基于层次模型的重要性编码算法

A Significance Coding Algorithm Based on Layer Model

该文提出了基于层次模型的重要性编码算法,该算法在不改变JPEG2000上下文模型前提下按照重要系数的空间聚集区域进行层次邻域编码。实验结果表明:新算法输出的码流的自相关性较JEG2000以及Hilbert曲线输出的码流的自相关性好;新算法的平均码率较JPEG2000的条带扫描以及Hilbert曲线扫描的平均码率分别提高了1.06%,0.57%;而且,新算法的平均码率较JPEG2000的上下文量化优化算法获得的平均码率提高了1.08%。

A significance coding algorithm based on layer model is proposed. Without changing the context template of JPEG2000, the proposed scheme uses an explicit layer template to dynamically arrange the scan path of wavelet coefficients. Experimental results show that the proposed scheme provides better autocorrelation outputted bit stream than stripe scanning of JPEG2000 and Hilbert curve scan; and the proposed scheme improves over stripe scanning and Hilbert curve scanning in terms of average bit rates by 1.06% and 0.57%, respectively. Moreover, the proposed scheme achieves an improvement 1.08% over context quantization optimization algorithms based on JPEG2000 in terms of average bit rates.