Unsupervised land-cover classification of interferometric SAR images

The current study evaluates an unsupervised segmentation method called fuzzy C-means for two multi-temporal interferometric SAR image datasets. Noise removal filters and a principal components transformation was carried out as a pre-processing step to reduce noise and input data amount. The final segmented images were optimally clustered into 2-3 classes/clusters where the classes are water-bodies, forested and non-forested areas. The actual classification accuracy has to be better validated with more up-to-date maps, but it seems to very promising for the two datasets respectively. The presented method is believed to be good for detection of forested areas, water-bodies and non-forested areas.