A class-oriented visualization method for hyperspectral imagery

Currently available hyperspectral image visualization methods can be considered as data-oriented approaches. For such approaches it is difficult to fully satisfy the needs of observers due to the lack of the display of classes. On the other hand, compared to the current methods, demand-oriented or class-oriented hyperspectral visualization approaches show more pertinence and would be more practical. In this paper, using supervised information, a class-oriented hyperspectral color visualization approach based on manifold methods is proposed. The method can simultaneously display data information and class information. First, coarse classification is carried out based on available supervised information. Then, dimensionality reduction is utilized for each category by the use of manifold methods. Then, hue labels are selected in the color space for each category. Finally, output images are visualized after considering the results of the dimensionality reduction and separability. Experiments on real data show that the visualization results by this approach can make full use of supervised information. Also, not only do the output images have a high inter-class separability, but they also have good distance-preserving properties within each class.