The High Resolution Imaging Spectrometer （HIRIS） produces tremendous amounts of raw data. Transmission of these data to a ground receiving station is constrained by downlink communication bandwidth. To alleviate limitation

the data must be compressed. In this paper

we present a compression algorithm for hyperspectral image by using

three dimensional discrete cosine transform coding （3 D DCT）. The 3 D DCT includes a 1 D DCT on spectral direction followed by 2 D DCT on the eigen images produced by 1 D DCT. The main effect of the 1 D DCT is to remove the data correlation on the spectral direction

and produces the spectral band corresponded eigen images which are consisted by the DCT coefficients. The resulting spectrally decorrelated eigen images are then compressed using the JPEG algorithm. The compression ratio is controlled by the quantization （Q table） of JPEG. Image edge features are of great significance to the application of hyperspectral image. During the compression

it is pursued to retain the imgae edge feature. We have designed a Q table for retaining the edge feature with maxmum compression.