Article
Computer Science, Software Engineering
R. Kosarevych, O. Lutsyk, B. Rusyn
Summary: This paper presents a novel method for detecting binary- and random-valued impulsive noise in contaminated images. The proposed noise detector classifies image pixels as corrupted or uncorrupted based on their relative position and the properties of random point patterns formed based on the noisy image. Extensive simulation experiments show that this detection approach outperforms many well-known methods.
Article
Computer Science, Interdisciplinary Applications
Yunlong Zhang, Xin Lin, Yihong Zhuang, Liyan Sun, Yue Huang, Xinghao Ding, Guisheng Wang, Lin Yang, Yizhou Yu
Summary: Synthesizing pathology-free images from pathological images is valuable for algorithm development and clinical practice. We propose a new discriminator called the segmentor to accurately locate lesions and improve the visual quality of synthetic images. The generated images are used for medical image enhancement and to address the low contrast problem in medical image segmentation. The proposed method outperforms existing methods in comprehensive experiments.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Geochemistry & Geophysics
Nicola Acito, Marco Diani, Michael Alibani, Giovanni Corsini
Summary: In this article, a new procedure is presented to locate residual defective pixels in hyperspectral images and an effective method is proposed to estimate missing radiance values. The procedure improves image quality and can be applied to various satellite hyperspectral sensors.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Information Systems
Mohd Rafi Lone, Ekram Khan
Summary: In this paper, a nearest neighbor filtering method is proposed for impulse noise reduction. The method leverages inter-pixel correlation in images and achieves superior denoising performance compared to existing methods. Additionally, the proposed method has low computational complexity.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Mathematics, Applied
Hossein Khodabakhshi Rafsanjani, Hossein Noori, Nasibe Naseri
Summary: In this paper, the ENI operator, which can distinguish well between impulse noise and signal, is analyzed and a efficient diffusion based method for impulse denoising is proposed. The experimental results confirm the efficiency of the proposed method according to subjective and objective criteria.
COMPUTERS & MATHEMATICS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Sagenela Vijaya Kumar, C. Nagaraju
Summary: The study introduces a novel filter for removing impulse noise from images in two stages: noise identification and denoising. By using a classifier trained with support vector neural network and genetic algorithm, the proposed hybrid filter achieved improved results.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2021)
Article
Computer Science, Information Systems
Cuicui Zhao, Jun Liu, Jie Zhang
Summary: This paper introduces a method for image restoration corrupted by a mixture of different types of noise, which outperforms existing single type denoisers in removing multiple noise types and preserving image details.
Article
Engineering, Electrical & Electronic
Shan Liao, Haoen Huang, Jiayong Liu, Xiuchun Xiao, Xiaoyang Li, Shubin Li
Summary: This paper proposes a modified Newton integration (MNI) algorithm to handle image deblurring problem with noise-tolerance ability and fast convergence performance. The algorithm considers the influence of noise perturbation from a control perspective, and theoretical analyses confirm its effectiveness. Multiple simulative experiments demonstrate the excellent image deblurring and noise-tolerance advantages of the MNI algorithm compared to other existing algorithms.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2021)
Article
Engineering, Multidisciplinary
Hengrui Cui, Zhoumo Zeng, Hui Zhang, Fenglong Yang
Summary: In this paper, a new DIC method is proposed by incorporating the 1-D and 2-D gradient operators into the same framework. By decomposing the gradient operator into gradient acquisition and gradient filtering, the systematic error is reduced to less than 10% of the original ICGN-DIC algorithm, improving robustness to noise variations. Validation with an open-source dataset shows that the proposed method can reduce the system error to less than 15%.
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
Optics
A. Baldi, P. M. Santucci, F. Bertolino
Summary: Digital Image Correlation (DIC) is an optical experimental method that assumes pixel intensity remains unchanged during motion. However, various noise sources can cause deviation from this assumption. DIC can be implemented using different formulations, such as the Forward-Additive Gauss-Newton (FA-GN) and the Inverse-Compositional Gauss-Newton formulation (IC-GN), which have differences in speed, converging characteristics, and noise robustness.
OPTICS AND LASERS IN ENGINEERING
(2022)
Article
Engineering, Multidisciplinary
Yusi Dai, Chunhua Yang, Hongqiu Zhu, Can Zhou, Kai Wang
Summary: This paper proposes a lens distortion correction method based on refined edge lines for the infrared images of electrolytic cell groups (ECGs). The method extracts ECG edges using morphological operators, separates the edge lines using the Hough transform method and an improved hierarchical method, refines the lines using the least squares method, and calibrates distortion parameters according to the refined lines. An interpolation strategy is then presented to restore pixel values of the corrected images. The proposed method achieves a root mean squared error of about 0.014 pixels for the corrected lines, demonstrating its advanced performance and practicality.
Article
Engineering, Multidisciplinary
Miyoun Jung
Summary: This article introduces a novel model for restoring color images degraded by blurring and mixed noise. The model effectively separates RV impulse and Gaussian noise using l(0)- and l(2)-norms, while maintaining edges with total variation regularization. Experimental results show the model's superior performance in image quality compared to existing models.
APPLIED MATHEMATICAL MODELLING
(2022)
Article
Computer Science, Artificial Intelligence
Yuchao Tang, Shirong Deng, Jigen Peng, Tieyong Zeng
Summary: This paper discusses the importance of image restoration with impulse noise and proposes an efficient iterative algorithm to solve nonconvex models arising from impulse noise.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Interdisciplinary Applications
Jingyan Xu, Frederic Noo
Summary: Deep-learning (DL) based CT image generation methods are usually evaluated using RMSE and SSIM, while conventional model-based image reconstruction (MBIR) methods are often evaluated using image properties such as resolution, noise, and bias. This study investigates the application of linearization to DL networks for efficient characterization of resolution and noise. The results show that network linearization works well under normal exposure settings and provides computational tools to implement this method for popularizing the physics-related image quality measures for DL applications.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)