Article
Computer Science, Artificial Intelligence
Fei Ye, Zebin Wu, Xiuping Jia, Jocelyn Chanussot, Yang Xu, Zhihui Wei
Summary: This study proposes a hyperspectral image super-resolution algorithm based on tensor factorization and Bayesian methods. By constructing nonlocal patch tensors and utilizing automatic relevance determination, the model can simultaneously explore global spectral correlation and nonlocal spatial similarity, and adaptively infer the latent rank of the tensors, thus improving fusion performance.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Yinjian Wang, Wei Li, Na Liu, Yuanyuan Gui, Ran Tao
Summary: This study proposes a novel Bayesian sparse learning-based tensor ring (TR) fusion model, named FuBay. By specifying a hierarchical sparsity-inducing prior distribution, this method becomes the first fully Bayesian probabilistic tensor framework for hyperspectral fusion. Extensive experiments demonstrate its superior performance when compared with state-of-the-art methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Instruments & Instrumentation
Jian Long, Yuanxi Peng, Jun Li, Longlong Zhang, Yunpeng Xu
Summary: The article introduces a fast low tensor multi-rank (FLTMR) hyperspectral super-resolution method, which optimizes the estimation of spectral coefficients, computation time of the algorithm, and applies iterative attenuation coefficients to accelerate and achieve better estimation results. The proposed method shows efficient fusion of LR-HSI and HR-MSI, consuming less time while achieving improved performance.
INFRARED PHYSICS & TECHNOLOGY
(2021)
Article
Geochemistry & Geophysics
Cong Liu, Zhihao Fan, Guixu Zhang
Summary: This article proposes a novel single hyperspectral image super-resolution method by combining a trainable grouped joint tensor dictionary and a low-rank prior. The method addresses the issue of limited training samples and achieves superior performance compared to traditional and advanced methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Le Sun, Qihao Cheng, Zhiguo Chen
Summary: The study proposed an HSI super-resolution model based on spectral smoothing prior and tensor tubal row-sparse representation, termed SSTSR, which reconstructs HSI with high spatial resolution and spectral resolution through nonlocal priors, tensor decomposition, and regularization. Experimental results showed that the method outperformed many advanced HSI super-resolution methods.
Article
Environmental Sciences
Meng Cao, Wenxing Bao, Kewen Qu
Summary: A new joint regularized low-rank tensor decomposition method is proposed for hyperspectral image super-resolution to preserve the spatial and spectral structure. The method transforms hyperspectral data using Tucker decomposition and introduces graph regularization and total variational regularization constraints on the dictionary to effectively handle the high-dimensional data.
Article
Computer Science, Artificial Intelligence
Lei Zhang, Jiangtao Nie, Wei Wei, Yong Li, Yanning Zhang
Summary: The study proposes an unsupervised deep framework for blind hyperspectral image super-resolution, addressing the issue of unknown degeneration in both spatial and spectral domains.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Clemence Prevost, Ricardo A. Borsoi, Konstantin Usevich, David Brie, Jose C. M. Bermudez, Cedric Richard
Summary: In this paper, a coupled LL1 block-tensor decomposition is proposed to jointly solve the hyperspectral super-resolution problem and the unmixing problem of the underlying super-resolution image. Exact recovery conditions for the image and mixing factors are provided with consideration of spectral variability phenomenon between the observed low-resolution images. Two algorithms, one unconstrained and another subject to nonnegativity constraints, are proposed to solve the problems, and the performance of the proposed approach is showcased on synthetic and real images.
