Article
Biochemical Research Methods
Phaneendra K. Yalavarthy, Sandeep Kumar Kalva, Manojit Pramanik, Jaya Prakash
Summary: Photoacoustic/Optoacoustic tomography aims to reconstruct maps of initial pressure rise caused by light pulse absorption in tissue, posing an ill-conditioned and under-determined problem with limited detection positions. A new inversion method integrating denoising procedure within iterative model-based reconstruction was developed to improve quantitative performance, with a non-local means step resulting in 2.5 dB signal-to-noise ratio improvement compared to TV-based reconstruction.
JOURNAL OF BIOPHOTONICS
(2021)
Article
Engineering, Electrical & Electronic
Ezgi Demircan-Tureyen, Mustafa E. Kamasak
Summary: A two-stage denoising framework was proposed for adaptive denoising using direction descriptors to guide structure tensor, enhancing sensitivity of STV. By efficiently capturing local geometry with a preprocessor, experiments demonstrate the benefits of incorporating directional information into STV.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2021)
Article
Mathematics
Kuan Li, Chun Huang, Ziyang Yuan
Summary: This paper investigates error estimations for total variation regularization, which is applied in various fields such as signal processing, imaging, and machine learning. The study focuses on properties of the minimizer for the TV regularization problem, including stability, consistency, and convergence rate. Both a priori and a posteriori rules are considered, with an improved convergence rate based on sparsity assumption. Additionally, the paper discusses non-sparsity conditions commonly found in practice, presenting corresponding convergence rates under mild conditions.
Article
Computer Science, Artificial Intelligence
Erich Kobler, Alexander Effland, Karl Kunisch, Thomas Pock
Summary: This paper combines the variational approach with deep learning, introducing a data-driven general-purpose deep variation regularizer. By using a convolutional neural network to extract local features, it allows for mathematical and stability analysis of inverse problems, achieving state-of-the-art results in various imaging tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Engineering, Electrical & Electronic
Yanyan Shi, Yuehui Wu, Meng Wang, Zuguang Rao, Bin Yang, Feng Fu, Yajun Lou
Summary: This article proposes a novel approach for image reconstruction of conductivity distribution in electrical impedance tomography (EIT). The approach introduces a fidelity term based on L-1 norm to stabilize the problem and enforce sparsity in the solution. It also introduces a hybrid penalty term combining first-order and high-order total variation to preserve sharp profiles and suppress the staircase effect.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Environmental Sciences
Bouthayna Msellmi, Daniele Picone, Zouhaier Ben Rabah, Mauro Dalla Mura, Imed Riadh Farah
Summary: This research focuses on remote sensing data analysis in high dimensional space formed by hyperspectral images, with a particular emphasis on spectral un-mixing methods. By introducing sub-pixel mapping techniques, the study aims to improve the spatial localization of different land cover classes in mixed pixels.
Article
Computer Science, Artificial Intelligence
Antonin Chambolle, Thomas Pock
Summary: This work introduces a general framework for discrete approximations of total variation in image reconstruction, showing consistency in the sense of Gamma-convergence. Algorithms for learning discrete total variation are proposed to achieve optimal reconstruction quality for specific image reconstruction tasks. The study reveals significant differences in learned discretizations for different tasks, indicating that there is no universal best discretization method for total variation.
SIAM JOURNAL ON IMAGING SCIENCES
(2021)
Article
Geochemistry & Geophysics
Xiaocheng Yang, Chaodong Lu, Jingye Yan, Lin Wu, Mingfeng Jiang, Lin Li
Summary: Synthetic aperture imaging radiometers (SAIRs) are powerful tools for high-resolution imaging, but their reconstruction process faces an ill-posed inverse problem. This letter presents a reweighted total variation (RTV) method to address the reconstruction issue in SAIRs, and numerical simulation experiments demonstrate the effectiveness and performance of the proposed method.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Optics
Guodong Chen, Sikun Li, Xiangzhao Wang
Summary: Optical proximity correction (OPC) is an important enhancement technique in optical lithography to improve image fidelity and process robustness, especially for advanced technology nodes. The proposed efficient OPC method based on virtual edge and mask pixelation with two-phase sampling effectively addresses imaging distortions and demonstrates superior modification efficiency.
