By analyzing two characteristics of an interference multispectral image data, a compression algorithm based on adaptive classification and curve-fitting is proposed. The image is partitioned adaptively into intensive interference region and weak interference region by the mean square deviation criterion. Different fitting functions are constructed for the two regions respectively. For the intensive interference region, some typical interference curves are selected to predict other curves, and they are fitted by least square method. For the weak interference region, the data of each interference curve are approximated independently. Finally all the approximating errors of two regions are entropy coded. The experimental results show that, compared with JPEG2000, the proposed algorithm not only decreases the average output bit-rate by about 0.2 bit/pixel for lossless compression, but also improves the reconstructed images and reduces the spectral distortion, especially at high bit-rate for lossy compression.