Arithmetic Coding Based on Reflected Binary Codes

Abstract

Arithmetic coding is a lossless data compression technology on a sequence of symbols in independent and identically distributed (i.i.d.) distributions. In this paper, we pointed out that the conventional arithmetic coding has the possibility of burst outputs between the encoder and the output buffer, and this brings up a number of potential risks. To solve this issue, we present a new arithmetic coding which encodes the output bitstream in reflected binary codes. The analysis shows that the proposed coding method can avoid the burst outputs during the encoding. The comparisons with Huffman codes, conventional arithmetic codes and Asymmetric numeral system are discussed. Simulations show that the proposed approach has better throughput performance than the conventional approach.