ABSTRACT
ABSTRACT: The problem of 2-D building extraction from high resolution satellite images has been addressed in this correspondence. Research in the domain of building detection and/or extraction has been going on for a long time, yet automatic building extraction has remained a difficult problem due to variations in the spectral and spatial image properties, presence of noise in the image etc. A novel object based automatic unsupervised method for extracting building regions from high resolution satellite images incorporating geometrical features of the region has been proposed here. The method is object based in the sense that the image is first segmented and the processing is performed thereafter at the object level. Experimental results show a higher accuracy of the proposed method in detecting rectangular or near rectangular shaped buildings from the image.
