A GEOMETRIC ACTIVE CONTOUR MODEL USING SYMMETRICAL KULLBACK-LEIBLER DISTANCE FOR SAR IMAGE SEGMENTATION

摘要

This paper addresses the problem of SAR image segmentation by using a variational method. The resistor-average distance which is a symmetrical Kullback-Leibler distance is utilized to measure the discrepancy between the distributions of the classes to be segmented. The regional energy is derived by maximizing the discrepancy from the entire shapes of the distributions of the segmented classes. The gradient flow of the regional energy is derived accordingly. The heterogeneity indicator map, computed from the coefficient of variation, assists to locate the boundary of the segmented objects. The experimental results on SAR images confirm its effectiveness and flexibility.