An enhancement to universal modeling algorithm context for real-time applications to image compression

Abstract

A universal modeling algorithm, Context, introduced by J. Rissanen (IEEE Trans. Info. Theory, vol. 29, no. 5, Sep. 1983) for binary strings, is generalized for nonbinary strings, which makes it applicable to modeling many types of random processes, such as those encountered in image compression, both lossless and lossy, chaotic systems, and generally whenever prediction is needed. This generalization includes two major improvements, the control of the size of the required tree and a modification of the original context selection rule to improve accuracy and speed, based upon the idea of stochastic complexity, which in the current implementation are combined. In addition to the description of the new version of the algorithm, its application to image compression is discussed.