依据GMS卫星红外气象云图的统计特性，提出了一种“亚抽样＋修改的JPEG＋图象放大”的静止图象编码方法，使有失真压缩比在50:1至100:1时，恢复图象峰值信噪比在35dB－40dB之间，主观质量无视觉可觉察到的失真。对于信息保持型GMS卫星红外云图数据压缩，选取训练集求取最小均方误差意义下的最佳线性预测器，并作为固定预测器应用于每幅云图，性能优于目前文献报导的固定预测器。

On the basis of the statistical characterestics of satellite meteorological infrared cloud-charts this paper presents a coding algorithm of“Sub-sampling+Modified JPEG+Image enlarging”,realizing a compression ratio of 50∶1～100∶1, with the PSNR of 35～40dB and visually unperceptible distortion. For error-free data compression of satellite cloud-charts, training set is selected to seek the minimum mean-square-error preditor,which is used as the fixed predictor and applied to the compression of any cloud-chart image, with a better performance than any other fixed predictors reported so far.