Image compression based on Gaussian mixture model constrained using Markov random field

Jianjun Sun, Yan Zhao, Shigang Wang, Jian Wei

Abstract: Highlights•An image compression scheme based on GMM-MRF is proposed for the first time in this paper.•A K-median algorithm, which can avoid being influenced by the extreme values, is proposed to replace K-Means in the initialization step of the Expectation-maximization algorithm.•A Mixture Model Optimization is proposed in this paper to determine the optimal number of distributions before the prediction.•The modeling performance of Gaussian Mixture Model is improved using Markov Random Field.