Including Spatial Information in Clustering of Multi-Channel Images

A huge number of clustering methods have been applied to many different kinds of data set including multivariate images, such as magnetic resonance images and remote sensing images. However, not many methods include spatial information of the image data. In this tutorial, the major types of clustering techniques are summarized. Particular attention will be devoted to the extension of clustering techniques to take into account both spectral and spatial information of the multivariate image data. General guidelines for the optimal use of these algorithms are given. The application of preand post-processing methods is also discussed.