SAR image segmentation based on global minimization Chan-Vese model
Bin Zhang · Jisuanji gongcheng yu sheji · 2012
Recently,active contours model has been one of most successful model of image segmentation.But the original Chan-Vese model level set segmentation method produced a lot of false segmentation in SAR image because of strong speckle noise.Therefore,the original Chan-Vese model is improved.Firstly,the nonconvex Chan-Vese model can be reformulated as convex optimization problem,that in turn allowed us to extract a global minmizer of the model.Then edge detector operator is incorporated into convex Chan-Vese model,hybrid model with global minimization based on edge and region information is proposed.Meanwhile,a new iteration terminal condition that is sensitive to variation of evolution contour is proposed,and this can stop the curve evolution automatically under the given rule.Finally,we apply the proposed model to synthetic images and SAR images,and the results prove that the proposed model can extract rapidly and correctly the target areas from SAR images with high robustness.