中文摘要：

结合人眼对比度感知特性和图像离散余弦变换特征，提出了一种图像的最小感知误差阈值的计算方法，利用该方法计算的阈值来量化源图像频谱图的各个子块的信息，使得编码时变换域系数中尽可能多的出现零值系数，再通过Huffman编码方法实现编码，并通过实验验证，结果表明平均码长为1．3933，接近于信源的熵1．1231，编码效率达到0．8823，压缩比达到13．2299，峰值信噪比为45．4775dB。说明衡量数据压缩性能的性能指标均达到较好效果，同时又因为结合了人眼视觉特性，保证了压缩图像的观测效果。表明提出的阂值计算方法和图像压缩方案是有效的、实用的。

英文摘要：

Combined the contrast sensitivity characteristics of the human vision system and image characteristics by the discrete cosine transform, the method is proposed to calculate the threshold value of human vision error detecting the spectrum image of different frequencies. And the original spectrum image information of each sub-block is quantified by the threshold calculated in order to occur more zero coefficients in the transform domain coefficients while encoding, and to carry out encoding through Huffman coding method. The experimental results show that the average code length is 1.393 3, close to the entropy 1.123 1 of the image source, and the coding efficiency is 0.882 3, while the compression ratio is 13.229 9, and the peak signal to noise ratio is 45.477 5 dB. The performance parameters which measure the data compression are good, at the same time, because it is combined with the visual characteristics, observed effect of the compressed image is ensured, which show that the proposed threshold calculation method and image compression scheme are effective and practical.