A variational method using Riemannian metric for SAR image segmentation
Na Li, Fang Liu, Jing Li, Junfeng Yang, Tuanjie Zheng · 2017
This paper presents a variational method for SAR image segmentation that unifies boundary- and region-based information into the geometric active contour model. A new Riemannian metric is introduced to construct region-based energy function derived by maximizing the geodesic distance of a new Riemannian metric in differential-geometric structure for spectral density functions. A heterogeneity indicator map is proposed for SAR images and incorporated into the boundary-based energy function. The curves are propagated toward the result under the influence of the boundary- and region-based forces. The experiment results on TerraSAR-X images confirm its effectiveness.