Significance of texture features in the segmentation of remotely sensed images

This paper proposes an efficient segmentation algorithm and the significance of texture features over segmentation in multispectral satellite images. In order to preserve the complete details about the earth's surface, initially texture features are extracted from the multispectral satellite images. Texture features are very important in remotely sensed images, since it has the capability to interpret micro and macro level features present in the images. This method has been effectively tested over multi-temporal LANDSAT data set of Hanoi and Balcoc region of vietnam. Qualitative observations made in the proposed work suggest that the addition of DOOG (Difference of Offset Gaussian) filter and texture features provide improved segmentation results compared to traditional methods. As per the quantitative investigations, the proposed method outperforms other contemporary methods in terms of Random Index (RI), Variance of Information (VOI) and Global Consistency Error (GCE).