摘要:
针对医学图像中存在的灰度不均匀现象,提出一种变分水平集分割模型。将邻域信息引入到基于Bayes决策准则的水平集分割框架中,以增强灰度不均匀条件下弱目标边界的识别能力。为缓解模型求解时易出现的局部极小问题,设计一种简单而有效的随机优化方法以搜索近似全局最优解。实验对比及分析验证了该水平集分割模型在多种灰度不均匀场景下均表现出较好的分割性能。
关键词:
变分水平集,
灰度不均匀,
医学图像分割,
曲线演化
Abstract:
To effectively perform medical images segmentation in the case of intensity inhomogeneity, this paper presents a novel variational level set method. Statistical features of object neighborhood are integrated into a level set segmentation framework based on Bayes decision principle, in order to increase the ability to capture weak border of objects. In addition, a simple and efficient stochastic optimization scheme is designed to search an approximately global optimal solution using global information of the segmented object. The purpose is to alleviate the problem of local minimum caused by gradient descent solution used in level set method. Experiments and contrast analysis evidently demonstrate robustness, accuracy of the proposed model in the case of intensity inhomogeneity scenes.
Key words:
variational level set,
intensity inhomogeneity,
medical image segmentation,
curve evolution
中图分类号:
林颖, 印桂生, 杨耘. 基于变分水平集的灰度不均匀医学图像分割[J]. 计算机工程, 2010, 36(24): 203-205.
LIN Ying, YI Gui-Sheng, YANG Yun. Segmentation of Medical Images with Intensity Inhomogeneity Based on Variational Level Set[J]. Computer Engineering, 2010, 36(24): 203-205.