摘要
Chan-Vese(CV)模型以其能较好地处理图像的模糊边界和复杂的拓扑结构而广泛运用于图像分割中。但由于核磁共振(MR)图像广泛存在强度不均匀性,因此CV模型不仅不能进行准确的分割,而且迭代过程需要对所有图像数据进行反复计算,分割效率很低。针对以上缺点,提出了一种基于局部统计信息的用于快速进行图像分割的CV模型,即首先在局部区域内,通过计算统计量来得到像素点归类的贝叶斯后验概率,并以此作为曲线演化的依据,这样,就能对强度不均匀的MR图像进行准确的分割;然后设置两个表分别存储曲线内外部邻点,并通过更新这两个表内的点来实现曲线演化,从而不但大幅提高了计算速度,并保持了水平集方法能自动处理拓扑结构变化的优点。
Chan-Vese (CV) model, which has good ability to handle the blurry boundary and complex topological structures in images, has been widely used in image segmentations. However, the MR image which has intensity inhomogeneity cannot be segmented accurately by the CV model. And it needs computing all the data of the image during the iterative course. Arming at these disabilities, a fast method of CV model based on statistics in local regions is proposed. First, by calculating the statistics in local regions, Bayesian posterior probabilities that decide which class the pixels belong to are obtained, which are the foundations of the evolution of curve. In this way, the MR images can be segmented accurately. And then two lists are set for storing joint points inside and outside of the curve, and it only need to update the two lists to evolve the curve. By this method, it not only saves lots of time but also preserves the advantages of level set methods, such as the automatic handling of topological changes.
出处
《中国图象图形学报》
CSCD
北大核心
2010年第1期69-74,共6页
Journal of Image and Graphics
基金
江苏省教育厅"青蓝工程"项目(JSK2006018)
关键词
CV模型
贝叶斯后验概率
水平集
MR图像分割
CV model, Bayesian posterior probabilities, level set, MR image segmentation