Band selection for hyperspectral images based on self-tuning spectral clustering

Hyperspectral imaging (HSI) is an emerging technique,which allows to consistently capture images in the visible as well as infrared light range. Many materials can be easily discriminated by means of their spectra, rendering HSI an interesting method for the reliable classification of contents in a scene. As the number of features for each pixel in hyperspectral images is considerably high, further processing and classification is time consuming and stresses resources. Thus, efficient methods to select useful bands are required. We present a novel two-step scheme based on a clustering approach followed by representatives selection from each cluster. The classification results of real hyperspectral images demonstrate that the proposed method easily outperforms common as well as state-of-the-art methods.