A fuzzy graph-based segmentation for marine and maritime applications in SAR images

In the context of sea monitoring, an important processing step is image segmentation. In this paper, the authors propose a new segmentation method that combines fuzzy and graph-based theories. The algorithm, starting from a single source element belonging to the region of interest, proceeds with a propagation mechanism that aims at finding a Minimum Path Spanning Tree (MPST). The process is automatic, unsupervised, adaptive to the image content and independent from the order of analysis. A TerraSAR-X image and Cosmo-SkyMed images are used for the experiments. The considered applications are oil spill detection, sea surface analysis, and ship detection.