基于多空间图像融合的白细胞自动分割  

Automatic Segmentation of Leukocyte Image Based on Multiple Spatial Image Fusion

作　　者：王康[1] 李晓春[1] 

机构地区：[1]中南大学物理与电子学院,湖南长沙410083

出　　处：《计算机仿真》2015年第3期258-262,共5页Computer Simulation

摘　　要：针对目前白细胞分割存在的准确率不高、粘连细胞不易分割等问题,提出基于多空间图像融合的白细胞自动分割算法。根据细胞显微图像饱和度分量S和蓝色分量B的分布特点,构造融合图像,减弱内部空洞对分割的影响,为细胞核和单细胞的初分割奠定基础。对于粘连细胞图像,利用轮廓相邻点的行列式检测细胞边缘轮廓凹点并根据凹点索引值关系式定位粘连区域,最后采用线性插值修补轮廓。实验结果表明,上述算法可精确分割白细胞,分割正确率达到96%。To solve the problems that white blood cell segmentation accuracy rate is not high and it is difficult to segment adhesion cell,this paper proposed an automatic segmentation method of leukocyte image based on multiple spatial image fusion. The algorithm was used to construct a fusion image based on the saturation and blue component according to the characteristics of the distribution of cell microscopic image and weaken the influence of the internal cavity of the division,which lays a foundation for segmentation of cell and nucleus. For the adherent cell image,the determinant of the adjacent points was used to detect concave points and locate the overlapping regions according to the relation of pit index values. Finally,the contour was re-constructed using linear interpolation. Experiments show that the algorithm can realize accurate segmentation of leukocyte image,and the accuracy rate is 96%.

关 键 词：细胞分割 图像融合 凹点索引值 粘连细胞 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]