Computational Imaging

Go to contentGo to searchGo to menu
You are here:

Optical Design & Image Reconstruction:Activities in Computational Imaging

The team is very active in two unconventional imaging modalities: digital holographic microscopy (DHM) and synthetic aperture radar imaging (SAR). Although applied at different scales (microscopy samples or Earth observation), these imaging techniques both use coherent illumination and require the development of specific digital reconstruction techniques that exploit the knowledge of the physics of image formation model:

Digital holographic microscopy (DHM)

In digital holographic microscopy, the team develops instruments and digital processing approaches for biomedical applications in microbiological diagnoses. Holography provides access to absorption and phase shift induced by samples, and can therefore be used to image transparent objects (phase imaging). This activity, funded via academic research projects (ANR, Région AURA) and industrial (bioMérieux) research projects, has led to techniques for accurate auto-focusing in DHM ([Brault et al. 2022], patent with bioMérieux), and aberration characterisation and compensation ([Brault et al. 2023], patent with bioMérieux) studied during Dylan Brault's thesis, as well as new reconstruction methods. Our reconstruction algorithms, based on inverse problem solving [Momey et al. 2019], enable fine calibration of the system (optical aberrations, focus) together with a more quantitative and repeatable reconstruction of biological samples [Berdeu et al. 2019]. It also makes possible the separation of the contribution of the background and objects of interest [Berdeu et al. 2021], or even produce a volume reconstruction from multi-angle illuminations (tomographic difffractive microscopy) [Denneulin et al. 2022].

A highlight in this field is the development of self-calibration and multispectral reconstruction methods for microbiological diagnosis [Brault et al. 2023].

Multi-wavelength holographic microscopy of Gram-labelled bacterially infected blood smears (A) traditional RGB image and series of 8-colour in-line holograms (a-gram-negative bacteria, b-gram-positive bacteria, c-red blood cells, d-calibration beads). (B) Inverse problem multi-wavelength reconstructions on a bacillus (top: sequential reconstruction uncorrected for aberrations, bottom: aberration-corrected multi-wavelength reconstruction corrected for aberrations). (C) details of aberration-corrected multi-wavelength inverse problem reconstructions: in transmission and in Optical Path Difference (OPD) [Brault et al., Scientific Reports 2023, link to the paper].

Our work in DHM since 2007 has led to the development of a Matlab toolbox dedicated to the reconstruction of in-line digital holograms based on Inverse Problems methodology.

→ Documentation and download here

Synthetic Aperture Radar imaging (SAR)

In SAR imaging, major advances have been made in speckle reduction methodologies: using multi-temporal filtering [Zhao et al. 2019][Gasnier et al. 2021a] or deep learning (supervised: [Dalsasso et al. 2020], semi-supervised: [Dalsasso et al. 2021], self-supervised: [Dalsasso et al. 2022], [Meraoumia et al. 2023], plug-and-play ADMM: [Deledalle et al. 2022], [Ulondu-Mendes et al. 2024]).

A highlight is the development of a self-supervised learning strategy based on the decomposition of complex amplitudes into their real and imaginary parts (algorithm MERLIN [Dalsasso et al. 2022]).

A SAR image from the TerraSAR-X satellite (spatial resolution: 3m), before and after speckle noise speckle noise reduction using our self-supervised approach [Dalsasso et al., IEEE TGRS 2022, link to the paper].

Several 3D reconstruction methodologies have also been proposed for SAR tomography: regularised inversion [Rambour et al. 2019a], surface reconstruction using graph-cuts [Rambour et al. 2019b], deep learning [Berenger et al. 2023]. Segmentation techniques for lakes and rivers have been developed [Lobry et al. 2019][Gasnier et al., 2021b] in preparation for the SWOT space mission (CNES/NASA) and are now integrated into the CNES operational chain for the generation of hydrological products. This work has had a significant impact (GRSS Symposium prize among more than 2000 papers submitted, invitation by foreign colleagues to participate in 2 review articles in the high-impact IEEE Geoscience and Remote Sensing Magazine [Rambour et al. 2020][Rasti et al. 2022], invited tutorial « Machine Learning for SAR Processing » at EUSAR2024).

 

Publications illustrative of our scientific activity :

  • in holographic microscopy

[Brault et al. 2022] Dylan Brault, Corinne Fournier, Thomas Olivier, Nicolas Faure, Sophie Dixneuf, et al.. Automatic numerical focus plane estimation in digital holographic microscopy using calibration beads. Applied optics, 2022, 61 (5), pp.B345. DOI. Preprint

[Brault et al. 2023] Dylan Brault, Thomas Olivier, Nicolas Faure, Sophie Dixneuf, Chloé Kolytcheff, et al.. Multispectral in-line hologram reconstruction with aberration compensation applied to Gram-stained bacteria microscopy. Scientific Reports, 2023, 13 (1), pp.14437. DOI. Preprint

[Momey et al. 2019] Fabien Momey, Loïc Denis, Thomas Olivier, Corinne Fournier. From Fienup’s phase retrieval techniques to regularized inversion for in-line holography: tutorial. Journal of the Optical Society of America. A Optics, Image Science, and Vision, 2019, 36 (12), pp.D62-D80. DOI. Preprint

