一种结合GVF和CV模型的水平集图像分割方法  

A Level Set Image Segmentation Method Combined with GVF and CV Model

作　　者：胡小为[1] 刘宏申[1] 徐国雄[1] 阮越[1] 刘恒[1] 潘祥[1] 

机构地区：[1]安徽工业大学计算机科学与技术学院,安徽马鞍山243032

出　　处：《安徽工业大学学报（自然科学版）》2016年第3期289-294,共6页Journal of Anhui University of Technology（Natural Science）

基　　金：安徽省自然科学基金项目(1308085QF113)

摘　　要：由于CV(Chan-Vese)模型是一个非凸性泛函,对该泛函求极值只能得到局部最优解,运用该模型进行图像分割时,很难在全局范围内得到理想的结果。鉴于此,提出一种结合梯度矢量流(gradient vector flow,GVF)和CV模型的水平集图像分割方法。该方法通过GVF将边缘梯度信息扩散至整幅图像,在保留CV模型基本优点的同时,融入GVF的全局性梯度信息,从而引导CV模型在全局范围内演化至准确的目标边缘。实验结果表明,该方法的分割效果和收敛速度均明显优于传统CV模型。Owning to the non-convex functional of with the Chan-Vese（CV） model, one can only obtain a localoptimal solution. It is difficult to achieve an ideal result for image segmentation in the global range. Therefore anew level set based image segmentation method that combining CV model and gradient vector flow（GVF） wasproposed. The edge gradient information is spreaded to the entire image with GVF, which guides the evolution ofCV model to the correct target edge in the global range and retains the basic advantages of CV model. Theexperimental results indicate that the present method are obviously better than the traditional CV model.

关 键 词：图像分割 水平集方法 CV模型 梯度矢量流 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]