Evaluation of Unsupervised Clustering Methods on Hyperspectral Image Data Sets

Classification and clustering of hyper spectral remote sensing images are keys to extract abundant information. Researchers have developed several popular clustering algorithms in the past years, and many learning based methods have been developed nowadays. Conventional clustering algorithms showed good performance with low dimensional data, like RGB images and database records. In hyperspectral image clustering, methods are usually composed of two steps, feature extraction and conventional clustering. This paper attempted to evaluate performances of conventional clustering methods based hyperspectral images without any feature extraction step.