SAR satellite image interpretation based on the multilayer level set approach

Synthetic aperture radar, SAR, is a remote sensing way to explore the ground truth in day and night. How to interpreter the given SAR images provides an import clue to study the characteristics of the imaged areas. However, the image interpretations for SAR images are difficult because of the effects of speckle signals shown in the images. In order to solve the problem, several algorithms have been proposed. The processed results show the proposed algorithms still have their own limits on reducing the effect of speckle signals. In this paper, the multilayer level set approach is employed to have the SAR images be grouped into several sub-regions such that the segmented regions are homogeneous. Based on this minimization of the energy, the multilayer level set method implicitly presents the regional boundaries as several nested level lines. By increasing iterations and preselected level values, these lines evolve close to the level boundaries based on the energy minimization. This method provides numerical stability and quick convergence. In order to implement the multilayer level set approach, several level values need to be established firstly. Those level values are determined with calculating the average values from the classified groups with applying K-means method. Based on the four-color theory, the multilayer level set method is able to generate an optimal piecewise continuous approximation for the SAR image such that each approximation sub-region is homogeneous. From the processed results, the multilayer level set approach can efficiently reduce the effect of the speckle signals, and quickly segment SAR images for further image interpretation.