Research And Implementation Of Pca Based Hyperspectral Image Compression Algorithm

With the development of science and technology, the use of hyperspectral image has become more and more widely. It has important practical value to compress the hyperspectral image effectively. As it has a strong spectral correlation, the hyperspectral image is compressed using principal component analysis (PCA) in this paper. Different from the traditional transformation based method, we use PCA and2DPCA for band selection. And then the selected bands are processed by prediction and entropy coding. In order to avoid the loss of local important information, we group bands into different clusters using the correlation coefficient between them. Then take band selection for each sub image separately. Thus we realize a segmented PCA based and a segmented2DPCA based band selection algorithm of hyperspectral image.Experimental results prove that the PCA based band selection compression method is not only need a small amount of calculation but also can keep the physical properties of original image effectively, which has positive significance for the follow-up processing2DPCA is faster than PCA, and the bands extracted by2DPCA are more representative. Segmented PCA and segmented2DPCA can extract important bands from a broader range with a shorter time than PCA. The bands selected by2DPCA contain more information of the original image. Its overall compression ratio is15.49:1.