Abstract: The hyperspectral images compression algorithm based on contourlet transform is proposed, which using multiscale geometric analysis for hyperspectral image space de-correlation. Firstly, wavelet based contourlet transform on each band of hyperspectral images, then predict current band by transform coefficients of the previous band. Finally,SPIHT coder used on prediction error and embedded data stream generated. Experimental results show that the proposed algorithm based on contourlet transform achieve well compression efficient and retain high spectral image texture information better than comparison algorithm.

Key words: Hyperspectral image compression, Multiscale geometric analysis, Contourlet, SPIHT coder