Adaptive binary arithmetic coding for multi-media applications

An adaptive binary arithmetic coder serves as a universal component for data compression when coupled with an appropriate model structure or context-selection for conditional probability distributions. Appropriate model structures have been developed for files of 8-b/byte data for digital gray-scale images, for binary images, and for digitized speech. The flexibility of arithmetic coding allows the modeling of the data to be done independently. With Huffman coding, statistical models must be selected in view of coding limitations. Moreover, arithmetic coding permits the adaptive determination of both the contexts and the statistics based on the changing statistics of the source data.<>