The text adopts two methods named local standard deviation and de-correlation to estimate the signal-to-noise ratio of operational modular imaging spectrometer(OMIS) multi-spectral images. The methods have reduced the influence of object texture change to a very low extent. In this way, after atmosphere correction the signal-to-noise ratio characteristic of OMIS images can reflect that of the OMIS instrument sufficiently. With regard to the image compression, the text proposes to control the peak signal-to-noise ratio of resumed image of each band exactly bigger than the signal-to-noise ratio of original image of the same band, which limits the noise brought by the compression algorithm of self within the noise of original images. Combining the thought dimensioned above, with discrete cosine transform-based and DWT-based compression algorithms, the OMIS multi-spectral images are compressed. The results indicate that the resumed image information is almost lossless in the bands of high signal-to-noise ratio is 45.86 dB.