Segmenting SAR Satellite Images With the Multilayer Level Set Approach

Synthetic aperture radar (SAR) provides a remote sensing method to explore the ground truth in all types of weather conditions. However, interpretation of SAR images is difficult because of the effects of speckle signals in the images. One method, the multilayer level set approach, segments an entire given SAR image into several sub-regions such that the segmented regions are homogeneous. This method employs two implicit functions with pre-selected level values. Based on the fact that the segmented regions are homogeneous and presented as regional constants, the energy defined by the segmented regions and their corresponding regional boundaries is minimized such that the relationships between the defined energy and the implicit functions can be transformed into the relationships between the implicit functions and time. By implementing the algorithm in terms of finite difference, this method offers an efficient and stable approach to a numerical solution. By increasing iterations and preselected level values, the implicit functions evolve close to the regional boundaries based on the energy minimization. From the processed results, the multilayer level set approach can efficiently segment the given SAR images, allowing further image interpretation to reveal the ground truth in the imaged areas.