Significant bit-plane clustering technique for JPEG2000 image coding - Electronics Letters

Introduction: The JPEG2000 standard [1] has shown better performance than the widely used JPEG standard [2]. Nevertheless, efforts to improve JPEG2000 never stop. Long et al. [3] modified the quantisation step-size selection schemes for the uniform scalar quantisation used in JPEG2000 to improve the lossy compression. Lian et al. [4] proposed two skipping methods applied to the embedded block coding to reduce JPEG2000 encoding time. Embedded block coding with optimised truncation (EBCOT) is the core technique in JPEG2000. EBCOT defines three fractional bit-plane passes: the significance propagation, magnitude refinement and cleanup passes. Each coefficient bit in a bit-plane is coded in one of the three fractional bitplane passes and the contexts are determined by the current coefficient and its neighbouring coefficients. EBCOT can achieve high coding efficiency using context-based adaptive arithmetic coding and rate distortion optimisation. The motivation of the proposed clustering method is to raise the proportion of the significance state, for which the bit value is 1, in the significance propagation pass. To achieve this, significant quantised coefficients are clustered together so that these significant bits in the bitplane scan can be encoded in the significance propagation pass, and thus the coding efficiency of EBCOT is improved.