Color Representation and Classification for Hyperspectral Imagery

A hyperspectral image contains information in hundreds of spectral channels. The information is sparsely distributed in such a huge 3D image cube. In practical applications, it may be helpful if such a high dimensional data can be displayed into an informative color image. This 3-band color image provides a vivid and quick overview of existing objects/materials and their distribution in the image scene. Condensing the data from the original high dimensional space into such a low dimensional space will lose information. Keeping this in mind, our objective is to display the useful information contained in an original data as much as possible and as distinctively as possible. Our display approaches are based on the techniques in the class of principal components analysis (PCA). The generated color image is evaluated using a detection algorithm, and the result is compared with the one using the original data.