摘要
傅里叶叠层成像(FPM)受硬件和算法等因素的限制,成像的整体性能有待提高。为解决传统FPM技术成像速度慢、成像质量低的问题,融入深度学习的FPM图像重建方法得到广泛关注。基于此,提出一种基于超分辨率对抗生成网络的FPM模型,在原有网络基础上通过增加密集块连接实现全局特征融合并且使用一种加权损失函数提高图像重建质量。分辨率板图像重构结果表明,所提深度学习方法较传统方法重建效果显著、重建速度更快。
Fourier ptychography microscopy(FPM) is limited by hardware and algorithm,and its overall performance needs to be improved.To address the issues of slow imaging speed and low imaging quality of traditional FPM technology,the FPM image reconstruction approach integrated with depth learning has been widely explored.Herein,based on this,a super-resolution countermeasure generation network-based FPM model is proposed.Furthermore,global feature fusion is obtained by adding dense block connections using the original network,and a weighted loss function is used to enhance the quality of image reconstruction.The reconstruction results of the resolution plate image demonstrate that the proposed depth learning method has a better reconstruction effect and faster reconstruction speed than the conventional method.
作者
王一
魏晓雨
刘保辉
苏皓
Wang Yi;Wei Xiaoyu;Liu Baohui;Su Hao(College of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,Hebei,China;Tangshan Technology Innovation Center of Intellectualisation of Metal Component Production Line,Tangshan 063210,Hebei,China;Tangshan Key Laboratory of Semiconductor Integrated Circuits,Tangshan 063210,Hebei,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2023年第20期168-174,共7页
Laser & Optoelectronics Progress
基金
河北省高等学校科学研究项目(ZD2022114)
唐山市科技计划项目(21130212C)。
关键词
显微
计算成像
傅里叶叠层显微成像
对抗生成网络
超分辨率重建
深度学习
microscopy
computational imaging
Fourier ptychography microscopy
generative adversarial network
super-resolution reconstruction
deep learning