Deep learning assisted variational Hilbert quantitative phase imaging
딥러닝 보조 변분 힐버트 정량 위상 영상
Imagen de fase cuantitativa de Hilbert variable asistida por aprendizaje profundo
Apprentissage profond assisté variationnel Hilbert Quantitative Phase Imaging
Количественная фазовая визуализация Гильберта
Zhuoshi Li 李卓识 ¹ ² ³, Jiasong Sun 孙佳嵩 ¹ ² ³, Yao Fan 范瑶 ¹ ² ³, Yanbo Jin 金彦伯 ¹ ² ³, Qian Shen 沈茜 ¹ ² ³, Maciej Trusiak ⁴, Maria Cywińska ⁴, Peng Gao 郜鹏 ⁵, Qian Chen 陈钱 ³, Chao Zuo 左超 ¹ ² ³
¹ Smart Computational Imaging Laboratory (SCILab), School of Electronic and Optical Engineering, Nanjing University of Science and Technology, Nanjing 210094, China
中国 南京 南京理工大学智能计算成像实验室
² Smart Computational Imaging Research Institute (SCIRI) of Nanjing University of Science and Technology, Nanjing 210094, China
中国 南京 南京理工大学智能计算成像研究院
³ Jiangsu Key Laboratory of Spectral Imaging and Intelligent Sense, Nanjing 210094, China
中国 南京 江苏省光谱成像与智能感知重点实验室
⁴ Institute of Micromechanics and Photonics, Warsaw University of Technology, 8 Sw. A. Boboli St., Warsaw 02-525, Poland
⁵ School of Physics, Xidian University, Xi'an 710126, China
中国 西安 西安电子科技大学物理学院
Opto-Electronic Science, 18 May 2023

We propose a high-accuracy artifacts-free single-frame digital holographic phase demodulation scheme for relatively low-carrier frequency holograms—deep learning assisted variational Hilbert quantitative phase imaging (DL-VHQPI). The method, incorporating a conventional deep neural network into a complete physical model utilizing the idea of residual compensation, reliably and robustly recovers the quantitative phase information of the test objects.

It can significantly alleviate spectrum-overlapping-caused phase artifacts under the slightly off-axis digital holographic system. Compared to the conventional end-to-end networks (without a physical model), the proposed method can reduce the dataset size dramatically while maintaining the imaging quality and model generalization.

The DL-VHQPI is quantitatively studied by numerical simulation. The live-cell experiment is designed to demonstrate the method's practicality in biological research. The proposed idea of the deep learning-assisted physical model might be extended to diverse computational imaging techniques.
Opto-Electronic Science_1
Opto-Electronic Science_2
Opto-Electronic Science_3
Reviews and Discussions
Top-down control of bottom-up material synthesis @ nanoscale
Deep learning enhanced NIR-II volumetric imaging of whole mice vasculature
Beam splitter benefits from topological antichiral edge states
Tailoring spatiotemporal dynamics of plasmonic vortices
Lensless complex amplitude demodulation based on deep learning in holographic data storage
High-performance warm white LED based on thermally stable all inorganic perovskite quantum dots
Highly efficient vectorial field manipulation using a transmitted tri-layer metasurface in the terahertz band
High performance "non-local" generic face reconstruction model using the lightweight Speckle-Transformer (SpT) UNet
Comparative analysis of NovaSeq 6000 and MGISEQ 2000 single-cell RNA sequencing data
Integrated liver proteomics and metabolomics identify metabolic pathways affected by pantothenic acid deficiency in Pekin ducks
Solvent-free fabrication of broadband WS₂ photodetectors on paper
Janus aramid nanofiber aerogel incorporating plasmonic nanoparticles for high-efficiency interfacial solar steam generation

Previous Article                                Next Article
Copyright © Hot Paper