Efficient Wavelet-Based Compression for Hyperspectral Images Using Precompression Algorithm

Since it´s difficult for DWT-based compression scheme to efficiently and directly utilize the various properties of hyperspectral images, such as their statistic characteristics, spectral correlation, and spatial correlation, a new framework of DWT-based compression is proposed and a so-called precompression process is built and introduced into it. Firstly, general relation between bands is deduced in accordance with the high spectral correlation. Based on the relation, precompression algorithm is designed to segment bands of the images into groups and remove spectral correlation and spatial correlation within each group by extremely few bits consumption. Afterwards, algorithm based on 3D packet DWT is adopted subsequently for further compression. Experiments show that the proposed schemes derived from the new DWT-based framework can significantly improve performance compared with the best known techniques, especially at low bitrates. And the new framework has excellent extensibility.