Abstract: Aiming at handling the difficulty of storage and transmission caused by the large volume of data, a hyperspectral image compression algorithm was proposed, which combines adaptive band regrouping and bit rate prcallocation.The bands regrouping preprocessing based on affinity propagation clustering and reference frame selection were utilized to exploit spectrum correlation sufficiently. The interband prediction was applied to dccorrelate the spectrum redundancy in different groups while bit ratio prcallocation was utilized for intraband compression according to the information of prediction residual. Experimental results show that the proposed approach achieves a good performance in quality and complexity, the average peak signal to noise ratio (PSNR) is higherthan the state of the art algorithms.

Key words: Hyperspectral image compression, Band regrouping, Affinity propagation clustering, Interband prediction