Coastline Detection in SAR Images Using a Hierarchical Level Set Segmentation

This paper proposes a new coastline detection method for both single-polarization and quad-polarization synthetic aperture radar (SAR) images. Unlike other coarse-to-fine methods using a combination of two boundary extraction techniques in two layers, the proposed method extracts coastlines by a hierarchical structure from coarse one to fine one using level set and narrow-band level set techniques. The number of layers in the hierarchical structure is adaptive and controlled by an offset analysis of the coarse result. In the method, coarse coastline is first extracted by the level set algorithm with filtered data because the scattering of land and ocean are varied. A multilayer level set segmentation, performed in a narrow band, is then applied to gradually locate more accurate coastlines to decrease the offset in the coarse result caused by the smoothing processing. Finally, the determination of the structure parameters of the proposed method is analyzed through the calculated offset. A number of real polarimetric SAR images are used for validation. Experimental results show that the proposed method is more accurate and is less sensitive to the initial parameters than the other typical method based on level set with filtered or unfiltered data. In addition, the experimental results also show that the proposed method is robust and accurate for data of different microwave bands and different coastal environments.