[Berdeu et al. 2019] Anthony Berdeu, Olivier Flasseur, Loïc Méès, Loïc Denis, Fabien Momey, et al.. Reconstruction of in-line holograms: combining model-based and regularized inversion. Optics Express, 2019, 27 (10), pp.14951. DOI. Preprint

[Berdeu et al. 2021] Anthony Berdeu, Thomas Olivier, Fabien Momey, Loïc Denis, Frédéric Pinston, et al.. Joint reconstruction of an in-focus image and of the background signal in in-line holographic microscopy. Optics and Lasers in Engineering, 2021, 146, pp.106691. DOI. Preprint

[Denneulin et al. 2022] L. Denneulin, F. Momey, D. Brault, M. Debailleul, A M Taddese, et al. GSURE criterion for unsupervised regularized reconstruction in tomographic diffractive microscopy. Journal of the Optical Society of America. A Optics, Image Science, and Vision, 2022, 39 (2), pp.A52. DOI. Preprint

  • in SAR imaging

[Zhao et al. 2019] Weiying Zhao, Charles-Alban Deledalle, Loïc Denis, Henri Maître, Jean-Marie Nicolas, et al.. Ratio-Based Multitemporal SAR Images Denoising: RABASAR. IEEE Transactions on Geoscience and Remote Sensing, 2019, DOI. Preprint

[Gasnier et al. 2021a] Nicolas Gasnier, Loïc Denis, Florence Tupin. On the use and denoising of the temporal geometric mean for SAR time series. IEEE Geoscience and Remote Sensing Letters, 2021, DOI. Preprint

[Dalsasso et al. 2020] Emanuele Dalsasso, Xiangli Yang, Loïc Denis, Florence Tupin, Wen Yang. SAR Image Despeckling by Deep Neural Networks: from a Pre-Trained Model to an End-to-End Training Strategy. Remote Sensing, 2020, 12 (16), pp.2636. DOI. Preprint

[Dalsasso et al. 2021] Emanuele Dalsasso, Loïc Denis, Florence Tupin. SAR2SAR: a semi-supervised despeckling algorithm for SAR images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, inPress, pp.1-1. DOI. Preprint

[Dalsasso et al. 2022] Emanuele Dalsasso, Loïc Denis, Florence Tupin. As if by magic: self-supervised training of deep despeckling networks with MERLIN. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60, pp.1-13. DOI. Preprint

[Meraoumia et al. 2023] Inès Meraoumia, Emanuele Dalsasso, Loïc Denis, Rémy Abergel, Florence Tupin. Multi-temporal speckle reduction with self-supervised deep neural networks. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61, DOI. Preprint

[Deledalle et al. 2022] Charles-Alban A Deledalle, Loïc Denis, Florence Tupin. Speckle reduction in matrix-log domain for synthetic aperture radar imaging. Journal of Mathematical Imaging and Vision, 2022, 64, pp.298-320. DOI. Preprint

[Ulondu-Mendes et al. 2024] Cristiano Ulondu-Mendes, Loïc Denis, Charles-Alban Deledalle, Florence Tupin. Robustness to spatially-correlated speckle in Plug-and-Play PolSAR despeckling. IEEE Transactions on Geoscience and Remote Sensing, 2024, 62, DOI. Preprint

[Rambour et al. 2019a] Clément Rambour, Loïc Denis, Florence Tupin, Hélène Oriot. Introducing spatial regularization in SAR tomography reconstruction. IEEE Transactions on Geoscience and Remote Sensing, 2019, DOI. Preprint

[Rambour et al. 2019b] Clément Rambour, Loïc Denis, Florence Tupin, Hélène Oriot, Yue Huang, et al.. Urban Surface Reconstruction in SAR Tomography by Graph-Cuts. Computer Vision and Image Understanding, 2019, 188, pp.102791. DOI. Preprint

[Berenger et al. 2023] Zoé Berenger, Loïc Denis, Florence Tupin, Laurent Ferro-Famil, Yue Huang. A Deep Learning Approach for SAR Tomographic Imaging of Forested Areas. IEEE Geoscience and Remote Sensing Letters, 2023, 20, pp.4007405. DOI. Preprint

[Lobry et al. 2019] Sylvain Lobry, Loïc Denis, Brent Williams, Roger Fjørtoft, Florence Tupin. Water Detection in SWOT HR Images Based on Multiple Markov Random Fields. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2019, 12 (11), pp.4315-4326. DOI. Preprint

[Gasnier et al., 2021b] Nicolas Gasnier, Loïc Denis, Roger Fjørtoft, Frédéric Liege, Florence Tupin. Narrow River Extraction from SAR Images Using Exogenous Information. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14, pp.5720 - 5734. DOI. Preprint

[Rambour et al. 2020] Clement Rambour, Alessandra Budillon, Angel Johnsy, Loïc Denis, Florence Tupin, et al.. From Interferometric to Tomographic Synthetic Aperture Radar. Scatterer unmixing in urban areas: A review of synthetic aperture radar tomography-processing techniques. IEEE geoscience and remote sensing magazine, 2020, 8 (2). DOI

[Rasti et al. 2022] Behnood Rasti, Yi Chang, Emanuele Dalsasso, Loic Denis, Pedram Ghamisi. Image Restoration for Remote Sensing: Overview and toolbox. IEEE geoscience and remote sensing magazine, 2022, 10 (2), pp.201-230. DOI. Preprint