|本期目录/Table of Contents|

[1]肖月,李朋伟△.基于梯度修正的改进分水岭模型在细胞分割中的应用*[J].生物医学工程研究,2020,04:330-336.
 XIAO Yue,LI Pengwei.Application of improved watershed model based on gradient correction in cell segmentation[J].Journal of Biomedical Engineering Research,2020,04:330-336.
点击复制

基于梯度修正的改进分水岭模型在细胞分割中的应用*(PDF)

《生物医学工程研究》[ISSN:1006-6977/CN:61-1281/TN]

期数:
2020年04期
页码:
330-336
栏目:
出版日期:
2020-12-25

文章信息/Info

Title:
Application of improved watershed model based on gradient correction in cell segmentation
文章编号:
1672-6278 (2020)04-0330-07
作者:
肖月李朋伟△
太原理工大学信息与计算机学院,晋中 030600
Author(s):
XIAO YueLI Pengwei
College of Information and Computer Engineering,Taiyuan University of Technology, Jinzhong 030600, China
关键词:
Canny算子结构元梯度修正分水岭变换细胞分割
Keywords:
Canny operatorStructural elementGradient correctionWatershed transformationCell segmentation
分类号:
R318
DOI:
10.19529/j.cnki.1672-6278.2020.04.02
文献标识码:
A
摘要:
显微细胞的准确分割是相关疾病在计算机辅助诊疗过程中的关键,良好的分割结果为疾病诊断提供了重要依据。为提高细胞图像的分割精度,本研究提出一种Canny算子结合形态学对梯度进行修正的改进分水岭算法。算法首先利用Canny算子提取细胞图像HSV颜色空间各分量的梯度图,并利用图像熵的概念加权拟合,然后对梯度图进行形态学平滑滤波,之后由灰度方差阈值获取前、后标记图像,最后用标记图像修改梯度图像并作分水岭变换求得分割结果。结果表明,本研究算法较经典分水岭算法,其交并比参数值提高了31.86%,分割精度提高了21.88%,与形态学分水岭算法相比,其值分别提高了14.69%和8.87%,且细胞边缘分割完整、连续,无过分割现象。
Abstract:
Accurate segmentation of the microscopic cell images is the key to computer-aided diagnosis and treatment of related diseases. Good segmentation results provide an important basis for disease diagnosis.To improve the segmentation accuracy of cell images, we proposed an improved watershed algorithm with Canny operator combined with morphology to modify the gradient.Firstly,the Canny operator was used to extract the gradient of the HSV component of the images.The concept of image entropy was used to weight fit and then the gradient maps was smoothed based on morphological transformation.By using the variance of the gray scale to obtain the local minimum value of the front and back marked images.Finally,the gradient image was modified by the marked images and the watershed transform was used to obtain the segmentation result.The experimental results show that compared with the classical watershed algorithm, this algorithm improves the IOU parameter value by 31.86% and the segmentation accuracy by 21.88%. Compared with the morphological watershed algorithm, the parameter value is increased by 14.69% and 8.87% respectively,and the edge segmentation is complete and continuous without over-segmentation.

参考文献/References

备注/Memo

备注/Memo:
(收稿日期:220-04-18)国家自然科学基金资助项目(61474079,11602159);山西省优秀人才科技创新项目(201605D211020);山西省国际合作项目(201803D421029)。△通信作者Email:lipengwei@tyut.edu.cn
更新日期/Last Update: 2021-02-05