Unsupervised segmentation of hyperspectral remote sensing images with superpixels

In this paper, we propose an unsupervised method for hyperspectral remote sensing image segmentation. The exploits the mean-shift clustering algorithm that takes as input a preliminary superpixels segmentation together with spectral pixel information. proposed does not require number of classes parameter, and it exploit any a-priori knowledge about type land-cover or land-use to be segmented (e.g. water, vegetation, building etc.). Experiments on Salinas, SalinasA, Pavia Center University datasets are carried out. Performance measured in terms normalized mutual information, adjusted Rand index F1-score. Results demonstrate validity comparison state art.