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。 According to the statistical characteristics of GMS satellite infrared meteorological cloud images, a method of still image encoding “sub-sampling + modified JPEG + image enlargement” is proposed to restore the image with the distortion compression ratio from 50: 1 to 100: 1 Peak signal to noise ratio between 35dB ~ 40dB, no subjective visual quality perceived distortion. For the information-keeping GMS satellite infrared cloud image data compression, the best linear predictor in the sense of minimum mean square error is selected as the training predictor and applied to each cloud as a fixed predictor. The performance is better than the fixed predictor reported in the literature .