Water Boundary Detection for SAR Images Using Region-Based Active Contours with Salient Edge Energy

Detecting water boundaries in synthetic aperture radar (SAR) images remains challenging in the field of boundary detection. To tackle this issue, a new region-based active contour model with salient edge energy is proposed. First, the mid-values of pixel gray scales in the inner and outer regions are employed as region fitting centers, which can suppress the influence of interference pixels. Second, a novel region-based signed pressure force (RSPF) based on the above mid-value region fitting centers is developed and used to construct the objective function. Third, a salient edge energy term is incorporated into the above objective function to improve the model’s segmentation performance. Finally, some regularization terms are also added to keep the stability of the model. Experimental results for real SAR images verify that the proposed model can detect water boundaries more effectively and efficiently compared with state-of-the-art models, in particular, its detection accuracies are higher than 0.95 and detection false alarms are smaller than 0.2. In addition, its curve initialization is free.