物理模型引导的智能相位成像

刘振; 朱昊; 周游; 马展; 曹汛

doi:10.11959/j.issn.2096-3750.2023.00345

您当前的位置：

首页 >

文章列表页 >

物理模型引导的智能相位成像

专题：智能感知技术 | 更新时间：2024-08-16

- 物理模型引导的智能相位成像
- Intelligent phase imaging guided by physics models
- 物联网学报 2023年7卷第2期页码：35-42
- 作者机构：
  
  1. 南京大学电子科学与工程学院，江苏南京 210023
  2. 南京大学计算机科学与技术系，江苏南京 210023
  3. 南京大学医学院，江苏南京 210093
- 作者简介：
  
  [ "刘振（2002- ），男，南京大学电子科学与工程学院与计算机科学与技术系联合培养硕士生，主要研究方向为隐式神经表示、无透镜成像等" ]
  [ "朱昊（1992- ），男，博士，南京大学副研究员，主要研究方向为物理启发神经网络和光场成像等" ]
  [ "周游（1990- ），男，博士，南京大学助理教授、特聘研究员，主要研究方向为光学显微成像（无透镜显微成像、内窥显微成像、光场显微成像）和智能重构算法等" ]
  [ "马展（1981- ），男，博士，南京大学教授，主要研究方向为类脑视频通信、视频编码、计算摄像、深度学习等" ]
  [ "曹汛（1983- ），男，博士，南京大学教授，主要研究方向为图像和视频处理、光谱成像、计算摄像等" ]
- 基金信息：
  
  国家自然科学基金资助项目;The National Natural Science Foundation of China(62025108);国家自然科学基金资助项目;The National Natural Science Foundation of China(62071219);国家自然科学基金资助项目;The National Natural Science Foundation of China(62101242)
- DOI：10.11959/j.issn.2096-3750.2023.00345
  中图分类号： TN26
- 纸质出版日期：2023-06-30，
  
  网络出版日期：2023-06，
- 稿件说明：
移动端阅览
刘振, 朱昊, 周游, 等. 物理模型引导的智能相位成像[J]. 物联网学报, 2023,7(2):35-42.

ZHEN LIU, HAO ZHU, YOU ZHOU, et al. Intelligent phase imaging guided by physics models. [J]. Chinese journal on internet of things, 2023, 7(2): 35-42.
刘振, 朱昊, 周游, 等. 物理模型引导的智能相位成像[J]. 物联网学报, 2023,7(2):35-42. DOI： 10.11959/j.issn.2096-3750.2023.00345.

ZHEN LIU, HAO ZHU, YOU ZHOU, et al. Intelligent phase imaging guided by physics models. [J]. Chinese journal on internet of things, 2023, 7(2): 35-42. DOI： 10.11959/j.issn.2096-3750.2023.00345.

摘要

隐式神经表示使用神经网络刻画了信号坐标到其属性的映射，通过将正向物理过程引入误差函数的设计中，可被用于求解各种逆问题，应用前景广阔。然而，对隐式神经表示的网络参数进行随机初始化会导致优化速度慢、求解精度低，因此，提出使用元学习算法为隐式神经表示提供一个具有强先验的初始化参数，从而提升求解逆问题时的优化速度和求解精度。针对无透镜相位成像这一重要问题，基于快照式无透镜感知模型提出一种智能相位成像方法，将光学衍射传播理论引入隐式神经表示的误差函数设计中，能够消除传统深度学习算法对大规模数据集的依赖，仅需要传感器记录的单张强度图像，即可实现对样本的高精度相位恢复。此外，通过在网络初始化中引入元学习模型，进一步提升网络训练的效率和精度。数值仿真结果表明，与传统方法相比，所提方法能够获得11 dB以上的峰值信噪比（PSNR

peak signal-to-noise ratio）提升；在真实数据中的实验结果表明，所提方法重建出的相位图像更加清晰，伪影更少。

Abstract

Implicit neural representation characterizes the mapping between the signal’s coordinate to its attributes

and has been widely used in the optimization of inverse problems by embedding the physics process into the loss function.However

the implicit neural representation is suffering the low convergence speed and accuracy from the random initialization of the network parameters.The meta-learning algorithm for providing implicit neural representation with a strong prior of network parameters was proposed

thus enhancing the optimization efficiency and accuracy for solving inverse problems.To address the important issue of lens less phase imaging

an intelligent method on phase imaging was proposed based on the snapshot lens less sensing model.By embedding the optical diffraction propagation theory into the design of loss function for implicit neural representation

the dependency of large-scale labelled dataset in traditional deep learning-based methods was eliminated and accurate phase image from a single diffraction measurement image was provided.Furthermore

the meta-learning model was introduced for initializing network to further improve the efficiency and accuracy of network training.Numerical simulation results show that the proposed method can achieve a PSNR improvement of more than 11 dB compared to the conventional method.The experimental results in real data show that the phase image reconstructed by the proposed method is clearer and has fewer artifacts.

