Article
Chemistry, Analytical
Piotr Jozwik-Wabik, Krzysztof Bernacki, Adam Popowicz
Summary: Monochromatic images are often affected by noise, which reduces the quality of the results. Deterministic algorithms like Non-Local-Means and Block-Matching-3D are commonly used to reduce noise. This article focuses on using machine learning to denoise monochromatic images, even without access to noise-free data, and demonstrates that ML-based methods can achieve high performance.
Article
Computer Science, Information Systems
Jiayi Ma, Chengli Peng, Xin Tian, Junjun Jiang
Summary: This paper proposes a new deep network architecture, named DBDnet, for image denoising. It generates a coarse noise map and gradually updates it through a boosting function. Experimental results show that DBDnet outperforms state-of-the-art methods in various image denoising tasks.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Artificial Intelligence
Yacov Hel-Or, Gil Ben-Artzi
Summary: This study provides a comprehensive explanation of the principles and optimization methods of wavelet denoising, suggesting that the shape of shrinkage functions should be adjusted when using redundant bases. The relationship between the transform used, optimal SFs, and the domains in which they are optimized is explored in detail. The study also shows that optimizing the SFs in the spatial domain is always better than or equal to optimizing them in the transform domain in subband optimization.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Erting Pan, Yong Ma, Xiaoguang Mei, Fan Fan, Jiayi Ma
Summary: Hyperspectral image denoising is a challenging problem, and prior knowledge about hyperspectral noise is essential for developing an effective denoising method. Most existing methods assume equal noise intensity in all bands, which contradicts practical HSIs and leads to unsatisfactory results. To address this, we propose a novel denoising framework called (N) over cap-Net, which utilizes the intrinsic properties of real HSI noise and employs a bootstrap mechanism for better denoising performance.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Information Systems
Yong Du, Guoqiang Han, Yinjie Tan, Chufeng Xiao, Shengfeng He
Summary: This paper introduces a dynamic dual learning network, DualBDNet, for blind image denoising. Unlike single-task prediction networks, DualBDNet investigates the relationship between residual estimation and nonresidual estimation, achieving better denoising performance.
IEEE TRANSACTIONS ON MULTIMEDIA
(2021)
Article
Mathematics, Applied
Zakaria Belhachmi, Thomas Jacumin
Summary: We introduce and discuss shape-based models for finding the best interpolation data in the compression of images with noise. The aim is to reconstruct missing regions by minimizing a data fitting term. We analyze the proposed models from two different points of view and provide numerical computations to confirm their usefulness in non-stationary PDE-based image compression.
COMMUNICATIONS IN NONLINEAR SCIENCE AND NUMERICAL SIMULATION
(2022)
Article
Computer Science, Information Systems
K. Shivarama Holla, Nokap Park, Bumshik Lee
Summary: In this paper, an edge-focused image denoising convolutional neural network is proposed to restore noisy images corrupted with additive white Gaussian noise (AWGN). The network obtains edge information for an input image using a simple Laplacian edge operator and performs image and edge processing using parallel residual convolutional blocks. The processed image and edge features are then mapped into the restored image using a single convolutional layer. Experimental results show that the proposed model is more effective in preserving textures and edges while removing noise compared to conventional methods, achieving up to 0.8 dB higher PSNR and 0.05 higher SSIM on a real image dataset.
Article
Computer Science, Artificial Intelligence
Ahlad Kumar, M. Omair Ahmad, M. N. S. Swamy
Summary: This paper introduces a new regularization term, LF-GGVF, based on L1-norm for image denoising, which improves denoising performance by utilizing fractional order variational method and overlapping group sparsity. The approximation of fractional order derivatives using Riemann-Liouville derivative boosts denoising performance within the optimization framework. Numerical optimization is carried out using ADMM and split Bregman techniques, demonstrating better noise suppression and edge preservation compared to other methods in experimental validation.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Jielin Jiang, Xiangming Hong, Yingnan Zhao, Xiaonglong Xu, Yan Cui
Summary: Convolutional neural networks (CNNs) have shown impressive performance in removing additive white Gaussian noise, but are unsatisfactory in removing mixed noise due to limited receptive field. Recent approaches using attention mechanism (AM) to capture global information still lack computational efficiency. This paper proposes a novel model called simple dual attention mechanism UNet (SDAUNet) for mixed noise removal, which uses UNet architecture to acquire multi-scale features and a simple dual attention convolutional block to capture global features efficiently. Experimental results demonstrate the superiority of SDAUNet in measurement metrics and visual performance compared to other state-of-the-art methods.
