Bi-level image compression using adaptive tree model

Abstract

Summary form only given. State-of-the-art methods for bi-level image compression rely on two processes of modelling and coding. The modelling process determines the context of the coded pixel based on its adjacent pixels and using the information of the context to predict the probability of the coded pixel being 0 or 1. The coding process will actually code the pixel based on the prediction. Because the source is finite, a bigger template (more adjacent pixels) doesn't always lead to a better result, which is known as "context dilution" phenomenon. The authors present a new method called adaptive tree modelling for preventing the context dilution. They discussed this method by considering a pruned binary tree. They have implemented the proposed method in software.