SIAM JOURNAL ON IMAGING SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Jize Xue, Yong-Qiang Zhao, Yuanyang Bu, Wenzhi Liao, Jonathan Cheung-Wai Chan, Wilfried Philips
Summary: This paper proposes a novel hyperspectral image super-resolution method that fully considers the spatial/spectral relationships between available HR-MSI and LR-HSI, achieving better performance on three benchmark datasets in terms of both visual and quantitative evaluation compared to state-of-the-art methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Automation & Control Systems
Xinya Wang, Qian Hu, Yingsong Cheng, Jiayi Ma
Summary: Hyperspectral image super-resolution aims to improve the spatial resolution of hyperspectral images by reconstructing high-resolution images from low-resolution observations. Recent advancements in deep learning have significantly contributed to the progress in this field, providing more reliable predicted results. However, there is a lack of comprehensive reviews and analysis focusing on the latest deep learning methods for hyperspectral image super-resolution.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2023)
Article
Geochemistry & Geophysics
Sen Jia, Shuangzhao Zhu, Zhihao Wang, Meng Xu, Weixi Wang, Yujuan Guo
Summary: With the rapid development of deep convolutional neural networks, super-resolution in hyperspectral images has made significant progress. However, current methods lack effective ways to extract spectral information and suffer from parameter redundancy and model complexity. In this study, we propose a diffused CNN approach that adds spectral convolutions into the enhanced convolutional neural block to better characterize spectral features and improve reconstruction efficiency through feature fusion and image enhancement.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Electrical & Electronic
Emma J. Reid, Lawrence F. Drummy, Charles A. Bouman, Gregery T. Buzzard
Summary: This paper presents a Multi-resolution Data Fusion algorithm that accurately interpolates low-resolution electron microscope data. The algorithm utilizes small amounts of unpaired high-resolution data to train a neural network denoiser and incorporates a Multi-Agent Consensus Equilibrium problem formulation to balance the denoiser with a forward model agent for fidelity to measured data.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2022)
Review
Computer Science, Information Systems
Na Liu, Wei Li, Yinjian Wang, Ran Tao, Qian Du, Jocelyn Chanussot
Summary: The ability of hyperspectral images (HSIs) to capture fine spectral discriminative information allows them to observe, detect, and identify objects with subtle spectral discrepancies. However, HSIs may not accurately represent the true distribution of ground objects due to environmental disturbances, atmospheric effects, and hardware limitations. These degradations significantly reduce the quality and usefulness of HSIs. Low-rank tensor approximation (LRTA) has gained attention in the HSI restoration community and is effective in addressing convex and non-convex inverse optimization problems. This survey provides a comprehensive technical assessment of LRTA for HSI restoration, covering topics such as denoising, fusion, destriping, inpainting, deblurring, and super-resolution.
SCIENCE CHINA-INFORMATION SCIENCES
(2023)
Article
Geography, Physical
Meilin Zhang, Guizhou Zheng, Zhiben Jiang, Qiqi Zhu, Linlin Wang, Qingfeng Guan
Summary: In this paper, a local-aware coupled network (LCNet) is proposed to address the challenging task of hyperspectral image super-resolution (HSI SR). LCNet adaptively learns the spectral response function (SRF) and point spread function (PSF) in the primary stage of the network to address the issue of unknown prior information. Experimental results demonstrate the superiority of LCNet in preserving both the texture details of MSI and the spectral characteristics of HSI.
GISCIENCE & REMOTE SENSING
(2023)
Article
Environmental Sciences
Jing Zhang, Minhao Shao, Zekang Wan, Yunsong Li
Summary: This study proposes a Multi-Scale Feature Mapping Network (MSFMNet) to adaptively learn the prior information of hyperspectral images (HSIs). By simplifying the network structure and designing multi-scale feature generation and fusion modules, MSFMNet aims to solve the issues of existing HSI super-resolution methods.
Article
Computer Science, Information Systems
Haidong Wang, Xuan He, Zhiyong Li, Jin Yuan, Shutao Li
Summary: This study proposes an end-to-end MOT network called joint detection and association network (JDAN) that can simultaneously perform object detection and data association, resulting in improved tracking performance by optimizing the overall task.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Jinyang Liu, Renwei Dian, Shutao Li, Haibo Liu
Summary: This study proposes a saliency guided deep-learning framework for pixel-level image fusion, which can simultaneously deal with different tasks and generate fusion images that are more in line with visual perception by extracting meaningful information through fusion weights.
INFORMATION FUSION
(2023)
Article
Computer Science, Artificial Intelligence
Xiaohui Wei, Ting Lu, Shutao Li
Summary: Graphs are crucial for improving graph-based machine learning methods, but existing approaches have limitations when dealing with multiple graphs. This article presents a novel method that can simultaneously capture global and local information, and achieves superior results in clustering tasks compared to previous state-of-the-art methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Environmental Sciences
Phonekham Hansana, Xin Guo, Shuo Zhang, Xudong Kang, Shutao Li
Summary: Heavy rains often cause serious floods in Laos, impacting agriculture, households, and the economy. Therefore, it is crucial to monitor the flooding to better understand its patterns and characteristics. This study analyzes the influence of flooding in Laos using multi-source data and applies various methods to detect flood areas, assess global impacts, forecast flood risk, and compare optical images. The findings reveal concentration of floods near the Mekong River with a decreasing trend over time, providing valuable insights for flood management and mitigation strategies in Laos. The validation results demonstrate notable performance across a five-year period.
