Probability Interval Partitioning Entropy Codes

A novel approach to entropy coding is described that provides the coding efficiency and simple probability modeling capability of arithmetic coding at the complexity level of Huffman coding. The key element of the proposed approach is given by a partitioning of the unit interval into a small set of disjoint probability intervals for pipelining the coding process along the probability estimates of binary random variables. According to this partitioning, an input sequence of discrete source symbols with arbitrary alphabet sizes is mapped to a sequence of binary symbols and each of the binary symbols is assigned to one particular probability interval. With each of the intervals being represented by a fixed probability, the probability interval partitioning entropy (PIPE) coding process is based on the design and application of simple variable-to-variable length codes. Probability modeling can either be fixed or adaptive while the PIPE code for each probability interval remains fixed and is decoupled from the modeling stage. The excess rate of PIPE coding is shown to be comparable to that of arithmetic coding.