A Novel Unsupervised Subpixel Mapping Algorithm to Enhance Accuracy of Remotely Sensed Image

The division of an image into meaningful structures, image segmentation, is often an essential step in image analysis, object representation, visualization, and many other image processing tasks. In this paper the multi-spectral satellite image is taken as input. Our aim is to enhance the accuracy and to remove the over segmentation..Mean-shift technique have been demonstrated to be capable of estimating the local density gradients of similar image pixels.Fuzzy c-means has been shown to work well for clustering based segmentation, however due to its iterative nature this approach has excessive computational requirements.By incorporating the unsupervised unmixing criterion of the FCM clustering algorithm and the maximal land cover spatial-dependence principle, the proposed usFCM_SPM can generate a subpixel land cover map without any prior endmember information. Therefore in order to remove the over segmentation the Minimum Spanning Tree (MST) clustering algorithm will cluster the objects in the feature space and eliminate the over segmentation. In this research work two important clustering algorithms namely modified mean shift and usFCM_SPM are compared. These algorithms are applied and performance is evaluated on the basis of the efficiency of clustering output.Experimental results on a multispectral image demonstrate that the usFCM_SPM clustering method exhibits better accuracy than modified mean shift algorithm.