摘要
在白细胞图像中,由于白细胞细胞核的存在,直接应用分水岭算法时,往往达不到较好的效果。本文提出一种结合EM聚类的改进分水岭算法。通过将EM聚类获得的图像中细胞核区域替换,然后使用基于距离变换的分水岭分割,确定白细胞区域。对距离变换后的图像采用形态学处理减少了细胞分割中的过分割现象。同时使用细胞核位置的先验条件,合并分水岭分割区域,进一步减小过分割的影响。本文方法提供一种新的将分水岭算法应用于白细胞分割的思路。同时实验证明,方法在分割精度上有着良好的表现。
In white blood cell image, the direct application of watershed algorithm in the gray scale images containing leukocytes often can not reach better results due to the presence of nuclear. So this paper proposed a new algorithm combining the EM clustering and watershed. The nuclear regions gotten by EM clustering are replaced, and then watershed seg- mentation based distance transform is used to determine the leukocyte region. The use of morphological processing in the images after distance transform reduces the over-segmentation. Moreover, prior conditions of nuclear position can merge watershed segmentation regions, and reduce the impact of over segmentation. The experiments show that the method has a good accuracy in the segmentation performance.
出处
《计算技术与自动化》
2016年第3期81-84,共4页
Computing Technology and Automation
基金
"十二五"国家科技支撑计划项目(子项目)(2012BAK06B04)
湖南省自然科学基金项目(14JJ2074)
关键词
白细胞
图像分割
EM聚类
分水岭变换
white blood cell
image segmentation
EM clustering
watershed
