Article
Engineering, Biomedical
Jingfei He, Peng Gao, Xunan Zheng, Yatong Zhou, Hao He
Summary: The denoising of magnetic resonance (MR) images is crucial for accurate organ tissue information recognition in medical diagnosis. This paper presents a 3D MR image denoising method based on weighted tensor nuclear norm minimization using balanced nonlocal patch tensors. A high-dimensional adaptive clustering technology is developed to construct highly correlated 3D MR image patches into nonlocal patch tensors, taking advantage of nonlocal self-similarity and correlation among different dimensions. The low-rank characteristics of the generated tensor are utilized for denoising using a tensor singular value decomposition framework, and the accuracy is improved by considering local noise level information and constructing balanced 3D nonlocal patch tensors.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2022)
Article
Computer Science, Information Systems
Run Tian, Guiling Sun, Xiaochao Liu, Bowen Zheng
Summary: An optimized scheme for edge detection is proposed in this study, which combines the WNNM image denoising algorithm with the Sobel edge detection algorithm to enhance anti-noise performance. Experimental results demonstrate that the algorithm can achieve better detection outcomes when processing noisy images.
Article
Computer Science, Hardware & Architecture
Lei Zhang, Cong Liu
Summary: The proposed AWTD method combines adaptive weight tensor and consideration of spatial and spectral information, achieving good performance in image denoising tasks.
Article
Engineering, Electrical & Electronic
Junwei Xu, Yuli Fu, Youjun Xiang
Summary: By exploring the nonlocal self-similarity of images, the WNNM based image denoising algorithm has achieved competitive performance; however, it requires more time in a larger search space. In this paper, an edge map-guided strategy is proposed to accelerate the MS-WNNM based image denoising, which trims off redundant patches and retains informative ones, achieving comparable denoising with less runtime.
DIGITAL SIGNAL PROCESSING
(2023)
Article
Environmental Sciences
Wenfeng Kong, Yangyang Song, Jing Liu
Summary: This paper proposes the framelet-based three-modal tensor nuclear norm (F-3MTNN) denoising model, which combines the basis redundancy of framelet and the low-rank characteristics of HSI to achieve a more flexible and comprehensive denoising effect.
Article
Computer Science, Information Systems
Xue Guo, Feng Liu, Xuetao Tian
Summary: With the increasing railway mileage, technologies such as vision and optical-fiber sensing are widely used in automatic railway inspections. A vision-based system called on-board track inspection system (OBTIS) is designed for this purpose. This study focuses on improving the defect detection accuracy of images collected by OBTIS, by modeling the noise in the images and proposing a denoising model named RA-WNNM. Experimental results demonstrate the superiority of RA-WNNM in denoising and image classification.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Geochemistry & Geophysics
Yuan Luo, Xiaorun Li, Shuhan Chen, Chaoqun Xia, Liaoying Zhao
Summary: This article proposes an improved infrared small target detection method that effectively utilizes spatial and temporal information, introduces new sparse prior map and tubewise sparse regularization term to simultaneously preserve targets and suppress background.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Zhang-Lei Shi, Xiao Peng Li, Chi-Sing Leung, Hing Cheung So
Summary: This study introduces an algorithm for portfolio optimization that explicitly controls the cardinality of the portfolio through a non-convex optimization problem. Results on real-world datasets demonstrate the superiority of the proposed algorithm over several existing algorithms.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Mathematics, Applied
Hanmei Yang, Jian Lu, Heng Zhang, Ye Luo, Jianwei Lu
Summary: Image denoising is an essential preprocessing step for various imaging technologies. Nonlocal low rank matrix approximation has been successfully applied to image denoising, but existing models ignore the correlation among image patches and their performance is degraded when encountering heavy noise. To address this, we propose a field of experts regularized nonlocal low rank matrix approximation denoising model that integrates a global field of experts regularization, a fidelity term, and a nonlocal low rank constraint into a unified framework. Experimental results demonstrate that the proposed model has superior performance.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Songze Tang, Zhenqiu Shu
Summary: In this paper, an adaptive weighted residual and nuclear-norm regularization approach is proposed for handling mixed image noise and conducting face hallucination. By self-identifying the mixed noise and alleviating its effects on the coding process, this method is able to generate suitable coefficients in the presence of mixed noise and exhibits better performance than existing methods in experiments.
APPLIED INTELLIGENCE
(2023)
Article
Mathematics, Applied
Linwei Fan, Huiyu Li, Miaowen Shi, Zhen Hua, Caiming Zhang
Summary: This paper proposes a novel two-stage enhanced low-rank prior model (TSLR) for efficient image denoising, dividing the denoising process into contour restoration and detail restoration stages. By preserving details and using the weighted sum method, it effectively retains image details and improves denoising outcomes.
