Hyperspectral Image Segmentation Method Based on Kernel Method

Hyperspectral image Contains rich information, so improving the classification ability of hyperspectral image has become a hot spot of research recently. Image segmentation is one of the most basic and important problems of the image processing and low level computer vision and the precondition of the image processing, some hyperspectral image classification method based on image segmentation. In this study, an image segmentation method based on kernel methods was proposed. First reduce the dimension of hyperspectral images by KPCA, then the k-means method are used to cluster, finally finish the image segmentation in the high-dimensional space by gaussian kernel mapping. In simulation experiment, changing the parameters of the experiment and comparing with standard hyperspectral image segmentation, the results show that the method is good enough in image segmentation, which can be used for visual analysis and pattern recognition, and realize hyperspectral image segmentation.