Geometrically guided Fuzzy C-Means clustering of multispectral images

Fuzzy C-means (FCM) is an unsupervised clustering technique and is often used for the unsupervised segmentation of multivariate images. The segmentation is based on spectral information only and geometrical relationship between neighbouring pixels is not used. In this paper, a semi-supervised FCM technique is used to add geometrical information during clustering. Geometrical information can be adapted from the local neighbourhood, or from a more extended shape, model such as the hough circle detection. Segmentation experiments with the Geometrically Guided FCM (GG-FCM) show improved segmentation above traditional FCM such as more homogeneous regions and less spurious pixels.