Regional context model and dynamic Huffman binarization for adaptive entropy coding of multimedia

Classical context modeling and binarization algorithms on multimedia do not fully exploit their spatial correlations under the sequential assumption. This paper proposes a novel entropy coding scheme incorporating regional context modeling (RCM) and dynamic Huffman binarization (DHB) for multimedia. RCM evaluates the context order with the line distance in Cartesian coordinate system, so that the 2-D spatial correlations can be directly exploited. RCM predicts each symbol with the weighted estimation based on the line distance and appearing frequency of its context, which coincides with the principle for context-based prediction. On the other hand, DHB is proposed to adaptively assign the length of bins based on the estimated probability of symbols for coding. For optimal prediction and coding, frequent symbols are assigned with short bins to constrain the propagation of code redundancy. Applied into multimedia coding, extensive experimental results show that the proposed scheme achieves better coding performance than the benchmark coders.