On the use of deep learning for phase recovery

Light Science & Applications, Год журнала: 2024, Номер 13(1)

Опубликована: Янв. 1, 2024

Phase recovery (PR) refers to calculating the phase of light field from its intensity measurements. As exemplified quantitative imaging and coherent diffraction adaptive optics, PR is essential for reconstructing refractive index distribution or topography an object correcting aberration system. In recent years, deep learning (DL), often implemented through neural networks, has provided unprecedented support computational imaging, leading more efficient solutions various problems. this review, we first briefly introduce conventional methods PR. Then, review how DL provides following three stages, namely, pre-processing, in-processing, post-processing. We also used in image processing. Finally, summarize work provide outlook on better use improve reliability efficiency Furthermore, present a live-updating resource ( https://github.com/kqwang/phase-recovery ) readers learn about

Язык: Английский

---

Insights into infrared crystal phase characteristics based on deep learning holography with attention residual network

Journal of Materials Chemistry A, Год журнала: 2025, Номер unknown

Опубликована: Янв. 1, 2025

Through a synergistic blend of infrared digital holography and deep learning, we introduce unconventional mechanistic insight, namely the crystal phase.

Язык: Английский

---

End-to-end infrared radiation sensing technique based on holography-guided visual attention network

Optics and Lasers in Engineering, Год журнала: 2024, Номер 178, С. 108201 - 108201

Опубликована: Март 27, 2024

Язык: Английский

---

Generative adversarial neural network for 3D-hologram reconstruction

Laser Physics Letters, Год журнала: 2024, Номер 21(4), С. 045201 - 045201

Опубликована: Фев. 14, 2024

Abstract Neural-network-based reconstruction of digital holograms can improve the speed and quality micro- macro-object images, as well reduce noise suppress twin image zero-order. Usually, such methods aim to reconstruct 2D object or amplitude phase distribution. In this paper, we investigated feasibility using a generative adversarial neural network 3D-scenes consisting set cross-sections. The method was tested on computer-generated optically-registered inline holograms. It enabled all layers scene from each hologram. is improved 1.8 times when compared U-Net architecture normalized standard deviation value.

Язык: Английский

---

Deep empirical neural network for optical phase retrieval over a scattering medium

Nature Communications, Год журнала: 2025, Номер 16(1)

Опубликована: Фев. 5, 2025

Abstract Supervised learning, a popular tool in modern science and technology, thrives on huge amounts of labeled data. Physics-enhanced deep neural networks offer an effective solution to alleviate the data burden by incorporating analytical model that interprets underlying physical processes. However, it completely fails tackling systems without solution, where wave scattering with multiple input output are typical examples. Herein, we propose concept empirical network (DENN) is hybridization model, which enables seeing through opaque medium untrained manner. The DENN does not rely data, all while delivering as high 58% improvement fidelity compared supervised learning using 30000 pairs for achieving same goal optical phase retrieval. might shed new light applications physics, information science, biology, chemistry beyond.

Язык: Английский

---

Physics-driven universal twin-image removal network for digital in-line holographic microscopy

Optics Express, Год журнала: 2023, Номер 32(1), С. 742 - 742

Опубликована: Ноя. 24, 2023

Digital in-line holographic microscopy (DIHM) enables efficient and cost-effective computational quantitative phase imaging with a large field of view, making it valuable for studying cell motility, migration, bio-microfluidics. However, the quality DIHM reconstructions is compromised by twin-image noise, posing significant challenge. Conventional methods mitigating this noise involve complex hardware setups or time-consuming algorithms often limited effectiveness. In work, we propose UTIRnet, deep learning solution fast, robust, universally applicable suppression, trained exclusively on numerically generated datasets. The availability open-source UTIRnet codes facilitates its implementation in various systems without need extensive experimental training data. Notably, our network ensures consistency reconstruction results input holograms, imparting physics-based foundation enhancing reliability compared to conventional approaches. Experimental verification was conducted among others live neural glial culture migration sensing, which crucial neurodegenerative disease research.

Язык: Английский

---

Single-shot inline holography using a physics-aware diffusion model

Optics Express, Год журнала: 2024, Номер 32(6), С. 10444 - 10444

Опубликована: Фев. 27, 2024

Among holographic imaging configurations, inline holography excels in its compact design and portability, making it the preferred choice for on-site or field applications with unique requirements. However, effectively reconstruction from a single-shot measurement remains challenge. While several approaches have been proposed, our novel unsupervised algorithm, physics-aware diffusion model digital (PadDH), offers distinct advantages. By seamlessly integrating physical information pre-trained model, PadDH overcomes need training dataset significantly reduces number of parameters involved. Through comprehensive experiments using both synthetic experimental data, we validate capabilities reducing twin-image contamination generating high-quality reconstructions. Our work represents significant advancements by harnessing full potential prior.

Язык: Английский

---

Multi-prior physics-enhanced neural network enables pixel super-resolution and twin-image-free phase retrieval from single-shot hologram

Opto-Electronic Advances, Год журнала: 2024, Номер 7(9), С. 240060 - 240060

Опубликована: Янв. 1, 2024

Язык: Английский

---

Ultrafast radiographic imaging and tracking: An overview of instruments, methods, data, and applications

Nuclear Instruments and Methods in Physics Research Section A Accelerators Spectrometers Detectors and Associated Equipment, Год журнала: 2023, Номер 1057, С. 168690 - 168690

Опубликована: Сен. 22, 2023

Ultrafast radiographic imaging and tracking (U-RadIT) use state-of-the-art ionizing particle light sources to experimentally study sub-nanosecond transients or dynamic processes in physics, chemistry, biology, geology, materials science other fields. These are fundamental modern technologies applications, such as nuclear fusion energy, advanced manufacturing, communication, green transportation, which often involve one mole more atoms elementary particles, thus challenging compute by using the first principles of quantum physics forward models. One central problems U-RadIT is optimize information yield through, e.g. high-luminosity X-ray sources, efficient detectors, novel methods collect data, large-bandwidth online offline data processing, regulated underlying statistics, computing power. We review highlight recent progress in: (a.) Detectors high-speed complementary metal-oxide semiconductor (CMOS) cameras, hybrid pixelated array detectors integrated with Timepix4 application-specific circuits (ASICs), digital photon detectors; (b.) modalities phase contrast imaging, diffractive four-dimensional (4D) tracking; (c.) algorithms neural networks machine learning, (d.) Applications ultrafast material XFELs, synchrotrons laser-driven sources. Hardware-centric approaches optimization constrained detector properties, low signal-to-noise ratio, high cost long development cycles critical hardware components ASICs. Interpretation experimental including comparisons models, frequently hindered sparse measurements, model measurement uncertainties, noise. Alternatively, make increasing learning algorithms, implementations compressed sensing. Machine artificial intelligence approaches, refined information, may also contribute significantly interpretation, uncertainty quantification optimization.

Язык: Английский

---

人工智能定量相位成像：从物理到算法再到物理（内封面文章·特邀）

Infrared and Laser Engineering, Год журнала: 2025, Номер 54(2), С. 20240490 - 20240490

Опубликована: Янв. 1, 2025

---