A finite-precision adaptation of bit recycling to arithmetic coding

The bit recycling compression technique has been introduced to minimize the redundancy caused by the multiplicity of encodings present in many compression techniques. It has achieved about 9% as a reduction in the size of the files compressed by Gzip. In prior work, we have proposed an arbitrary-precision technique to adapt bit recycling to arithmetic code instead of Huffman code. We have shown that this adaptation enables bit recycling to achieve better compression and a much wider applicability. We have also presented a theoretical analysis that estimates the average amount of data compression that can be achieved by this adaptation. In this paper, we propose the finite-precision version of this adaptation so that it can be implemented efficiently using conventional computer registers.