IET IMAGE PROCESSING
(2023)
Article
Chemistry, Analytical
Yong Li, Chenguang Liu, Xiaoyu You, Jian Liu
Summary: In this paper, a white Gaussian noise estimation algorithm based on pixel-level low-rank, low-texture subblocks and principal component analysis is proposed. The algorithm utilizes adaptive clustering and eigenvalue selection methods to improve the accuracy and robustness of noise level estimation.
Article
Mathematics, Applied
Zhihui Tu, Jian Lu, Hong Zhu, Huan Pan, Wenyu Hu, Qingtang Jiang, Zhaosong Lu
Summary: In this paper, we propose a new nonconvex low-rank tensor approximation method to eliminate complex noise in hyperspectral images. By leveraging the low-rank prior and exploring the intrinsic structure of the underlying HSI, the proposed method outperforms state-of-the-art methods in terms of denoising performance.
Article
Engineering, Electrical & Electronic
Ali Abbasian Ardakani, Afshin Mohammadi, Fariborz Faeghi, U. Rajendra Acharya
Summary: This study aims to evaluate 67 denoising filters and select the best one for ultrasound image denoising. A new filter evaluation method, Rank Analysis, was introduced and utilized. The best filter identified was the Spatial correlation (SCorr) filter.
INTERNATIONAL JOURNAL OF IMAGING SYSTEMS AND TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Rajesh Bhatt, Naren Naik, Venkatesh K. Subramanian
Summary: In image processing, the structural similarity index (SSIM) has emerged as a popular perceptual measure for image quality assessment, but its non-convex nature makes it challenging to apply in model based applications. Our research aims to discuss issues associated with SSIM and attempts to address them by developing a generalized convex framework for various image processing tasks.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Information Systems
Prabhishek Singh, Manoj Diwakar, Reena Gupta, Sarvesh Kumar, Alakananda Chakraborty, Eshan Bajal, Muskan Jindal, Dasharathraj K. Shetty, Jayant Sharma, Harshit Dayal, Nithesh Naik, Rahul Paul
Summary: Medical imaging is a complex process that captures images created by X-rays, ultrasound imaging, angiography, etc. The proposed work addresses the challenge to eliminate Gaussian additive white noise from CT images, achieving commendable efficiency in denoising.
Article
Computer Science, Artificial Intelligence
Dong Zhao, Lei Liu, Fanhua Yu, Ali Asghar Heidari, Mingjing Wang, Guoxi Liang, Khan Muhammad, Huiling Chen
Summary: By enhancing the selection mechanism of the ACOR method and introducing random spare strategy and chaotic intensification strategy, the convergence speed and accuracy can be significantly improved, effectively avoiding local optima. Through a series of experiments, these improved methods demonstrate superior performance in problem-solving, and compared to other techniques, RCACO has a more reliable ability to step out of local optima.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Shiqin Wang, Xin Xu, Lei Liu, Jing Tian
JOURNAL OF REAL-TIME IMAGE PROCESSING
(2020)
Article
Engineering, Electrical & Electronic
Yong Thiang Ng, Chung Ming Huang, Qing Tao Li, Jing Tian
SIGNAL IMAGE AND VIDEO PROCESSING
(2020)
Article
Engineering, Biomedical
Siddharth Pandey, Pranshu Ranjan Singh, Jing Tian
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2020)
Article
Computer Science, Artificial Intelligence
Sougata Deb, Youheng Ou Ynag, Matthew Chin Heng Chua, Jing Tian
JOURNAL OF AMBIENT INTELLIGENCE AND HUMANIZED COMPUTING
(2020)
Article
Environmental Sciences
Jayanthi Rajarethinam, Joel Aik, Jing Tian
INTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH
(2020)
Article
Construction & Building Technology
Eugene Chian, Weili Fang, Yang Miang Goh, Jing Tian
Summary: The study proposes two computer vision-based detection approaches to automatically detect missing barricades on construction sites, aiming to improve efficiency and reduce labor-intensive tasks. The results show that one of the approaches, MODA, has better performance and several implementation advantages compared to MCA, with an average precision of 57.9% and an average recall of 73.6%. These methods can help site managers take prompt action to mitigate the risks of falls from height accidents.
