面向纤维显微成像的图像清晰度评价算法  

Image sharpness evaluation algorithm for fiber microscopic imaging

作　　者：瞿瑞德 李子印 汪小东 叶飞 金君 QU Ruide;LI Ziyin;WANG Xiaodong;YE Fei;JIN Jun(School of Optics and Electronic Technology,China Jiliang University,Hangzhou,Zhejiang 310018,China;Huzhou Institute of Quality and Technical Supervision and Inspection(Huzhou Fiber Quality Monitoring Center),Huzhou,Zhejiang 313099,China)

机构地区：[1]中国计量大学光学与电子科技学院,浙江杭州310018 [2]湖州市质量技术监督检测研究院(湖州市纤维质量监测中心),浙江湖州313099

出　　处：《毛纺科技》2022年第12期85-94,共10页Wool Textile Journal

基　　金：浙江省市场监督管理局雏鹰计划培育项目(CY2022352);湖州市科技计划立项项目(2021GZ38);浙江省基础公益研究计划项目(LGN20F050001);浙江省市场监督管理局科研计划项目(20210146)。

摘　　要：针对基于Tenengrad梯度函数等算法不能很好适应图像的清晰度评价,所找到的“最清晰图像”是纤维边缘出现粗黑边线的图像,与GB/T 10685—2007《羊毛纤维直径试验方法投影显微镜法》要求不一致的问题,提出一种新的适用于羊毛、羊绒图像的清晰度评价算法。首先通过图像预处理、融合形态学和连通域滤波的目标分割提取纤维目标区域作为ROI(感兴趣区域),再通过基于Z轴图像序列二值边缘直方图的清晰度算法完成图像清晰度的计算。清晰度计算先对Canny算子检测的边缘进行二值边缘点数计算,并将结果进行归一化处理和校正,最后根据二值边缘直方图计算清晰度值,得到凸型清晰度曲线,并取其最大值对应的图像为最清晰图像。试验结果表明:该算法的平均准确率达94.5%,找到的最清晰图像符合GB/T 10685—2007的要求。Aiming at the problem that the algorithm based on Tenengrad gradient function and other algorithms cannot adapt well to the image sharpness evaluation,the"clearest image"found is the image with thick black edge,which is inconsistent with the requirements of GB/T 10685-2007"Wool Fiber Diameter Test Method Projection Microscope Method".A new resolution evaluation algorithm for wool and cashmere images was presented in this paper.Firstly,the fiber target region was extracted as ROI(region of interest)by image preprocessing,fusion morphology and connected domain filtering,and then the definition algorithm based on the Z-axis image sequence binary edge histogram was used to calculate the image clarity.For image clarity calculation,firstly,the binary edge points are calculated for the edges detected by Canny operator,and the results are normalized and corrected.Then,the definition value is calculated according to the binary edge histogram,and the convex definition curve is obtained,and the image corresponding to the maximum value is the clearest image.The experimental results show that the average accuracy of this algorithm is 94.5%,and the clearest image found conforms to the requirement of GB/T 10685-2007.

关 键 词：羊毛羊绒纤维图像 目标分割 Z轴图像序列 二值边缘直方图 图像清晰度评价 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]