摘要
针对局部分割方法对初始轮廓敏感的问题,本文提出一种基于模糊主动轮廓的鲁棒局部分割方法.该方法利用图像的局部信息,定义一种新的平均模糊能量函数.通过对演化曲线进行形态学膨胀和腐蚀运算构建窄带,并在窄带范围内求解模糊能量函数的最小值来实现局部分割.为防止演化曲线陷入局部极小值,在迭代过程中加入对比度约束判断条件,进一步提高了分割方法对初始轮廓的鲁棒性.对合成图像和医学图像的分割实验结果表明,与已有的几种局部分割方法相比,本文方法在分割精度和鲁棒性等方面都有较大提高.
Since the local segmentation method is sometimes sensitive to the initial contour, a robust local segmentation method based on average fuzzy-energy based active contour is proposed in this paper. A new average fuzzy energy function is defined by using the local image information. In order to achieve local segmentation, the minimization of energy function is solved in a narrow band constructed by morphological dilation and erosion operations. A contrast constraint condition is added in the iterative process to prevent the curve from falling into local minimum, which can further improve the robustness of the segmentation model against initial contour. Experimental results on synthetic images and medical images show that, compared to several existing local segmentation methods, the proposed method has considerable improvement in segmentation accuracy and robustness.
出处
《自动化学报》
EI
CSCD
北大核心
2017年第4期611-621,共11页
Acta Automatica Sinica
基金
国家自然科学基金(81371635
81671848)
山东省重点研发计划项目(2016GGX101017)
教育部高等学校博士学科点专项科研基金(20120131110062)资助~~
关键词
模糊聚类
主动轮廓
局部分割
隶属度
对比度约束
Fuzzy clustering, active contour, local segmentation, membership, contrast constraint
