摘要
研究了多阈值Otsu对血管组织覆盖灰度区间分类算法和基于Hessian矩阵的管状物增强算法。通过多阈值Otsu分类算法对三维人体头部MRA图像进行分类,得到包含少量其它边界过渡区域的脑血管图像,将修正的基于Hessian矩阵的增强算法作用于过渡区灰度数据,使管状物得以保留并增强;将Hessian矩阵的特征值和特征向量应用于血管特征的响应函数及形态学修补和光顺各个环节,与血管灰度接近的片状和点状数据得以去除或削弱;选择合理的尺度空间范围及尺度空间增量和调节因子以平滑非线状区域和锐化增强血管区域。该脑血管提取方法在同等准确率下具有较高的稳定鲁棒性。
An integrated cerebral vascular enhancement method is researched based on multi-threshold Otsu for the organ＇s gray transition interval region classification and multi-scale Hessian feature for the tubular object enhancement. Multi-threshold Otsu algorithm is implemented to get the cerebral vascular relevant gray voxels, and these pixel＇ s geometric characteristics are exploited by Hessian matrix. And Hessian matrix＇ s eigenvalues and vectors forming a tubular object response function are used for further mathematical morphology processing to smooth and verify vessel region. Compared with other tubular object enhancement methods, it behaves higher accurateness with stability robustness.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第5期1709-1712,1749,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61262031)
江西省教育厅科研基金项目(GJJ09212)
江西省研究生创新专项基金项目(YC2012-S081)
