Design and Implementation of Frame Work for Building Extraction using Fast Accurate Bilateral Filter, Fast Corner Method and Improved Gaussian Blur

Identifying and detection of buildings from satellite imagery is beneficial in mapping, environmental training, disaster management, city planning, geo references, and studies purposes. In this paper first, Gaussian mask is used in pre-processing to avoid blurring effects in computer vision to enhance image structures at different scales then Gaussian blur smoothens irregular pixel values by excising the extreme outliers. With fast accurate bilateral filter, quality of edge preserves in buildings has been smoothened, Local key point’s extraction, dominant directions and weights were extracted to detect corners of different shaped buildings in image very quickly. Based upon weightage and correct directions in voting, Transforms applied for optimum results. We processed satellite images taken from Mumbai, Chennai cities, and Nagapattinam town and rural of Nannilam in Tamilnadu, India. Kernel density distribution become extensively utilized to find the density flow for metropolitan cities also. In results, buildings detection without shadow in 28 to 35 seconds in cities and 5 to 8 seconds in urban areas and improved accuracy of 40 % in metro and 8 % in urban than existing detection techniques.