A lossless image coder with context-based minimizing MSE prediction and entropy coding

In this paper, a novel gray-scale lossless image coder, based on context-based minimizing mean squared error (MMSE) prediction and entropy coding, is proposed. The proposed algorithm consists of a prediction stage and an entropy coding stage. To obtain the context of prediction, classification of directional edges is employed, which characterizes two-dimensional image behaviors, such as edge and smooth region. Based on the context model, adaptive DPCM prediction coefficients are obtained in the MMSE sense. The MMSE method based on a context-by-context basis decorrelates the neighboring pixels very effectively in the prediction stage. In the entropy coding stage, the context modeling techniques also provide an improved coding gain. To exploit the statistical redundancy still remaining in the residual image, many contexts are preset in the entropy coding stage, and merged into a small number of contexts for complexity reduction. From experimental results, it is found that the proposed lossless coding scheme slightly outperforms the CALIC which is known to achieve the highest compression ratio in entropy coding