JOURNAL OF SCIENTIFIC COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Ying Shi, Tianyu Liu, Dong Hu, Chuan Li, Zhi Wang
Summary: Our proposed multi-channel optimization model demonstrates superior performance in denoising real-world color images, balancing noise level and preserving image details through the use of weighted Schatten p-norm and data fidelity terms.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Zekun Liu, Siwei Yu
Summary: This paper proposes an ADMM-based optimization algorithm to solve the MMV problem. The key innovation is the introduction of an l(2,0)-norm sparsity constraint, which differs from the widely used l(2,1)-norm constraint. The proposed algorithm is shown to solve a larger range of MMV problems even under adverse conditions, as demonstrated by comparisons with other algorithms using simulated examples.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2023)
Article
Computer Science, Information Systems
Kunhao Zhang, Yali Qin, Huan Zheng, Hongliang Ren, Yingtian Hu
Summary: A compressive sensing image reconstruction algorithm is proposed, which combines bilateral total variation and nonlocal low-rank regularization for better image reconstruction performance. Experimental results demonstrate the algorithm's superiority in image quality and peak signal-to-noise ratio compared to conventional methods.
Article
Engineering, Electrical & Electronic
Hao Wang, Wanying Zhang, Yuxin He, Wenming Cao
Summary: This paper proposes a novel short-term sparse portfolio optimization (SSPO) model based on pound 0 -norm. The model selects portfolios based on the short-term increasing potential of assets and introduces a pound 0 -norm constraint to control the maximum number of non-zero assets. The model allows for direct use of no-short-selling constraints and introduces a sparse regularization term to eliminate trivial trades in the system. An algorithm based on the alternating direction method of multipliers (ADMM) is developed to solve the non-convex optimization system and its convergence is investigated. The effectiveness of the approach is demonstrated through numerical experiments on real-world datasets.
Article
Geochemistry & Geophysics
Zhenghua Huang, Zifan Zhu, Qing An, Zhicheng Wang, Qin Zhou, Tianxu Zhang, Ali Saleh Alshomrani
Summary: This letter proposes a novel enhancement framework for remotely sensed images to correct luminance guided by weighted least squares (WLS), which separates the image into base and detail layers for enhancement. Experimental results show that the proposed method outperforms current techniques in contrast improvement and detail preservation.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Correction
Geochemistry & Geophysics
Zhenghua Huang, Zifan Zhu, Qing An, Zhicheng Wang, Qin Zhou, Tianxu Zhang, Ali Saleh Alshomrani
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Zhenghua Huang, Lei Wang, Qing An, Qin Zhou, Hanyu Hong
Summary: This paper proposes a new enhancement framework for remotely sensed images called Global-Local Enhancement Network (GLE-Net). The framework corrects the intensity of the images by learning extra information from collected training data, improving both the low-frequency and detail components, and producing high-quality images. The GLE-Net method performs well in preserving brightness and fine details, outperforming existing techniques.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Optics
Xuan Li, Yuhang Xu, Zhenghua Huang, Lei Ma, Zhi Yang
Summary: A novel visual privacy measurement framework based on analytic hierarchy process is proposed in this paper to achieve instance-level visual privacy measurement using multi-cues. Experimental results show that the proposed methods provide reliable visual cues and the measurement results are highly consistent with human evaluation results.
Article
Geochemistry & Geophysics
Zhenghua Huang, Zifan Zhu, Zhicheng Wang, Yu Shi, Hao Fang, Yaozong Zhang
Summary: This article proposes a deep gradient descent network (DGDNet) method, which stabilizes the gradient descent model by estimating the learning rate and residual part. The learning rate is designed using the eigenvalues of the Hessian matrix of remotely sensed images and their local weighted factor, while the residual part is calculated by a U-shaped network. The experimental results show that DGDNet can efficiently obtain a stable solution and achieve competitive denoising performance.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Optics
Ziling Lu, Zhenghua Huang, Qiong Song, Hongyin Ni, Kun Bai
Summary: A novel joint constraint local contrast measure algorithm is proposed in this research to detect small IR targets. It combines two contrast measures, namely, ratio-difference measure to enhance the small target and suppress the background, and constrained difference measure to suppress clutter and enhance the target. The proposed method achieves better detection performance than other state-of-the-art methods, as demonstrated in experiments on real IR images and sequences.
