一种自适应宫颈细胞显微图像分割算法  

An adaptive cervical cell microscopic image segmentation algorithm

作　　者：赵英红[1,2] 胡磊[1] 孙存杰[1,2] 平杰[2] 

机构地区：[1]徐州医学院影像学院,江苏徐州221004 [2]徐州医学院附属医院,江苏徐州221004

出　　处：《军事医学》2012年第12期946-949,共4页Military Medical Sciences

摘　　要：目的提出一种有效算法应用于宫颈细胞显微图像的分割,实现细胞核、细胞质与背景的分离。方法采用最大类方差自适应阈值法将细胞整体从背景中分离出来,再利用最佳直方图熵法与改进的遗传算法相结合,将图像的总熵的计算作为适应度函数,结合遗传算法的寻优特性,实现了细胞核与细胞质的有效分割。结果与结论该算法有效实现了宫颈细胞显微图像的分割,分割后细胞图像边界完整清晰,未出现失真。大量实验证明,该算法具有分割速度快,精度高,算法稳定性好等特点。Objective To propose an efficient algorithm for segmenting the cell nucleus and cytoplasm from cervical cells in microscopic images.Methods The maximum between-class variance algorithm was adopted to separate cells from their background.With the maximum entropy of greyscale image as fitness function,image segmentation was optimized.Incorporating the robustness,adaptability and efficiency of the genetic algorithm,this method was able to obtain an optimal threshold,and achieve effective segmentation of the nucleus and cytoplasm.Results and Conclusion The experimental results showed that the segmentation algorithms used in this research are effective.Compared to traditional genetic algorithms,this algorithm can not only improve the optimization capability,but greatly diminish the searching time of the threshold.

关 键 词：宫颈细胞 最大类间方差法 最佳直方图熵法 遗传算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]