Synthetic aperture radar image segmentation based on edge-region active contour model

An energy functional is proposed based on an edge-region active contour model for synthetic aperture radar (SAR) image segmentation. The proposed energy functional not only has a desirable property to process inhomogeneous regions in SAR images, but also shows satisfactory convergence speed. Our proposed energy functional consists of two main energy terms: an edge-region term and a regularization term. The edge-region term is derived from a Gamma model and gradient term model, which can process the speckle noises and drive the motion of the curves toward desired locations. The regularization term is not only able to maintain a desired shape of the evolution curves but also has a strong smoothing curve effect and avoid the occurrence of small, isolated regions in the final segmentation. Finally, the gradient descent flow method is introduced for minimizing our energy functional. A desirable feature of the proposed method is that it is not sensitive to the contour initialization. Compared with other methods, experimental results show that the proposed approach has promising edge detection results on the synthetic and real SAR images.