摘要
提出一种新的交互式图像分割方法并应用于骨髓瘤细胞克隆斑图像分割。从图分割的理论出发,首先由用户选定部分像素作为前景对象和背景,其他部分为未知区域,然后构造能量函数,通过K均值初始化函数,进行迭代使能量函数最小化,最后从形态学的角度进行腐蚀、膨胀处理,有效地将感兴趣的克隆斑部分分割出来。对于分割错误的部分,利用类似画笔的工具重新标记为前景或者背景,然后再次进行分割。试验结果表明,不论采用主观评价方法,或采用图像分割质量判断准则进行评价,分割结果都较为满意。
A new interactive image segmentation method used in the multiple myeloma cloning spots image segmentation is presented in the paper.Based on the theory of graph cuts,some pixels are selected as the front object and the background seeds,and the other parts are treated as the unknown region.Then,an energy function is constructed and initialized through K-means,and the minimum cut method is used in the segmentation by energy minimization. Last,the image is eroded and dilated,and the cloning separate parts coul...
出处
《中国医疗器械杂志》
CAS
2008年第5期319-322,共4页
Chinese Journal of Medical Instrumentation
基金
上海市重点学科建设项目(T0102)资助。
关键词
骨髓瘤细胞克隆斑
图分割
K均值
multiple myeloma cloning spot
graph segmentation
K-means