中文摘要：

针对原EZW算法未能很好利用图像小波系数特点及按照频率特性量化小波系数的不足，提出了对图像小波系数进行信噪分离、阈值化处理以及对低、高频图像信息进行分阈值量化的改进算法，并给出了在保证复原图像质量情况下扫描终止的判别条件，以节省压缩时间，在实时传输中能有效地提高图像压缩效率。仿真实验结果表明，改进算法无论在扫描相同次数下的信噪比，还是相近信噪比下的压缩比都获得了较大改善，为小波变换下的图像压缩方法提供了新的思路。

英文摘要：

Aiming at disadvantage and insufficiency in primal EZW algorithm whose character of wavelet coefficient is not quantized according to frequency characteristic, a new improved algorithm is presented which includes signal-noise separation of wavelet coefficient , the valve disposition and different threshold quantization between low frequency and high frequency image information, in the case that make sure quality of rebuilt image, it proposes the judgment condition of stopping scanning, which saves scanning time more and can improve efficiency of image compression effectively during process of instant transmittal. The simulations show that the improved algorithm makes progress not only in PSNR of the same scanning times, but also in CR of approximate signal to noise ratio. Meanwhile, it provides new way of image compression under wavelet transform.