极坐标描述的显微白细胞图像分割算法  

Polar Coordinates Representation Based Leukocyte Segmentation of Microscopic Cell Images

作　　者：顾广华[1] 崔冬[1] 郝连旺[2] 

机构地区：[1]燕山大学信息科学与工程学院,秦皇岛066004 [2]燕山大学电气工程学院,秦皇岛066004

出　　处：《生物医学工程学杂志》2010年第6期1237-1242,共6页Journal of Biomedical Engineering

基　　金：河北省自然科学基金资助项目(F2008000891);秦皇岛市科学技术研究与发展指导计划资助项目(201001A14)

摘　　要：本文提出了一种显微细胞图像中粘连白细胞的分割算法。利用指纹图平滑算法对白细胞图像的饱和度分量的直方图进行平滑,获得有意义的谷点,据此分割出白细胞核;根据细胞核的等效半径划定一个圆形区域,使其包含白细胞的全部区域;用极坐标表示粘连白细胞的边缘,通过检测边缘像素点幅角的变化来检测细胞的粘连区域;结合粘连区域的梯度信息重构白细胞边缘,最终精确分割出白细胞。实验表明,本文分割算法具有很好的检全率、检准率和像素误差率性能。We propose an algorithm for segmentation of the overlapped leukocyte in the microscopic cell image.The histogram of the saturation channel in the cell image is smoothed to obtain the meaningful global valley point by the fingerprint smoothing method,and then the nucleus can be segmented.A circular region,containing the entire regions of the leukocyte,is marked off according to the equivalent sectional radius of the nucleus.Then,the edge of the overlapped leukocyte is represented by polar coordinates.The overlapped region by the change of the polar angle of the edge pixels is determined,and the closed edge of the leukocyte integrating the gradient information of the overlapped region is reconstructed.Finally,the leukocyte is exactly extracted.The experimental results show that our method has good performance in terms of recall ratio,precision ratio and pixel error ratio.

关 键 词：白细胞分割 细胞粘连 指纹图平滑 极坐标描述 幅角检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]