焦炭光学组织离焦显微图像去模糊算法研究  

Study on the deblurring algorithm of off-focus microscopic images of coke optical tissue

作　　者：刘怀广 董渊 汤勃 黄千稳 LIU Huaiguang;DONG Yuan;TANG Bo;HUANG Qianwen(Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学机械传动与制造工程湖北省重点实验室,武汉430081

出　　处：《智能计算机与应用》2023年第9期25-32,共8页Intelligent Computer and Applications

基　　金：国家自然科学基金面上项目(51874217,52272377)。

摘　　要：微观组织结构是判定焦炭质量的重要依据,通常需要偏光显微镜对焦炭光片成像,但显微镜头的特殊性,使得图像中间清晰而周围存在离焦模糊,降低了图像的质量,影响对焦炭组织成分的判定。因此,本文结合局部模糊分割算法和离焦去模糊网络,提出一种局部去模糊算法。首先对焦炭图像进行基于LBP的模糊检测,得到模糊度量图;通过Alpha抠图算法分割度量图,获得α模板矩阵;最后,以α模板矩阵作为融合比列,对原图和全处理图像进行线性融合,得到去模糊图像。通过实验对比,证明本文方法在累加概率(CPBD)等4个质量评价标准上面都得到了提升,相比于传统方法,去模糊效果更佳。Microstructure is an important basis for determining the quality of coke,which usually requires polarized light microscopy to image the coke light sheet,but the special characteristics of the microscope head make the image clear in the middle and out-of-focus blurring around,which reduces the image quality and affects the determination of coke tissue composition.Therefore,this paper proposes a local deblurring algorithm by combining the local fuzzy segmentation algorithm and the out-offocus deblurring network.First,LBP-based blur detection is performed on the coke image to obtain the blurred metric map;then the metric map is segmented by the Alpha matting algorithm to obtain the alpha template matrix;finally,the original image and the fully processed image were linearly fused with the alpha template matrix as the fusion ratio to obtain the final deblurred image.Through experimental comparison,the method in this paper achieves significant improvement over four quality criteria such as cumulative probability(CPBD),and has better deblurring effect compared with traditional methods.

关 键 词：焦炭图像 局部离焦模糊 模糊检测 IFAN网络 图像去模糊 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]