The Novel Compression Method Of Hyperspectral Image Based On The Wavelet Transformation

There is very abundant spatial and spectral information in the three dimension hyperspectral data. With the technical development of information processing, the spectral band numbers of the imaging spectrometer have been increased and the spectral resolution of the object has been improved. The hyperspectral data has been playing an important role in the application of the remote sensing.In this dissertation, the application of KL transformation and wavelet transformation to hyperspectral images compression is studied based on analyzing the characteristic of the hyperspectral images. The algorithm of the advanced KLT/IWT/SPIHT is implemented in the dissertation. Used the characteristic the hyperspectral images, the algorithms of the 32-PKLT/IWT/SPIHT and the Auto-adapt PKLT/IWT/SPIHT are proposed. In order to compare with the performance of the compression methods mentioned above, the 3D IWT/3D SPIHT is realized. The main research work is followed:Firstly, the transformation methods and encoding methods are introduced. KL transformation, the second generation wavelet transformation and the encoding methods based on the wavelet transformation is presented. Then the compression algorithm of the advanced KL transformation/integer wavelet transformation/SPIHT (KLT/IWT/SPIHT) is proposed. The advanced KLT is used to remove the spectral redundancy of the hyperspectral images at first. Then the 9/7 IWT is used to eliminate the spatial redundancy of the hyperspectral images. At last, the SPIHT method is used to compress the hyperspectral data after KLT and IWT.Secondly, through the analysis of the characteristic of the hyperspectral images, it is found the hyperspectral image has weaker spatial correlation and stronger spectral correlation. It is proposed that the spectral redundancy should be eliminated in the compression of the hyperspectral images.Finally, combined with the conclusion of the characteristic the hyperspectral images, the equal spectral band partition or the auto-adapt spectral band partition before the advanced KL transformation is propounded. The algorithms of the advanced KLT/IWT/SPIHT, 32-PKLT/IWT/SPIHT, the Auto-adapt PKLT/IWT /SPIHT and the 3d IWT/3d SPIHT are designed. The four algorithms mentioned above has been compared in the compression performance and the computing time under the different bit ratios, with the hyperspectral cube of 128 columns by 128 rows by 224 bands which is treated as the 16-bit unsigned data.The experiment results show that when the bit ratio is greater than 0.8bpp, the performance of the Auto-adapt advanced KLT/IWT/SPIHT algorithm is outperformed the others, while the 3d IWT/3d SPIHT algorithm can be run faster than the others.