Embedded image-domain compression using context models

Transform coding techniques are popular for their excellent lossy compression for natural images but fall short with “simple” images, i.e. images represented by a sparse subset of the available intensity values. In this paper we propose a novel image-domain compression technique aiming at simple images. The proposed method is a context-adaptive bitplane coder where each bitplane is encoded using a binary arithmetic coder. Our novel context modeling scheme allows to exploit the strong correlation across the bitplanes of simple images. Together with the scalable bitplane reduction technique proposed for increased compression efficiency, our bitplane coder produces an SNR-scalable embedded bitstream that is lossless when completely decoded