A Parallel Lossless Binary linage Compression Technique Using Bayesian Networks

Abstract

Lossless compression of binary images has been an important scope of inter­ est for many years. Being lossless is crucial in many applications such as geo­ physics, telemetry, nondestructive evaluation and medical imaging, where huge amounts of data must be stored, and accurately recovered preserving each fine de­ tail in the original data. Reaching the maximum compression in minimum time has always been the goal of researchers in this field. As well as efficiency and speed, maintaining security has also been a significant issue that is to be achieved in the compression process. In this work the concept of Bayesian networks is pro­ posed to be used as a probabilistic model, making it possible to exploit the two dimensional relationship between adjacent pixels. Due to the high time and com­ putation complexity of Bayesian networks, parallel processing is used to achieve maximum compression with minimum time consumption. A distributed file stor­ ing system rather than a central one is also employed to achieve security of the stored compressed images, making it impossible to recover the original file un­ less it is authorized to access the whole parallel system. The experimental results show that the proposed approach achieves very promising results compared to other binary compression techniques. The system outperformed other techniques in some cases and achieved comparable results in the rest of the cases.