AUTOMATION IN CONSTRUCTION
(2021)
Article
Computer Science, Interdisciplinary Applications
Yang Miang Goh, Jing Tian, Eugene Yan Tao Chian
Summary: This study proposes a computer vision-based smart monitoring system to automatically detect worker breaching safe distancing rules on construction sites. Through a case study, it is demonstrated that monitoring of safe distancing can be automated using this approach, and the anchorless detection model CenterNet outperforms current state-of-the-art approaches in the real-time detection of workers.
COMPUTERS & INDUSTRIAL ENGINEERING
(2022)
Article
Engineering, Electrical & Electronic
Yazhi Zhao, Gui Peng David Yam, Jiahao Lu, Zhen-Peng Bian, Jing Tian
Summary: Pedestrian attribute recognition aims to recognize human attributes based on the appearance of pedestrians in images. This paper proposes a new approach called FLSRNet to tackle the challenges of variety, ambiguity, and imbalanced distribution of attributes. The approach includes label smoothing mechanism and focal mechanism to improve the performance.
SIGNAL IMAGE AND VIDEO PROCESSING
(2022)
Article
Engineering, Biomedical
Jing Tian
Summary: This study evaluates the adversarial vulnerability of deep neural network-based gait event detection models by generating adversarial examples. The experimental results show that adversarial examples significantly reduce the performance of the models.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2022)
Article
Chemistry, Analytical
Hesen Feng, Lihong Ma, Jing Tian
Summary: This paper proposes an image super-resolution method that improves image reconstruction performance through regularized patterns and dynamic convolution kernel generation. Experimental results show that this method outperforms other methods in terms of PSNR and SSIM performance.
Article
Chemistry, Analytical
Tianming Xie, Zhonghao Zhang, Jing Tian, Lihong Ma
Summary: In this paper, a novel target-aware token design for transformer-based object detection is proposed. The proposed method introduces a target-aware sampling module and a target-aware key-value matrix to tackle the target attribute diffusion challenge. Experimental results show that the proposed Focal DETR achieves superior performance over existing models on the COCO object-detection benchmark dataset.
Editorial Material
Chemistry, Analytical
Jing Tian
Article
Health Care Sciences & Services
Yubin Wu, Qianqian Lin, Mingrun Yang, Jing Liu, Jing Tian, Dev Kapil, Laura Vanderbloemen
Summary: This paper proposes a computer vision-based approach for grading yoga poses, using contrastive skeleton feature representations. The approach extracts skeleton keypoints from input images and encodes them using a contrastive triplet method. The paper also suggests a novel strategy for composing contrastive examples in pose feature encoding. Extensive experiments demonstrate the superior performance of the proposed approach.
Proceedings Paper
Computer Science, Artificial Intelligence
Longxi Li, Hesen Feng, Bing Zheng, Lihong Ma, Jing Tian
Summary: The study aims to improve the interpretability of mapping in single image super-resolution reconstruction through variable local dense blocks and a dense in dense network, achieving successful results. This approach demonstrates superior performance in experiments compared to most state-of-the-art methods in terms of accuracy and visual perception.
2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)
(2021)
Article
Computer Science, Artificial Intelligence
Yan Tao Eugene Chian, Jing Tian
Summary: This paper proposes a defect detection approach using statistical data fusion to break large images into smaller patches and use neural networks to detect defects, enhancing system performance and robustness. The method is evaluated using three benchmark datasets, demonstrating superior performance in individual patch inspection and whole image inspection.
