Bi-Level Image Compression Estimating the Markov Order of Dependencies

Abstract

This paper presents a bi-level image compression method based on chain codes and entropy coders. However, the proposed method also includes an order estimation process to estimate the order of dependencies that may exist among the chain code symbols prior to the entropy coding stage. For each bi-level image, the method first obtains its chain code representation and then estimates its order of symbol dependencies. This order value is used to find the conditional and joint symbol probabilities corresponding to our newly defined Markov model. Our order estimation process is based on the Bayesian information criterion (BIC), a statistically based model selection technique that has proved to be a consistent order estimator. In our experiments, we show how our order estimation process can help achieve more efficient compression levels by providing comparisons against some of the most commonly used image compression standards such as the Graphics Interchange Format (GIF), Joint Bi-level Image Experts Group (JBIG), and JBIG2.