An Image Compression System
J. Rissanen · 1986
A lossless image compression system is described, which consists of a statistical model and an arithmetic code. The model first performs a prediction of each pixel by a plane, and then it collects the occurrence counts of the prediction errors, conditioned on past pixels forming a "context". The counts are collected in a tree, constructed adaptively, and the size of the context with which each pixel is encoded is optimized.