A new dynamic thresholding method for detection of water regions in multispectral VHR images

In this study, a novel water area detection method is presented for high resolution multispectral satellite images. Proposed algorithm automatically detects a threshold value on the NFSE image using the newly proposed Mean Area of Connected Component (MACC) method. By utilizing this threshold, a rough water mask is extracted from the image. Proposed method is tested on 20 different VHR satellite images obtained from Worldview, Ikonos and Geoeye-1. These extensive experiments show the robustness of the MACC approach and competitive results with state-of-the-art are obtained.