Article
Geochemistry & Geophysics
Ghassem Khademi, Hassan Ghassemian
Summary: Pansharpening is an inverse problem that estimates a high-resolution multispectral image based on a low-resolution multispectral image and a panchromatic image. This letter proposes a variational model based on the observation model of satellite imaging system and second-order TGV, and solves it using a primal-dual algorithm. Compared to other variational methods, the proposed algorithm is more efficient and can preserve fine details of the panchromatic image in the sharpened multispectral image.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Jinming Duan, Xi Jia, Joseph Bartlett, Wenqi Lu, Zhaowen Qiu
Summary: In this work, we propose a variational model for image registration that combines SAD and a new total variation regularization term. We linearize the model to tackle the non-convexity and non-differentiability and use ADMM to solve the resultant convex optimization. Our proposed algorithm is more efficient than existing methods and performs better in experiments. The code is publicly available.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Dan Yao, Stephen McLaughlin, Yoann Altmann
Summary: This paper presents a scalable approximate Bayesian method for image restoration using Total Variation (TV) priors, with the ability to offer uncertainty quantification. The method utilizes the Expectation Propagation (EP) framework for estimation and automatically adjusts the regularization parameter through Expectation Maximization (EM).
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Xueyan Ding, Yafei Wang, Zheng Liang, Xianping Fu
Summary: This study proposes a unified total variation method based on an extended underwater imaging model, aiming to achieve good performance in underwater image enhancement by eliminating the dual-path light attenuation in underwater images.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Mathematics, Applied
Yiming Gao, Zhengmeng Jin, Xu Li
Summary: In this paper, a variational model based on dynamic optimal transportation and total generalized variation is proposed for CT reconstruction. It aims to reduce the radiation dose for patients and improve image quality and structure preservation. The final state image of the optimal transport problem is reconstructed through CT inversion, utilizing the given initial state as a template for structural information. The proposed model is solved numerically using a first-order algorithm based on the primal-dual method and demonstrated to achieve high-quality and structurally preserved image reconstruction in sparse-view CT.
Article
Mathematics
Tao Zou, Guozhang Li, Ge Ma, Zhijia Zhao, Zhifu Li
Summary: This paper proposes a new image restoration method D-TGV, which achieves a balance between detail preservation and denoising through the combination of derivative fidelity and total generalized variation. The ADMM algorithm is used to solve the model equations for rapid convergence. Experimental results show that the method outperforms currently established methods in terms of detail preservation and denoising.
Article
Optics
Yunping Zhang, Yanmin Zhu, Edmund Y. Lam
Summary: This paper proposes a one-stage network (OSNet) for 3D particle volumetric reconstruction, which allows high-resolution reconstruction with low latency and improved processing speed. The experimental results demonstrate the feasibility and robustness of the method under different particle concentrations and noise levels. Additionally, the study explores other applications of 3D particle tracking and the potential extension to similar computational imaging problems.
Article
Geochemistry & Geophysics
Peiyan Guan, Edmund Y. Lam
Summary: A hyperspectral pansharpening method called MDA-Net is proposed in this paper, which utilizes a three-stream structure and dual-attention guided fusion block to extract and fuse important information.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Optics
Zhou Ge, Haoyu Wei, Feng Xu, Yizhao Gao, Zhiqin Chu, Hayden K. -H. So, Edmund Y. Lam
Summary: This study presents a new approach to achieve fast autofocusing in microscopic imaging using neuromorphic event sensing technology. The proposed method can detect sparse brightness changes asynchronously and respond quickly to specimen movement, enabling autofocusing in just tens of milliseconds, which is thousands of times faster than current technologies. Experimental results demonstrate a substantial performance improvement and capability for biopsy specimen inspections.
OPTICS AND LASERS IN ENGINEERING
(2023)
Article
Optics
Xinming Guo, Yixuan Li, Jiaming Qian, Yuxuan Che, Chao Zuo, Qian Chen, Edmund Y. Lam, Huai Wang, Shijie Feng
Summary: Temporal phase unwrapping is an important method for recovering discontinuous surfaces or spatially isolated objects in fringe projection profilometry. Existing algorithms can be classified into three types, but all require additional fringe patterns of different frequencies to retrieve the absolute phase. Image noise limits the efficiency and speed of phase unwrapping. This work shows for the first time that a generalized framework using deep learning can effectively mitigate noise and enhance phase unwrapping reliability, without increasing the number of auxiliary patterns. The proposed method has great potential for developing powerful and reliable phase retrieval techniques.
