A compression algorithm based on 3D-DWT and 3DSPIHT in conjunction with ROI is proposed for hyperspectral images of LASIS according to its interference hyperspectral properties. Firstly, the hyperspectral image sequence with asymmetric 3D-DWT is decomposed. Secondly, its major hyperspectrum coefficients are scaled up by ROI method to protect the hyperspectral information. Finally, the general 3DSPIHT algorithm is adapted to encode the transformed images. The numerical experiment results show that the PNSR is more than 40 dB at 8∶1 compression rate, and the hyperspectral information can be protected.