Improving LZ77 bit recycling using all matches

Abstract

There exist lossless compression techniques, such as LZ77, that have the particularity that some original file may be compressed in more than one way, e.g. by choosing other matches than the closest longest ones only. The existence of multiple encodings per original file causes redundancy, i.e. it tends to make compressed files longer than necessary, on average. Recently, a technique called bit recycling was introduced to help reduce the redundancy caused by the multiplicity of encodings. It has been used to improve LZ77 compression. It exploits the fact that there often exists more than one longest match and it is called longest-match bit recycling. This work presents a more general, and more powerful, bit recycling technique that exploits shorter matches also. We call the technique all-match bit recycling. Our experiments demonstrate that at least 1 bit out of 11 results from the multiplicity of encodings, in LZ77 compression.

Constructing optimal whole-bit recycling codes

“…The side-channel is general purpose and it may carry any useful data. In previous work, it has been used for information hiding, or steganography [2], for authentication [1], for error detection or correction [9], and for bit recycling [4], [6], [13]. This last application is described in greater detail next.…”

“…Bit recycling aims at improving the rates achieved by compression techniques, such as LZ77 [14], that suffer from the redundancy caused by the multiplicity of the encodings [4], [6], [13]. Bit recycling takes the pragmatic approach of detecting and exploiting the multiplicity in order to recover a compensation instead of trying to eliminate the multiplicity at the source (e.g., Kawabata's approach for LZ77 [8]).…”

Adaptation of bit recycling to arithmetic coding

“…It turned out that bit recycling was able to achieve better compression by exploiting the multiplicity of encoding feature rather than systematically selecting the shortest codeword. Variants of bit recycling have been applied on LZ77 algorithm in [6] and [8], the experimental results of [7] showed that bit recycling has achieved a reduction of about 9% in the size of files that has been compressed by Gzip [9]. The authors of bit recycling have pointed out that their technique could not minimize the redundancy perfectly since it is built on Huffman coding [10], which does not have the ability to deal with codewords of fractional lengths, i.e.…”