Plug-and-play pixel super-resolution phase retrieval for digital holography

Chang, Xuyang; Bian, Liheng; Gao, Yunhui; Cao, Liangcai; Zhang, Jun

doi:10.1364/ol.458117

Cited by 20 publications

(13 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It extended microscopy's bandwidth to billion pixels through multiple illumination angles which correspond to different sub-regions in the Fourier domain (Supplementary Note 3). Different from the direct wavefront enhancement in KKR holography, we applied CI-CDNet to the iterative FPM reconstruction which is based on the large-scale plug-andplay (PNP) optimization framework [20,33] (Supplementary Note 2).…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Complex-domain enhancing neural network for large-scale coherent imaging

Bian

Chang

Zhao

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Large-scale computational imaging allows an unprecedented resolution beyond the space-bandwidth production (SBP) limit of the optical system. In coherent imaging (CI), the joint reconstruction of amplitude and phase further expands the information throughput and sheds light on label-free observation of biological samples at micro or even nano level. The existing large-scale CI techniques usually require scanning/modulation multiple times to guarantee measurement diversity and long exposure time to achieve a high signal-to-noise ratio (SNR). Such cumbersome procedure restricts clinic applications for rapid and low-phototoxicity cell imaging. In this work, a complex-domain enhancing neural network for large-scale coherent imaging termed CI-CDNet is reported for various large-scale CI modalities with outstanding reconstruction image quality and efficiency. CI-CDNet is able to exploit the latent coupling information between amplitude and phase, realizing multi-dimensional representation of complex wavefront. The cross-field characterization framework empowers strong generalization and robustness for various coherent modalities, allowing high-quality and efficient imaging under extremely few data volume and low exposure time. We applied CI-CDNet in various large-scale CI modalities including Kramers-Kronig-relations holography, Fourier ptychographic microscopy, and lensless coded ptychography. A series of simulations and experiments validate that CI-CDNet can reduce data volume requirement and exposure time by more than one order of magnitude. We further demonstrated that the high-quality reconstruction of CI-CDNet benefits the subsequent high-level semantic analysis.

show abstract

Section: Resultsmentioning

confidence: 99%

“…Large-scale CI techniques generally require certain types of diversity measurements in the spatial domain (e.g. lensless onchip system [16][17][18][19][20] ) or the Fourier domain (e.g. Fourier ptychography [2,3] ).…”

Section: Introductionmentioning

confidence: 99%

Complex-domain enhancing neural network for large-scale coherent imaging

Bian

Chang

Zhao

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…By directly incorporating the neural network as a regularization operator into optimization, PnP [18,46] alternatives iterations between the model-based fidelity and learning-based regularization (Fig. 8 a).…”

Section: Combining Deep Learning and Optimizationmentioning

confidence: 99%

Large-scale Global Low-rank Optimization for Computational Compressed Imaging

Li¹,

Xu²,

Chen³

et al. 2023

Preprint

View full text Add to dashboard Cite

Computational reconstruction plays a vital role in computer vision and computational photography. Most of the conventional optimization and deep learning techniques explore local information for reconstruction. Recently, nonlocal low-rank (NLR) reconstruction has achieved remarkable success in improving accuracy and generalization. However, the computational cost has inhibited NLR from seeking global structural similarity, which consequentially keeps it trapped in the tradeoff between accuracy and efficiency and prevents it from high-dimensional large-scale tasks. To address this challenge, we report here the global low-rank (GLR) optimization technique, realizing highly-efficient large-scale reconstruction with global self-similarity. Inspired by the self-attention mechanism in deep learning, GLR extracts exemplar image patches by feature detection instead of conventional uniform selection. This directly produces key patches using structural features to avoid burdensome computational redundancy. Further, it performs patch matching across the entire image via neural-based convolution, which produces the global similarity heat map in parallel, rather than conventional sequential block-wise matching. As such, GLR improves patch grouping efficiency by more than one order of magnitude. We experimentally demonstrate GLR's effectiveness on temporal, frequency, and spectral dimensions, including different computational imaging modalities of compressive temporal imaging, magnetic resonance imaging, and multispectral filter array demosaicing. This work presents the superiority of inherent fusion of deep learning strategies and iterative optimization, and breaks the persistent dilemma of the tradeoff between accuracy and efficiency for various large-scale reconstruction tasks.

show abstract

“…As a result, to tackle the ill-posedness, we introduce another regularization term R(x) for the reconstruction problem. Apart from the total variation function adopted in this paper, the regularization function can take many other forms, such as BM3D [42,[61][62][63] and deep denoiser priors [64,65].…”

Section: Regularized Inversionmentioning

confidence: 99%

Pixel Super-Resolution Phase Retrieval for Lensless On-Chip Microscopy via Accelerated Wirtinger Flow

Gao

Yang

Cao

2022

Cells

Self Cite

View full text Add to dashboard Cite

Empowered by pixel super-resolution (PSR) and phase retrieval techniques, lensless on-chip microscopy opens up new possibilities for high-throughput biomedical imaging. However, the current PSR phase retrieval approaches are time consuming in terms of both the measurement and reconstruction procedures. In this work, we present a novel computational framework for PSR phase retrieval to address these concerns. Specifically, a sparsity-promoting regularizer is introduced to enhance the well posedness of the nonconvex problem under limited measurements, and Nesterov’s momentum is used to accelerate the iterations. The resulting algorithm, termed accelerated Wirtinger flow (AWF), achieves at least an order of magnitude faster rate of convergence and allows a twofold reduction in the measurement number while maintaining competitive reconstruction quality. Furthermore, we provide general guidance for step size selection based on theoretical analyses, facilitating simple implementation without the need for complicated parameter tuning. The proposed AWF algorithm is compatible with most of the existing lensless on-chip microscopes and could help achieve label-free rapid whole slide imaging of dynamic biological activities at subpixel resolution.

show abstract

Plug-and-play pixel super-resolution phase retrieval for digital holography

Cited by 20 publications

References 19 publications

Complex-domain enhancing neural network for large-scale coherent imaging

Complex-domain enhancing neural network for large-scale coherent imaging

Large-scale Global Low-rank Optimization for Computational Compressed Imaging

Pixel Super-Resolution Phase Retrieval for Lensless On-Chip Microscopy via Accelerated Wirtinger Flow

Contact Info

Product

Resources

About