基于SRCNN的AFM超分辨率成像  

AFM Super-resolution Imaging Based on SRCNN

作　　者：林泓 林晨宇 吴腾 韩国强 LIN Hong;LIN Chenyu;WU Teng;HAN Guoqiang(Ocean School,Fuzhou University,Fuzhou 350108,China;School of Mechanical Engineering and Automation,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学海洋学院,福建福州350108 [2]福州大学机械工程及自动化学院,福建福州350108

出　　处：《机械制造与自动化》2022年第6期89-92,共4页Machine Building & Automation

基　　金：国家级大学生创新创业训练计划项目(SRTP)(202110386038)。

摘　　要：传统的光栅扫描法需要花费大量的时间来获得高分辨率图像,超分辨率算法可以用来提高原子力显微镜(AFM)图像的质量。但基于插值的方法容易产生图像伪影和边缘模糊,基于重构的图像处理方法也需要更好的先验知识和重构算法。为了尽可能获得AFM图像中更详细的纹理和特征信息,采用基于卷积神经网络的方法实现AFM超分辨率成像,并与传统的超分辨率方法进行对比,通过对重建图像的主客观评价验证了所提算法的可行性。As traditional raster scanning method is time consuming in obtaining high-resolution image,super resolution algorithm is applied to improve the quality of AFM image.However,due to the fact that interpolation based methods tend to cause image artifacts and edge blur,it is of necessity to acquire better prior knowledge and reconstruction algorithm for image processing methods based on reconstruction.In order to obtain more detailed texture and feature information in AFM image as much as possible,the method based on convolutional neural network is proposed to realize AFM super-resolution imaging,with which the traditional super-resolution methods are compared.The proposed algorithm is verified to be feasible through the subjective and objective evaluations of the reconstructed image.

关 键 词：原子力显微镜 超分辨率 卷积神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]