摘要
提出了一种适合医学图像分割的改进分水岭算法,针对分水岭算法存在的过分割问题,采用数学形态学重构滤波器和Ostu阈值分割方法对图像中感兴趣的目标和背景进行标记。根据标记的二值图像,运用形态学极小值标定技术对原有梯度图像进行修正。最后,使用分水岭算法对修正的梯度图像进行分割。该方法能有效抑制过分割现象,而且计算复杂程度较低、分割效果较好。
A modified watershed algorithm was proposed for medical image segmentation, with regard to the drawback of over-segmentation based on traditional watershed algorithm, a new methodology incorporated morphological reconstruction filters and Ostu threshold segmentation was used to obtain the foreground marker image and background marker image.Then the makers are imposed on the original gradients as its minima.Finally, the watershed algorithm is applied to the modified gradients to segment the image.Compared with other methods, the proposed algorithm can more efficiently reduce the over-segmentation and require less computation.
出处
《长江大学学报（自科版）（上旬）》
CAS
2008年第2期96-98,共3页
JOURNAL OF YANGTZE UNIVERSITY (NATURAL SCIENCE EDITION) SCI ＆ ENG
基金
国家自然科学基金项目(30370393)
关键词
数学形态学
控制标记符
医学图像分割
分水岭
mathematic morphology
marker-controlled
medical image segmentation
watershed
