Region Similarity Measures for Automatic Detection of Building Roofs from Orthophotoplans

In this paper, an efficient and automatic method for detection of multiple-objects of interest from images is presented. This method is based on using region similarity measures. The method starts by constructing two knowledge databases in which significant and distinctive textures extracted from both objects of interest and background are respectively represented. The proposed procedure continues by an initialization step in which the processed image is segmented into homogeneous regions. In the purpose of separating objects of interest from image background, an evaluation of the similarity between regions of the segmented image and those of the constructed knowledge databases is then performed. The main advantages of this method are simplicity, applicability and suitability. Applying this method on building roof detection from orthophotoplans has shown its robustness and performance.