Effective implementation of maximin clustering for remotely-sensed images

Abstract

Abstract Maximin clustering is widely used for the segmentation of remotely-sensed images. In general, all the clustering methods demand more computational efforts. In order to reduce the image segmentation time some computational logics were used with the Maximin clustering algorithm. They are: (1) the partial sum approach; (2) the pre-calculaled squares approach; and (3) the nearest neighbouring distance (NND) approach. Experimental work is carried out with 4-channel MSS (Multi-Spectral Scanner) data and 6-channel TM (Thematic Mapper) data. The logical combination of all the above logics with Maximin clustering is showing better results. Further, performance of this combined algorithm is observed by varying the number of variables and samples.