结合自适应TV模型和分水岭变换的图像分割算法  

Image segmentation algorithm of combining adaptive TV model and watershed transform

作　　者：关豪然 宋卫东[1] 张丰收[1] GUAN Haoran;SONG Weidong;ZHANG Fengshou(School of Medical Technology and Engineering,Henan University of Science and Technology,Luoyang 471023,China)

机构地区：[1]河南科技大学医学技术与工程学院,河南洛阳471023

出　　处：《电子设计工程》2023年第4期33-37,42,共6页Electronic Design Engineering

基　　金：国家自然科学基金(31800836);中国博士后科学基金(2020M682285)。

摘　　要：针对传统分水岭算法对噪声敏感,易出现过分割的现象,提出一种自适应全变分模型和标记分水岭算法相结合的图像分割算法。采用自适应全变分模型对原始图像进行滤波处理,平滑去噪的同时保留图像的边缘信息;求解其多尺度形态学梯度图像,并用基于最大熵的扩展极小值技术获得的前景和背景标记并对其多尺度梯度图像修改;对修改后的梯度图像进行分水岭变换,实现准确的分割。对比常用和相似的图像分割算法,实验结果表明,该算法在抗噪性、运行时间和分割交并比上有一定的优势。尤其是在噪声强、灰度值接近的医学图像上能够获得合理有意义的分割区域,效果良好。The traditional watershed algorithm is sensitive to noise and prone to over-segmentation. An image segmentation algorithm that combines an adaptive total variation model and a labeled watershed algorithm is proposed. The original image is filtered by the adaptive total variation model to smooth and denoise while retaining the edge information of the image;The multi-scale morphological gradient image is solved,and the foreground and the minimum value obtained by the extended minimum technique based on maximum entropy are solved. The background label is used to modify the multi-scale gradient image;The modified gradient image is subjected to watershed transformation to achieve accurate segmentation. Comparing commonly used and similar image segmentation algorithms,the experimental results show that the algorithm has certain advantages in noise resistance, running time, and segmentation and intersection. Especially in medical images with strong noise and close gray values,a reasonable and meaningful segmentation area can be obtained,and the effect is good.

关 键 词：图像分割 标记分水岭算法 自适应全变分模型 多尺度形态学梯度图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]