Abstract: With the rapid development of aerial and space remote sensing technique, the digital camera has been explored,and mapping experiments with this kind of digital images are processing as the regular mode of aerophotogrammetry. On the other hand, the technique of integration of 3S (GPS,RS and GIS) and digital photogrammetric system have been put in actual use as a mature technique. Meanwhile, the information superhighway will be the new key technique in the coming 21 century. All these developments mean that for the spatial geographical data require muti-level remote sensing image data. In RS, GIS and DPS(digital photogrammetric system), one of key techniques is how to deal with the real time transmitting of huge remote sensing data and how to build image data base. So the investigation technique of compression of remote sensing data is a very urgent task. Normal data encoding technique comes to loss-less compression, while the compression ratio only ranger about 2 and can't be satisfied with actual requirement. The new developing compression techniques, such as wavelet compression and fractal compression will reach a high compression ratio, but they belong to degraded compression and need much more CPU time.So for the time being, these methods have not much actual use in the field of the remote sensing.In this paPer, the technique of guasi-lossless compression based on the image restoration is presented. The technique of compression described here will include three steps, namely, bit compression, removeing correlation and image restoration based on the theory of modulation transfer function(MTF). The guasi-lossless compression comes to high speed and the quality of reconstructed image under restoration reached the guasi-lossless level with higher compression ratio.The test of TM and SPOT remote sensing images show that the average compression ratio is about 4~5, the fidelity reaches 0.99 and Peak value signal-noise ratio(PSNR) is over 42. All of the results confirm that the technique is reasonable and applicable.