摘要
目的:为了抑制分水岭算法的过分割问题,提出一种适合MRI脑肿瘤分割的分水岭算法的改进方法。方法:使用交互的方式对脑肿瘤目标区域进行前景标记,使用形态学滤波器与区域生长算法相结合的方法对脑肿瘤目标的背景区域进行动态背景标记,最终使用肿瘤目标的标记和背景标记合成的标记图像与原图像进行分水岭算法分割。结果:当形态学滤波器的模板尺寸大小为14个像素时,脑肿瘤背景区域的标记效果较好且算法稳定,亦能有效地抑制过分割现象。结论:该方法操作简便,分割效果好,过分割的抑制效果明显。
Objective To improve the watershed algorithm against the over segmentation in order to make it suitable for MRI image segmentation of the brain tumors.Methods The foreground marking of the brain tumor target area was performed with an interactive method,and the morphological filter and region growing algorithm were used to realize dynamic marking of the background area,then the marking images and the original ones underwent watershed segmentation.Results The marking of the background behaved well and stably when the template size of the morphological filter was 14 pixels,with the over segmentation compressed.Conclusion The improved algorithm functions well in the segmentation and the compression of the over segmentation,and is easy to use.
出处
《医疗卫生装备》
CAS
2013年第7期12-14,共3页
Chinese Medical Equipment Journal
基金
国家863重大项目(2011AA040406)
国家自然科学基金(61271119)
广西自然科学基金(2011jjA40078)
关键词
形态学重构
分水岭
标记
分割
morphological reconstruction
watershed
marking
segmentation