Article
Computer Science, Artificial Intelligence
Fuyan Ma, Bin Sun, Shutao Li
Summary: Facial Expression Recognition (FER) in the wild is challenging due to various factors, but the proposed Visual Transformers with Feature Fusion (VTFF) approach demonstrates superior performance by leveraging attentional selective fusion (ASF) and modeling relationships with global self-attention. Extensive experiments on three facial expression datasets show that VTFF sets new state-of-the-art results. The method also exhibits promising generalization capability on cross-dataset evaluation.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Geochemistry & Geophysics
Zhuoyi Zhao, Xiang Xu, Jun Li, Shutao Li, Antonio Plaza
Summary: Nowadays, CNN-based DL models have gained popularity in HSIC and achieved high accuracy due to their hierarchical and nonlinear feature learning patterns. However, deeper network structures may demand more parameters and training samples. To overcome these problems, we propose a lightweight network model using the GSC module, which reduces parameters and is suitable for HSI data. Experimental results show that our model has low training cost and achieves competitive accuracy with fewer samples compared to existing models.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Ze Song, Xiaohui Wei, Xudong Kang, Shutao Li, Jinyang Liu
Summary: In this study, we propose a cross-temporal context learning network called CCLNet, which leverages intratemporal and intertemporal long-range dependencies to fully exploit cross-temporal context information. Our method achieves improved change detection performance, especially in complex and diverse changing scenes.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Anjing Guo, Renwei Dian, Shutao Li
Summary: In recent years, fusing a low-resolution hyperspectral image (LR-HSI) with a high-resolution multispectral image (HR-MSI) from different satellites has been proven to be an effective method for improving HSI resolution. However, the LR-HSI and HR-MSI obtained from different satellites may not satisfy existing observation models and it is difficult to register them. To address these issues, a deep-learning-based framework is proposed, which includes image registration, blur kernel learning, and image fusion. The proposed framework demonstrates superior performance in HSI registration and fusion accuracy through extensive experiments.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Haolong Fu, Shixun Wang, Puhong Duan, Changyan Xiao, Renwei Dian, Shutao Li, Zhiyong Li
Summary: Visible-infrared object detection aims to improve detector performance by fusing the complementarity of visible and infrared images. To overcome the limitation of existing methods that only utilize local intramodality information, we propose a feature-enhanced long-range attention fusion network (LRAF-Net) that leverages the long-range dependence between different modalities.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Geochemistry & Geophysics
Puhong Duan, Xudong Kang, Pedram Ghamisi, Shutao Li
Summary: Oil spill detection is a topic of increasing importance due to its negative impact on the environment and coastal communities. This study proposes an unsupervised method for oil spill detection using hyperspectral remote sensing images. The method utilizes isolation forest to estimate the probability of each pixel belonging to seawater or oil spills, and generates pseudolabeled training samples using clustering algorithms. The proposed method achieves superior detection performance compared to other state-of-the-art approaches.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Shuo Zhang, Puhong Duan, Xudong Kang, Yan Mo, Shutao Li
Summary: In this article, a feature-band-based unsupervised underwater target detection method (FBUD) is proposed, which uses the normalized difference water index (NDWI) and unmixing technique to find the target and background pixels. The spectral difference between the target and background is used to determine the optimal feature bands. The probability map of the underwater target can be obtained through a simple math operation on the feature bands. Additionally, a new UAV-borne hyperspectral image dataset named HNU-UTD is built for real-world underwater target detection, and the experimental results confirm the accuracy and effectiveness of the proposed method, which outperforms supervised detection methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Ze Song, Xudong Kang, Xiaohui Wei, Haibo Liu, Renwei Dian, Shutao Li
Summary: This paper proposes a two-stage focus scanning network for camouflaged object detection. It first designs a novel encoder-decoder module to determine the regions where focus areas may appear. Then, it utilizes multi-scale dilated convolution to obtain discriminative features with different scales in the focus areas, and a dynamic difficulty aware loss is designed to guide the network's attention to structural details. Experimental results on various benchmark datasets demonstrate the superior performance of the proposed method.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Geochemistry & Geophysics
Xudong Kang, Bin Deng, Puhong Duan, Xiaohui Wei, Shutao Li
Summary: Hyperspectral oil spill mapping is achieved through the self-supervised spectral-spatial transformer network (SSTNet), which can distinguish different types of oil spills. The method utilizes a transformer-based contrastive learning network to extract deep discriminative features, which are then transferred to a downstream classification network fine-tuned with a small number of labeled samples. Experimental results on a self-constructed hyperspectral oil spill database (HOSD) demonstrate that the proposed method outperforms several state-of-the-art oil spill classification techniques in discriminating different types of oil spills, such as thick oil, thin oil, sheen, and seawater.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Haibo Liu, Chenguo Feng, Renwei Dian, Shutao Li
Summary: A spatial-spectral transformer-based U-net (SSTF-Unet) approach is proposed in this paper to achieve the fusion of high-resolution hyperspectral images and high-resolution multispectral images by capturing the association between distant features and exploring the intrinsic information of the images. The method utilizes spatial and spectral self-attention and incorporates multiple fusion blocks for multiscale feature fusion.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Qiya Song, Bin Sun, Shutao Li
Summary: Automatic speech recognition (ASR) is a crucial interface in intelligent systems, but its performance is often affected by external noise. Audio-visual speech recognition (AVSR) utilizes visual information to enhance ASR in noisy conditions. This article proposes a multimodal sparse transformer network (MMST) that incorporates sparse self-attention mechanism and motion features to improve AVSR performance.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)