Improved principal component analysis based hyperspectral image compression method

Due to the huge amount of hyperspectral image (HSI), the compression of HSI is an important topic in remote sensing. The state of art methods use principal component analysis ( PCA ) as spectral decorrelatior and wavelet transformation as spatial decorrelator. How to determine the number of principal component and how to allocate storage of every principal component are two problems not solved. In this paper, we use hyperspectral signal subspace identification by minimum error ( HySime ) algorithm to estimate the the number of principal component and eigenvalue as the metric to allocate the storage. The experimental results demonstrate that the proposed algorithms can get better results than traditional PCA based methods.