Satellite image segmentation with Shadowed C-Means

Satellite images often require segmentation in the presence of uncertainty, caused due to factors like environmental conditions, poor resolution and poor illumination. Since any subsequent image analysis depends on the quality of such segmentation, one has to obtain an efficient algorithm for the purpose. Pixel clustering is a popular way of determining the homogeneous image regions, corresponding to the different land cover types, based on their spectral properties. In this paper we map the newly developed shadowed clustering algorithm to the problem of segmenting remotely sensed images. It is observed that shadowed clustering can efficiently handle overlapping among segments while modeling uncertainty among the boundaries. Unlike rough clustering, here the choice of user-defined parameters is fully eliminated. The number of segments is automatically optimized in terms of validity indices. The algorithm is robust in the presence of outliers. The superiority of the system is demonstrated in segmenting a synthetic image, along with land cover types from the Indian Remote Sensing (IRS) images of the cities of Mumbai and Kolkata and the SPOT image around Kolkata. The algorithm is found to efficiently and accurately extract the different homogeneous regions in the presence of uncertainty. The results are analyzed both qualitatively and quantitatively.