An Isotropic Minimal Path Based Framework for Segmentation and Quantification of Vascular Networks
An Isotropic Minimal Path Based Framework for Segmentation and Quantification of Vascular Networks
复制标题
DOI：
10.1007/978-3-319-78199-0_33
复制
发表时间：
2017-10
期刊：
影响因子：
--
通讯作者：
E. Cohen;L. Cohen;T. Deffieux;M. Tanter
中科院分区：
文献类型：
--
作者：
E. Cohen;L. Cohen;T. Deffieux;M. Tanter
图像分析的最小路径方法旨在提取最小化能量泛函的曲线。路径的能量对应于根据相关度量函数的其加权曲线长度。在这项研究中，我们设计了一个二进制各向同性度量模型与使用基于海森的血管增强滤波器，以提取血管网络的几何特征。我们介绍了一种约束关键点搜索方法，能够提取亚像素血管中心线，直径和分叉。视网膜图像上的实验表明，该框架实现了类似的，甚至更好的分割性能相比，使用更复杂的度量设计的方法。
Minimal path approaches for image analysis aim to extract curves minimizing an energy functional. The energy of a path corresponds to its weighted curve length according to a relevant metric function. In this study, we design a binary isotropic metric model with the use of a Hessian-based vascular enhancement filter in order to extract geometrical features from vascular networks. We introduce a constrained keypoint search method able to extract subpixel vessel centrelines, diameters and bifurcations. Experiments on retinal images demonstrated that the proposed framework achieves similar even better segmentation performances as compared with methods using more sophisticated metric designs.