A creative chain coding technique for bi-level image compression inspired by the NetLogo HIV agent-based modeling simulation

In this paper, we develop a new chain coding application stimulated by the existing NetLogo HIV agent-based model and utilize it in bi-level image compression. Our paper is an extended version of our previously published paper in the International Conference on Computational Science 2021. Our method considers converting an image into a virtual environment, which maps to the original image and consists of HIV+ and HIV- female agents depending on the distribution of the pixels. Then, the algorithm introduces HIV+ male agents the purpose of which is to move around and infect other HIV- female agents. The movements of the HIV+ male agents are designed in a way that follows the relative coding approach, utilized in different chain coding projects. The relative coding increases the likelihood of generating consecutive codes that are encoded in a similar manner, and therefore, helps in providing better compression ratios. The algorithm tracks certain HIV+ male movements and uses them along with other pieces of information to reconstruct the original image back. As the literature shows, agent-based modeling can be advantageous over mathematical techniques and it can be effectively applied in some domains. The outcomes revealed that we could outperform standardized benchmarks used by researchers in the image processing community. Additionally, the paired samples t-tests reveal that the mean differences between our results and the ones generated by the other benchmarks (e.g. JBIG2) are statistically significant.

Automated summary

This paper presents a chain coding application based on an agent-based model for image compression. The method converts the image into a virtual environment and utilizes relative coding to achieve better compression ratios. The results show that this method outperforms other benchmarks and has statistically significant differences.