Optimal Multiple-Bit Huffman Decoding

Abstract

In this paper, we proposed a new optimal multi-bit Huffman decoding method that combines the barrel shifter and look-ahead approaches. Specifically, we developed an approach to partition the state diagram corresponding to a Huffman decoder in such a way that the expected decoding throughput rate is maximized for a given K-bit lookahead decoding. A distinct approach taken in this paper is to exploit the probabilistic state transition probability of a Huffman entropy coder to estimate the expected decoding throughput rate of a variable-bit look-ahead decoder. We proposed a heuristic algorithm that yields very good optimization results in linear computation time.