Automatic building extraction is a suitable tool for sustainable development in the field of urban development and post disaster-mitigation. Laser scanning data, which provides the height of the ground objects can be used for developing models to extract the manmade features in a complex urban environment. The height variation along the boundary of trees is more significant than buildings, this parameter can be used for their classification. The Standard Deviation (SD) is a measure to find the variation, Modified Standard Deviation (MSD) method is applied for this purpose. Using MSD method SD was determined with respect to a best fit line in the local domain i.e. calculating MSD of a small number of consecutive boundary pixels at a time. Since the MSD method is a pixel based classification, so the tree part of a mixed segment can be removed in this process, which is a difficult task in the polygon based classification. Some of the tree parts were remained in the result due to their less height variation. But these can be further removed by analyzing the shape of the resulted polygons. MSD method was also applied to the high-resolution spectral data. It was found that the spectral variations of the object boundaries are not same as their height variations in Laser data. So only shape analysis was applied to the extracted polygons of the spectral data for extraction of manmade objects. The present study shows that on the basis of shape, buildings and manmade features can be extracted reliably from spectral data.