Analysis and architecture for memory efficient JBIG2 arithmetic encoder

JBIG2 is the latest international standard for bilevel image compression. It partitions a bilevel image into three types of region - text, halftone, and generic region. This paper addresses the problem of large memory requirements for the contexts. The arithmetic encoders of text region in JBIG2 require a large memory for the contexts. In this work, we proposed several algorithms to reduce the memory requirements. A large portion of total memory requirements are reduced by the proposed algorithms. Moreover, the proposed algorithms are implemented as memory efficient hardware architecture, the unified arithmetic encoder. The experimental results show that the proposed algorithms are efficient. The proposed algorithms can reduce the memory requirements for the contexts by 98.7%. Moreover, the drop of coding gain due to memory reduction for the contexts is only 4.83% in average.