基于等周图割的细胞图像自适应多阈值分割算法  

Automatic Multilevel Threshold Cell Image Segmentation Algorithm Based on Isoperimetric Graph Cuts

作　　者：张瑞华[1] 付云飞[1] ZHANG Rui-hua;FU Yun-fei(School of Physics and Information Engineering,Jianghan University,Wuhan 430000,China)

机构地区：[1]江汉大学物理与信息工程学院,武汉430000

出　　处：《内蒙古大学学报（自然科学版）》2018年第5期534-539,共6页Journal of Inner Mongolia University：Natural Science Edition

基　　金：国家自然科学基金(61575085);湖北省教育厅教研项目(2016281);武汉市教育局教研项目(2017086)

摘　　要：针对粘连细胞图像,提出一种自适应多阈值分割算法.该算法采用等周图割的等周率来确定最佳阈值,基于最佳组距迭代策略用多级阈值来分割粘连细胞,此外,引入动态确定节点聚类数的方法来自动确定阈值个数,采用灰度级权值矩阵来加快等周率的计算.实验结果表明,与几种经典阈值分割算法和等周率算法相比,该算法对严重粘连的多细胞图像的分割效果更好,在复杂的细胞图像分割中具备一定的有效性和可行性.For the adherent cells,automatic multi threshold segmentation algorithm based on isoperimetric graph cuts is proposed.The optimal threshold is determined by the isoperimetric rate of the isoperimetric graph and the adhesive cells are segmented based on the optimal interval iteration strategy using multilevel thresholds.In addition,the method of dynamic determining the number of the node cluster is introduced to determine the threshold number automatically.The computing speed is accelerated by the grayscale weight matrix.Experimental results show that the new algorithm performs better compared with several classical threshold segmentation algorithms.The new algorithm has certain validity and feasibility in the complex cell image segmentation.

关 键 词：图像处理 图像分割 细胞粘连 谱图 阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP37[自动化与计算机技术—计算机科学与技术]