A variational level set approach for automatic target extraction of SAR images
Zongjie Cao, Ling-Li Pang, Yi-ming Pi · 2007
A variational level set approach for automatic target extraction of SAR images is presented in this paper. Target extraction of SAR image is a hard work due to the presence of speckle noise. The proposed approach defines an energy criterion that consists of a region term derived from a mixed Gamma model of speckle noise and a boundary term related to the image gradient. Both of the terms are directly defined on the level set function. It is obviously different from the energy functional defined on parameterized curve in early level set approach. Target extraction is implemented by minimizing the energy criterion via variational level set approach. The performance of the approach is verified by MSTAR data. The experimental results show that the approach takes account of the speckle noise effect and sufficiently uses the boundary information of SAR images, thus accurately extracts the target but does not need speckle pre processing step.