Article
Optics
L. Song, Edmund Y. Lam
Summary: In this paper, a learning-based recursive dual alternating direction method of multipliers (RD-ADMM) is proposed for phase retrieval. The method solves the phase retrieval problem by solving the primal and dual problems separately. By designing a dual structure, the information embedded in the dual problem is utilized to help with solving the phase retrieval problem, and it is shown that the same operator can be used for regularization in both the primal and dual problems. Experiments demonstrate that our method is effective and robust, providing higher-quality results than other commonly-used phase retrieval methods for this setup.
Article
Optics
Yunping Zhang, Stanley H. H. Chan, Edmund Y. Lam
Summary: Digital holography (DH) is a powerful imaging modality that captures object wavefront information, but in low-light situations, the quality of holograms suffers. We developed a snapshot DH system that operates at an ultra-low photon level by using a quanta image sensor and computational reconstruction. Our method expands DH to the photon-starved regime and enables more advanced holography applications.
Article
Biology
Ziqi Zhang, Kelvin C. M. Lee, Dickson M. D. Siu, Michelle C. K. Lo, Queenie T. K. Lai, Edmund Y. Lam, Kevin K. Tsia
Summary: This article introduces an image-based approach that can quantify multiple biophysical fractal-related properties of single cells at subcellular resolution. This technique has high-throughput single-cell imaging performance and offers sufficient statistical power for cell classification, drug response assays, and cell-cycle progression tracking.
COMMUNICATIONS BIOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Shansi Zhang, Nan Meng, Edmund Y. Lam
Summary: This paper proposes an efficient Low-light Restoration Transformer (LRT) for LF images under low-light conditions. The method utilizes multiple heads to perform intermediate tasks within a single network, achieving progressive restoration from small scale to full scale. Experimental results show that the proposed method achieves state-of-the-art performance on low-light LF restoration with high efficiency.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Engineering, Electrical & Electronic
Pei Zhang, Zhou Ge, Li Song, Edmund Y. Lam
Summary: Bio-inspired neuromorphic cameras asynchronously record visual information of dynamic scenes by discrete events. Due to the high sampling rate, they are capable of fast motion capture without causing image blur, overcoming the drawbacks of frame-based cameras that produce blurry recordings of dynamic objects. However, highly sensitive neuromorphic cameras are also susceptible to interference, and can generate a lot of noise in response. Such noisy event data can dramatically degrade the event-based observations and analysis.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Correction
Optics
Boyi Huang, Jia Li, Bowen Yao, Zhigang Yang, Edmund Y. Lam, Jia Zhang, Wei Yan, Junle Qu
Article
Optics
Boyi Huang, Jia Li, Bowen Yao, Zhigang Yang, Edmund Y. Lam, Jia Zhang, Wei Yan, Junle Qu
Summary: Super-resolution optical imaging is crucial to studying cellular processes. This study presents a deep-learning-based super-resolution technique for confocal microscopy. The proposed algorithm, using a two-channel attention network, can handle changes in pixel size and imaging setup, and demonstrate live-cell super-resolution imaging of microtubules.
Article
Computer Science, Artificial Intelligence
Pei Zhang, Chutian Wang, Edmund Y. Lam
Summary: This paper proposes a new graph representation for event data and couples it with a Graph Transformer for accurate neuromorphic classification. Results show that this approach performs well in challenging realistic situations with limited computational resources and a small number of events.
Article
Engineering, Electrical & Electronic
Zhenxing Zhou, Vincent W. L. Tam, Edmund Y. Lam
Summary: Continuous sign language recognition (CSLR) is a challenging task that utilizes multiple input modalities to improve recognition accuracy. However, the modality differences make it difficult to define an integrative framework. To address this, a novel deep learning framework called CA-SignBERT is proposed, which utilizes multiple BERT models and a special cross-attention mechanism to analyze information from different modalities.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Computer Science, Information Systems
Chang Liu, Xiaojuan Qi, Edmund Y. Lam, Ngai Wong
Summary: This study introduces a novel event data representation (ACE) and a framework based on branched network (BET) for handling noisy, sparse, and nonuniform event data in both static and dynamic scenes. Experimental results show that the method performs well in various classification and recognition tasks, surpassing the accuracy of existing methods.
Article
Computer Science, Artificial Intelligence
Li Song, Zhou Ge, Edmund Y. Lam
Summary: This paper proposes a novel inverse imaging scheme, called D-ADMM, by combining duality and the alternating direction method of multipliers (ADMM). By solving the dual problem, this method can find the estimated nontrivial lower bound of the objective function and guide the primal iterations. In image super-resolution, D-ADMM-TV shows comparable or slightly better results compared to other methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)