Article
Environmental Sciences
Zhenghua Huang, Zifan Zhu, Yaozong Zhang, Zhicheng Wang, Biyun Xu, Jun Liu, Shaoyi Li, Hao Fang
Summary: This paper proposes a model-driven deep denoising (MD3) scheme that combines sparse-coding-model-based and deep-neural-network-based approaches for noise reduction. The MD3 model is decomposed into subproblems, which are solved by learnable denoisers to efficiently produce a stable solution. The proposed MD3 approach achieves effective and efficient denoising performance, surpassing other advanced methods in preserving rich textures.
Article
Environmental Sciences
Zhenghua Huang, Zifan Zhu, Zhicheng Wang, Xi Li, Biyun Xu, Yaozong Zhang, Hao Fang
Summary: In this paper, a novel denoising strategy called Dual Denoiser Driven Convolutional Neural Networks (D(3)CNNs) is proposed to remove both random and stripe noise, which has been proven to be effective and outperforms the state-of-the-art methods.
Article
Physics, Multidisciplinary
Xi Li, Jingwei Han, Quan Yuan, Yaozong Zhang, Zhongtao Fu, Miao Zou, Zhenghua Huang
Summary: This paper proposes a novel end-to-end denoising network called Fourier embedded U-shaped network (FEUSNet), which learns Fourier features and embeds them into a U-shaped network to reduce noise while preserving multi-scale details.
Article
Engineering, Electrical & Electronic
Zifan Zhu, Chen Huang, Menghan Xia, Biyun Xu, Hao Fang, Zhenghua Huang
Summary: In this article, a novel recurrent feature refinement network based on short-term dense connection modules for optical flow estimation (RFRFlow) is proposed. This network utilizes global context information and multiscale correlation information to achieve better performance and object shape preservation compared to existing methods.
IEEE SENSORS JOURNAL
(2023)
Article
Instruments & Instrumentation
Yu Shi, Zhigao Huang, Zhenghua Huang, Xia Hua, Hanyu Hong, Lirong Li
Summary: In this paper, a passive millimeter-wave image restoration network called HINRDNet is proposed to address the issues of blur and noise in millimeter-wave images. The network utilizes local features through the design of Half Instance Normalization Block and employs HINBlock at different levels for image restoration. The fusion of high-level and low-level features for global feature cross fusion further enhances the performance of the proposed method, which outperforms existing state-of-the-art methods.
INFRARED PHYSICS & TECHNOLOGY
(2023)
Article
Instruments & Instrumentation
Biyun Xu, Shaoyi Li, Shaogang Yang, Haoran Wei, Chaojun Li, Hao Fang, Zhenghua Huang
Summary: This paper proposes a multi-stage progressive visible and infrared image fusion strategy (MSPIF) to improve image quality. The strategy includes enhancing the visible image using a weighted fusion algorithm, decomposing the infrared and enhanced visible images using Retinex_Net, decomposing the reflectance components using discrete wavelet transform, and fusing the components using weighted information entropy and local energy strategies. The proposed MSPIF achieves good results with structures preservation and outperforms existing methods.
INFRARED PHYSICS & TECHNOLOGY
(2023)
Article
Environmental Sciences
Ziling Lu, Zhenghua Huang, Qiong Song, Kun Bai, Zhengtao Li
Summary: In this paper, a novel model for infrared small-target detection is proposed. It utilizes the Laplace operator and local features to address the issues of complex backgrounds and clutter. Experimental results demonstrate that the model achieves excellent performance in suppressing strong edge clutter and estimating small targets.
Article
Geochemistry & Geophysics
Xuan Li, Yuhang Xu, Lei Ma, Zhenghua Huang, Haiwen Yuan
Summary: Salient object detection plays a crucial role in remote sensing applications, and current methods mainly rely on large pixel-wise datasets for training, which is time-consuming and limits their flexibility. This research proposes an efficient approach for salient object detection in optical remote sensing images based on easily accessible weak supervision. The proposed framework utilizes progressive attention and scribble supervision to better locate challenging salient objects and restore their details, outperforming other weakly supervised methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Zhenghua Huang, Zhicheng Wang, Zifan Zhu, Yaozong Zhang, Hao Fang, Yu Shi, Tianxu Zhang
Summary: This letter proposes a deep image denoising scheme called deep low-rank prior (DLRP) that utilizes the low-rank property and deep convolutional neural network (DCNN) to address the issue of additive white Gaussian noise (AWGN) degradation in remotely sensed images. Experimental results demonstrate that DLRP outperforms state-of-the-art methods in terms of performance.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)