Article
Computer Science, Artificial Intelligence
Xianghai Wang, Xinying Wang, Ruoxi Song, Xiaoyang Zhao, Keyun Zhao
Summary: To improve the reliability and accuracy of optical imaging, this paper proposes a novel multi-hierarchical cross transformer for hyperspectral and multispectral image fusion. The proposed method extracts multi-scale features of HSI and MSI and performs cross-modality information interaction between them using a multi-hierarchical cross transformer. Experimental results show the superior performance of the proposed method on multiple datasets.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Environmental Sciences
Yulei Wang, Qingyu Zhu, Yao Shi, Meiping Song, Chunyan Yu
Summary: An improved LSE-SFIM algorithm is proposed in this study, combining least square estimation (LSE) with SFIM to enhance spatial information quality in the fused image. By using four spatial filters to extract fine spatial information from the MSI image, the experimental results show that the improved LSE-SFIM algorithm performs better than traditional SFIM.
Article
Engineering, Electrical & Electronic
Jiabao Li, Yuqi Li, Chong Wang, Xulun Ye, Wolfgang Heidrich
Summary: In this paper, an unsupervised blind fusion network is proposed to reconstruct high-resolution HSIs from a single pair of HSI and RGB images, without known degradation models or any training data. The method utilizes unrolling network and coordinate encoding to provide state-of-the-art HSI reconstruction, and accurately estimates the degradation parameters through neural representation and implicit regularization. Experimental results demonstrate its effectiveness in simulations and real experiments, outperforming other state-of-the-art fusion methods on popular HSI datasets.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)
Article
Geochemistry & Geophysics
Zhixi Feng, Xuehu Liu, Shuyuan Yang, Kai Zhang, Licheng Jiao
Summary: Most existing classification methods for hyperspectral images (HSIs) rely on complicated and large deep neural network (DNN) models, which suffer from limited training samples and high computational costs in real scenarios. To address these issues, we propose a simple spectral hierarchical feature fusion and selection network (HFFSNet) that utilizes 1-D grouped convolution for dimensionality reduction and multilevel feature extraction. The multilevel features are fused using the soft attention mechanism to assist adaptive feature selection, and the selected features are further fused to enhance the overall feature representation. Extensive experimental results on three hyperspectral datasets demonstrate the effectiveness of our proposed network.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Remote Sensing
Chenming Li, Tingting Fan, Zhonghao Chen, Hongmin Gao
Summary: This paper proposes a lightweight convolutional neural network structure called DSD-HAFF, which improves hyperspectral image classification performance by constructing global dense dilated CNN branches and a hierarchical attention feature fusion branch. The structure can fully incorporate hierarchical features and significantly reduce network parameters.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2022)
Article
Engineering, Biomedical
Yuncong Feng, Jie Wu, Xiaohan Hu, Wenjuan Zhang, Guishen Wang, Xiaotang Zhou, Xiaoli Zhang
Summary: This paper proposes a medical image fusion algorithm based on structural similarity detection, saliency detection, and bilateral texture filter to solve the issues of texture loss, low contrast, and pseudo-edges in medical image fusion. The experimental results show that the proposed algorithm outperforms 17 other algorithms.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Chemistry, Applied
Hai-Dong Yu, Li-Wei Qing, Dan-Ting Yan, Guanghua Xia, Chenghui Zhang, Yong-Huan Yun, Weimin Zhang
Summary: The study investigated the potential of two different hyperspectral imaging systems, Vis-NIR and NIR, in determining TVB-N contents in tilapia fillets during cold storage. By comparing the calibration models established with variable selection and data fusion methods, it was found that low-level fusion data based variable selection methods resulted in superior models. Mid-level fusion data, particularly based on CARS, achieved the best model. Ultimately, the study demonstrated the great feasibility of hyperspectral imaging combined with data fusion analysis for nondestructive evaluation of tilapia fillet freshness.
Article
Computer Science, Information Systems
Xiaosong Li, Fuqiang Zhou, Haishu Tan, Wanning Zhang, Congyang Zhao
Summary: Multimodal medical image fusion is an important technique for biomedical diagnosis, and a novel method with two-layer decomposition and local gradient energy operator is proposed in this paper to achieve better fusion performance and computational efficiency. Extensive experiments demonstrate the method's superiority over state-of-the-art methods in both visual quality and quantitative evaluation.
INFORMATION SCIENCES
(2021)
Article
Chemistry, Applied
Zeyi Cai, Zihong Huang, Mengyu He, Cheng Li, Hengnian Qi, Jiyu Peng, Fei Zhou, Chu Zhang
Summary: This study used hyperspectral imaging and a convolutional neural network with attention mechanism to identify the geographical origins of Radix Paeoniae Alba. The results showed that the combination of hyperspectral imaging and deep learning strategies had good prospects for identifying the origins of Radix Paeoniae Alba.
Article
Computer Science, Artificial Intelligence
Wei He, Naoto Yokoya, Xin Yuan
Summary: This paper proposes a fusion model for HSI reconstruction by combining CASSI and RGB measurements, utilizing low-dimensional spectral subspace property and patch processing strategy to improve the speed and quality of reconstruction. Extensive experiments demonstrate that the proposed method outperforms previous state-of-the-art methods and speeds up the reconstruction process significantly.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Engineering, Electrical & Electronic
Bing Tu, Wangquan He, Wei He, Xianfeng Ou, Antonio Plaza
Summary: This article proposes a global-local hierarchical weighted fusion end-to-end classification architecture that enhances the discrimination of spectral-spatial features. Experimental results demonstrate its competitiveness in terms of accuracy and generalization.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Sunsi Fu, Rushan Zheng, Xiong Chen
Summary: This article proposes an adaptive infrared and visible image fusion method based on visual saliency and hierarchical Bayesian (AVSHB), which preserves the highest similarity between fused images and source images. By utilizing a salient edge-preserving filter (SEPF) and an adaptive fusion scheme, AVSHB outperforms other fusion methods in both qualitative and quantitative evaluations.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Geochemistry & Geophysics
Xiyou Fu, Sen Jia, Meng Xu, Jun Zhou, Qingquan Li
Summary: In this letter, a novel sparsity constrained fusion method based on matrix factorization is proposed for fusing hyperspectral and multispectral images. By imposing l(1) norm constraint and inserting a prior, this method can effectively handle localized changes between multiplatform images.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Chemistry, Analytical
Jiangwei Li, Dingan Han, Xiaopan Wang, Peng Yi, Liang Yan, Xiaosong Li
Summary: A multi-sensor medical-image fusion technique, integrating useful information from single-modal images of the same tissue, is crucial for clinical diagnosis and treatment planning.
Article
Engineering, Electrical & Electronic
Hongmin Gao, Zhonghao Chen, Chenming Li
Summary: This paper introduces a deep learning-based hyperspectral image classification method, which achieves more effective performance improvement with a newly designed multiscale feature extraction network and feature fusion scheme.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)