关键词

隐式神经表示物理模型相位成像元学习自监督学习

Keywords

implicit neural representationphysics modelphase imagingmeta learningunsupervised learning

references

REN Z B, XU Z M, LAM E Y M . End-to-end deep learning framework for digital holographic reconstruction[J]. Advanced Photonics, 2019,1(1): 016004.

REN Z B, XU Z M, LAM E Y . Learning-based nonparametric autofocusing for digital holography[J]. Optica, 2018,5(4): 337.

WANG H, LYU M, SITU G H . eHoloNet:a learning-based end-to-end approach for in-line digital holographic reconstruction[J]. Optics Express, 2018,26(18): 22603.

LYU M, WANG W, WANG H ,et al. Deep-learning-based ghost imaging[J]. Scientific Reports, 2017,7:17865.

WANG F, WANG H, WANG H C ,et al. Learning from simulation:an end-to-end deep-learning approach for computational ghost imaging[J]. Optics Express, 2019,27(18): 25560.

GOY A, ARTHUR K, LI S ,et al. Low photon count phase retrieval using deep learning[J]. Physical Review Letters, 2018,121(24): 243902.

ZHANG G, GUAN T, SHEN Z Y ,et al. Fast phase retrieval in off-axis digital holographic microscopy through deep learning[J]. Optics Express, 2018,26(15): 19388.

MONAKHOVA K, TRAN V, KUO G ,et al. Untrained networks for compressive lensless photography[J]. Optics Express, 2021,29(13): 20913.

BARBASTATHIS G, OZCAN A, SITU G H . On the use of deep learning for computational imaging[J]. Optica, 2019,6(8): 921.

LECUN Y, BENGIO Y, HINTON G . Deep learning[J]. Nature, 2015,521(7553): 436-444.

WANG F, BIAN Y M, WANG H C ,et al. Phase imaging with an untrained neural network[J]. Light:Science ＆ Applications, 2020(9): 77.

BOSTAN E, HECKEL R, CHEN M ,et al. Deep phase decoder:self-calibrating phase microscopy with an untrained deep neural network[J]. Optica, 2020,7(6): 559.

ZHANG X Y, WANG F, SITU G H . BlindNet:an untrained learning approach toward computational imaging with model uncertainty[J]. Journal of Physics D:Applied Physics, 2022,55(3): 034001.

ZHU H, LIU Z, ZHOU Y ,et al. DNF:diffractive neural field for lensless microscopic imaging[J]. Optics Express, 2022,30(11): 18168.

ZHOU Y, WU J M, SUO J L ,et al. Single-shot lensless imaging via simultaneous multi-angle LED illumination[J]. Optics Express, 2018,26(17): 21418-21432.

SHECHTMAN Y, ELDAR Y C, COHEN O ,et al. Phase retrieval with application to optical imaging:a contemporary overview[J]. IEEE Signal Processing Magazine, 2015,32(3): 87-109.

FIENUP J R . Phase retrieval algorithms:a comparison[J]. Applied Optics, 1982,21(15): 2758.

ZHOU Y, HUA X, ZHANG Z ,et al. Wirtinger gradient descent optimization for reducing Gaussian noise in lensless microscopy[J]. Optics and Lasers in Engineering, 2020(134): 106131.

MAIDEN A, JOHNSON D, LI P . Further improvements to the ptychographical iterative engine[J]. Optica, 2017,4(7): 736.

HOSPEDALES T, ANTONIOU A, MICAELLI P ,et al. Meta-learning in neural networks:a survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021,44(9): 5149-5169.

FINN C, ABBEEL P, LEVINE S . Model-agnostic meta-learning for fast adaptation of deep networks[C]// Proceedings of the 34th International Conference on Machine Learning - Volume 70. New York:ACM Press, 2017: 1126-1135.

TANCIK M, MILDENHALL B, WANG T ,et al. Learned initializations for optimizing coordinate-based neural representations[C]// Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2021: 2845-2854.

RONNEBERGER O, FISCHER P, BROX T . U-Net:convolutional networks for biomedical image segmentation[C]// International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham:Springer, 2015: 234-241.

RAHAMAN N, BARATIN A, ARPIT D ,et al. On the spectral bias of neural networks[EB]. 2018.

MILDENHALL B, SRINIVASAN P P, TANCIK M ,et al. NeRF:representing scenes as neural radiance fields for view synthesis[C]// European Conference on Computer Vision. Cham:Springer, 2020: 405-421.

KINGMA D P, BA J L . Adam:a method for stochastic optimization[J]. 3rd International Conference on Learning Representations,ICLR 2015-Conference Track Proceedings, 2015.

DENG J, DONG W, SOCHER R ,et al. ImageNet:a large-scale hierarchical image database[C]// Proceedings of 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2009: 248-255.

LIU Z W, LUO P, WANG X G ,et al. Deep learning face attributes in the wild[C]// Proceedings of 2015 IEEE International Conference on Computer Vision (ICCV). Piscataway:IEEE Press, 2016: 3730-3738.

浏览量

106

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据