纳米晶SEM图像复原的非凸高阶全变差方法  

Nonconvex High-Order Total Variation Method for Restoring Nanocrystalline SEM Images

作　　者：赵瑞妮 ZHAO Ruini(College of ASEAN Studies,Guangxi Minzu University,Nanning 530006,China)

机构地区：[1]广西民族大学东盟学院,广西南宁530006

出　　处：《广西民族大学学报（自然科学版）》2024年第3期78-84,共7页Journal of Guangxi Minzu University ：Natural Science Edition

基　　金：湖南省教育厅重点项目(22A0339);广西民族大学科学研究项目(2023KJQD10)。

摘　　要：针对获得的纳米晶SEM图像中存在泊松噪声的现象,建立一个基于非凸高阶全变差正则化的去噪模型。提出的新模型融合了非凸势函数和高阶全变差约束的优点,这种结合有助于克服阶梯假象并保持图像的细节特征。为了求解新模型,结合算子分裂法和迭代重加权L1算法,提出一个改进的交替最小化算法。与其他模型相比,模拟实验和真实实验结果证明了新方法的有效性和优越性。Regarding the phenomenon of Poisson noise in the obtained nanocrystalline SEM images,this paper establishes a denoising model based on nonconvex high-order total variation regularization.The proposed new model incorporates the advantages of nonconvex function and high-order total variation,and this combination helps to overcome the staircase artifacts and maintain detail features of the images.To solve the designed new model,this paper develops an improved alternating minimization algorithm by combining the operator splitting method and iteratively reweighted L1 algorithm.Compared with other models,several simulations and real experimental results demonstrate the effectiveness and superiority of our new scheme.

关 键 词：图像复原 泊松噪声 非凸函数 交替最小化算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]