摘要： 基于核密度估计的活动轮廓模型如果没有适当的扰动机制，往往不能在弧度突变的边缘上获得较好的收敛结果，且在大噪声环境下鲁棒性较差。针对该问题，提出一个新的代价函数。该函数通过融合边缘映射的曲率信息，改善原算法在突变边缘的收敛效果，降低算法对初始轮廓的依赖。

关键词: 活动轮廓模型, 图像分割, 核密度估计, 非参数方法, Snake模型

Abstract: If active contour model based on Kernel Density Estimation(KDE) has not proper interruption method, it is hard to obtain a desirable result on the edge changed violently, and its robustness is bad under big noise environment. In order to solve the problem, this paper proposes a new cost function. By combining the curvature information of edge mapping, it improves the convergence effect of the exsisted algorithm on break edge, and decreases its dependence on initial contour.

Key words: active contour model, image segmentation, Kernel Density Estimation(KDE), nonparametric method, Snake model

中图分类号: 

• TN911.73