Abstract: A novel representation and coding method of wavelet coefficients is proposed, which explores a new approach to improving image compression performance. Entropy values of various representations are calculated. Trinary representation is selected as an appropriate method because of its characteristics. By utilizing new symbols and states of trinary representation, more precise prediction model is built which uses multi-kind of correlations between wavelet coefficients. This prediction model is adopted to design image compression algorithm. Experimental results show that the proposed method outperforms the traditional bit-plane coding in quality of reconstructed image, which removes more redundancy among coefficients and obtains more potential to improve coding efficiency.

Key words: image compression, prediction coding, wavelet coefficient, trinary representation

摘要： 提出了一种新的小波系数表示及编码方法，探索了提高图像压缩性能的新途径。在统计各进制表示下小波系数熵值的基础上，分析了三进制表示的可行性和合理性；针对三进制表示的符号特点，充分利用小波系数之间的相关性，建立了高效的预测分类模型，设计实现了相应的图像压缩算法。实验结果表明，在各压缩倍数下，相较于典型的位平面编码方法，该方法的恢复图像主客观质量均有提高，充分说明其有效性及在提升压缩性能方面的潜力。

关键词: 图像压缩, 预测编码, 小波系数, 三进制表示