A novel dynamic thresholding and categorizing approach to extract water objects from VHR satellite images

In this study, we propose a new water detection and categorization approach for VHR multispectral satellite images. In order to detect the water regions, the novel Mean Area of Connected Component (MACC) thresholding algorithm is used, which utilizes the structural properties of rivers and lakes. At the same time, the algorithm categorizes the detected waters and refines the threshold value in order to extract narrow rivers. Proposed method is tested on different VHR satellite images obtained from Worldview, Ikonos and Geoeye-1. These extensive experiments demonstrate the robustness of the MACC approach compared to other dynamical thresholding selection approaches.