Building detection using local Gabor features in very high resolution satellite images

Building detection from very high resolution satellite imagery is an important task for land planners. However, manually locating buildings from these images is a difficult and time consuming process. Therefore, researchers focused on building detection using automated image processing and computer vision techniques. The main problems here are as follows. Buildings have diverse characteristics and their appearance (illumination, viewing angle, etc.) is uncontrolled. On the other hand, buildings often have similar cues like parallel edges and roof corners that can be merged. In this study, we propose an automated approach for building detection based on Gabor filters and spatial voting. We extract features (representing buildings) using Gabor filter responses. Using these features, we form a spatial voting matrix to detect buildings. We tested our algorithm on very high resolution grayscale Ikonos satellite images and obtained promising results.