Towards Clean Reversible Lossless Compression

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

Zip and unzip are everyday tools in today’s digital world. Since they are inherently inverse to each other, they are ideal for studying reversible computing methods on real-world problems. In this work-in-progress study, we take steps to develop a reversible zip tool. As a proof of concept, we designed clean (garbage-free) reversible versions of two algorithms, which are officially recognized by the zip-specification. Our design goal was not merely to achieve reversibility, but rather to maintain the asymptotic complexity of the irreversible counterparts. Because of their efficiency and different approaches to compression, we chose the dictionary-based Lempel–Ziv–Welch Compression (LZW) and the transformation-based Burrows–Wheeler Transform (BWT). As part of the challenge, we found a way to zero-clear the LZW dictionary and reversibly sort rotations for BWT. We have successfully created clean reversible versions of both algorithms and fully implemented and tested them in the reversible language Janus. Our reversible LZW has a worst-case runtime of Θ(n), just like the most efficient irreversible version. Our reversible BWT is, in the worst case, a factor n2 slower than the most efficient irreversible version. There are currently no better trace-free reversible methods for lossless compression.

OriginalsprogEngelsk
TitelReversible Computation - 16th International Conference, RC 2024, Proceedings
RedaktørerTorben Aegidius Mogensen, Lukasz Mikulski
ForlagSpringer
Publikationsdato2024
Sider94-102
ISBN (Trykt)9783031620751
DOI
StatusUdgivet - 2024
Begivenhed16th International Conference on Reversible Computation, RC 2024 - Torun, Polen
Varighed: 4 jul. 20245 jul. 2024

Konference

Konference16th International Conference on Reversible Computation, RC 2024
LandPolen
ByTorun
Periode04/07/202405/07/2024
NavnLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Vol/bind14680 LNCS
ISSN0302-9743

Bibliografisk note

Funding Information:
The last author was supported by JSPS KAKENHI Grant No. 22K11983 and Nanzan University Pache Research Subsidy I-A-2 for 2023.

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

ID: 397032700