基于改进Canny算法的浑浊菌液显微图像边缘检测  

Edge Detection of Turbid Bacterial Liquid Microscopic Images Based on a Modified Canny Algorithm

作　　者：于章清 宋克纳 薛云[1] 胡志刚[1] 祖向阳 王新征 YU Zhangqing;SONG Kena;XUE Yun;HU Zhigang;ZU Xiangyang;WANG Xinzheng(School of Medical Technology and Engineering,Henan University of Science and Technology,Luoyang 471000)

机构地区：[1]河南科技大学医学技术与工程学院,洛阳471000

出　　处：《计算机与数字工程》2024年第12期3710-3714,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:32072899,12104134);河南省科技发展计划(编号:212102310741,212102310887);河南省高等学校重点科研项目(编号:21A416005)资助。

摘　　要：浑浊菌液的低识别度图像边缘检测是困扰显微图像的分割与识别的一大难题,而传统的Canny算法存在过度平滑图像、人为设定双阈值、误判率高和漏判等问题。论文提出了一种改进的Canny算法。采用双边滤波器替代高斯滤波器进行预处理,达到保边去噪的目的;增加了45°与135°方向的梯度模板计算梯度幅值和方向;使用线性插值改进非极大值抑制过程,增加定位精度;阈值选取由人工选取变为3×3窗口均值自动生成阈值,增强了算法的适应性;最后,采用递归边界跟踪法连接边缘。选取高、低对比度显微图像对论文算法进行验证。结果表明,改进的算法效果明显优于传统算法,实现了浑浊样本细胞显微图像的边缘准确提取,为后续细胞显微图像的分割与识别提供了保障。The edge detection of low-resolution images taken from turbid bacterial liquid is a confusing problem in the seg-mentation and recognition of microscopic images.In traditional Canny algorithm,there are some problems in edge detection,such as excessive image smoothing,manual determination of threshold,high error rate and missed judgment.This paper presents an im-proved Canny algorithm.In this paper,the bilateralfilter instead of Gaussian filter for preprocessing is used to preserve the edge and remove the noise.The gradient template in 45°and 135°directions is added to calculate the gradient amplitude.Linear interpo-lation is used to improve non-maximum suppression,which enhances the accuracy of edge detection.Threshold selection is changed from manual selection to 3x3 window mean automatic generation threshold,which enhances the adaptability.The edges are connected by the recursive boundary tracking method.The algorithm is verified by high contrast and low contrast microscopic imag-es.The results show that improved Canny algorithm is better than traditional algorithm.This algorithm can accurately extract the edge of the microscopic image of turbid sample cells,which is significate for the subsequent cell microscopic image segmentation and recognition.

关 键 词：显微图像 双边滤波 自适应阈值 局部均值 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]