Hyperspectral Image Compression Method Based on Spectral Statistical Correlation

The hyperspectral imaging technology is one of the most important focuses of the remote sensing domain. Research on hyperspectral image compression method has important practical significance. Compared with other traditional remote sensors´ data, hyperspectral images include both spatial and spectral redundancies. Most popular image coding algorithms attempt to transform the image data so that the transformed coefficients are largely uncorrelated. Then these coefficients can be quantized and coded. In many applications, Karhunen-Loeve transform (KLT) is the famous way to decorrelate spectral redundancies. This paper has analyzed the spectral statistical correlation of hyperspectral images and found that the eigenvectors of covariance matrix of hyperspectral images that are composed of the similar objects have almost the same distribution regularity. Based on this regularity this paper advances the approximate KLT theory and realizes approximate KLT (AKLT) + two-dimensional wavelet transformation (2DWT) + two-dimensional set partitioning embedded block (2DSPECK) compression algorithm. Experiment proves that this method is effective.