Investigation Of Hyperspectral Image Compression

In some industrial applications, high-quality color information is always required in display, reproduction, and quality control. As the traditional trichromatic system can't satisfy the required accuracy, the technique of hyperspectral image (HSI) is always needed in these applications. However, the data volume of multispectral images is so huge that it makes its storage and communication very difficult. Therefore, it is necessary to compress the HSI data. In this study, the HSI compression is a kind of lossless or near-lossless compression, and its target is to obtain appropriate compression ratio while keep required color accuracy.In the beginning, this thesis presents the fundamentals of image compression, color theory, and the state of art of HSI compression techniques. Then, the investigation of HSI compression was conducted with the consideration of the spectral redundancy and spatial redundancy of HSI data. The spectral redundancy in HSI is removed by principle component analysis (PCA), and the spatial redundancy is removed by JPEG2000. We use the proposed technique to compress the several textile HSIs. The experimental results show that, when the compression parameters are appropriately chosen, the performance of the HSI compression is quite efficient. In addition, this thesis also presents methods in improving compression speed and guiding compression parameter selection.