Improved JPEG 2000 system using LS prediction and grouping context coding scheme

In this paper, two coding strategies are proposed to improve the coding efficiency of the JPEG2000 system. First, instead of using the embedded block coding with optimized truncation (EBCOT) scheme in all subbands, we apply the algorithm based on least square prediction in the LL part of the discrete wavelet transform domain. Moreover, in the LH, HL, and HH parts, instead of using the MQ coder and the fixed probability table, we group the 19 contexts into 7 classes. Since the characteristics of the contexts in EBCOT are different from one another, it is proper to use different probability tables for different classes. Simulation results demonstrate that the proposed methods significantly improve the coding efficiency of the JPEG2000 system.