are dynamically adjusted after reading each symbol during the process of compression or decompression. Compared to static Huffman coding, the adaptive model is much more close to the real situation of the source after initial steps. Adaptive Huffman coding works on dynamic statistical models. The statistical models may be adopted to work more closely with the coders. The probability distribution is computed applying a frequency count and adjustment of alphabet after each new symbol being input into the coder. Two types of codes are used and a switch codeword is used to flag the alternative use of the codes.