Article
Remote Sensing
Tianlin Zhang, Hongzhen Chen, Shi Chen, Chunjiang Bian
Summary: Super-resolution reconstruction is important in remote-sensing image processing. To improve efficiency and performance, we propose an edge-enhanced efficient network (EESR) and construct a dataset for experimentation. Experimental results show that EESR outperforms other methods in terms of restoration accuracy and inference efficiency.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Kan Chang, Hengxin Li, Yufei Tan, Pak Lun Kevin Ding, Baoxin Li
Summary: This paper proposes an effective joint demosaicking and super-resolution method through a two-stage CNN architecture. Experimental results demonstrate the superiority of this method over other state-of-the-art demosaicking and super-resolution methods, with the added benefit of smaller model size.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Environmental Sciences
Ziqian Liu, Wenbing Wang, Qing Ma, Xianming Liu, Junjun Jiang
Summary: In this paper, a full 3D convolutional neural network (F3DUN) is proposed for hyperspectral image super-resolution (HSISR) tasks. The F3DUN model combined with the U-Net architecture achieves state-of-the-art performance on HSISR tasks by utilizing skip connections and multi-scale features. Additionally, the paper compares F3DUN with a 3D/2D mixed model and finds that the full 3D CNN has a larger capacity and can obtain better results with the same number of parameters. Furthermore, experimental results demonstrate that the full 3D CNN model is less sensitive to data scaling and outperforms the 3D/2D mixed model on small-scale datasets.
Article
Computer Science, Artificial Intelligence
Keunsoo Ko, Yeong Jun Koh, Soonkeun Chang, Chang-Su Kim
Summary: A novel light field super-resolution algorithm is proposed to improve the spatial and angular resolutions of light field images. Experimental results demonstrate that the algorithm outperforms state-of-the-art algorithms on various light field datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Saeed Anwar, Nick Barnes
Summary: The research proposes a compact and accurate super-resolution algorithm DRLN, which achieves deep supervision learning through cascading residual structures and densely concatenated residual block settings, and models inter-level and intra-level dependencies between crucial features using Laplacian attention. Comprehensive evaluations on various test datasets show that the DRLN algorithm performs significantly better in terms of visual quality and accuracy compared to other methods.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Chang-Wei Jing, Zhi-Xing Huang, Zai-Ying Ling
Summary: A multi-scale and multi-stage self similar fusion image super-resolution reconstruction algorithm is proposed in this paper to improve the accuracy and efficiency of super-resolution reconstruction. The algorithm utilizes feature extraction network to obtain low-frequency features and coding network to obtain structural feature information of low-resolution images. It also uses a multi-path feedforward network to extract high-frequency features and removes unnecessary components in the discriminator to enhance the reconstruction quality.
Article
Green & Sustainable Science & Technology
Amin Mahdavi-Meymand, Wojciech Sulisz
Summary: In this study, nested artificial neural networks were developed and applied to predict significant wave height at twenty selected locations of the North Sea, using wind speed and wind direction as input parameters. The results showed that the derived models were 18.39% more accurate than linear regression, and the nested artificial neural network could increase the accuracy of traditional models by up to 34%. Among all applied models, the nested artificial neural network developed based on the integration of particle swarm optimization algorithm and adaptive neuro-fuzzy inference system provided the most accurate prediction of wave heights, with RMSE = 0.525m and R2 = 0.84. The high accuracy of the results suggests that the application of nested artificial neural networks may be recommended for modeling wave parameters and other complex problems, if computational time is not critical for users.
Article
Geochemistry & Geophysics
Shi Chen, Lefei Zhang, Liangpei Zhang
Summary: The proposed MSDformer method utilizes CNN for local spatial-spectral information and Transformer for global spatial-spectral information, achieving excellent SR performance and outperforming state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Information Systems
Heewon Kim, Seokil Hong, Bohyung Han, Heesoo Myeong, Kyoung Mu Lee
Summary: This study proposes a novel neural architecture search (NAS) algorithm for designing efficient deep neural networks in image super-resolution (SR) tasks. By constructing a supernet, the algorithm allows flexibility in choosing the number of channels and per-channel activation functions, and improves efficiency and accuracy through channel pruning. The searched model, FGNAS, outperforms state-of-the-art NAS-based SR methods.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2022)
Article
Environmental Sciences
Luis Salgueiro, Javier Marcello, Veronica Vilaplana
Summary: This study introduces a single-image super-resolution model based on convolutional neural networks, which can enhance low-resolution bands captured by Sentinel-2 satellites to reach the maximal resolution. The proposed model, named Sen2-RDSR, outperforms other state-of-the-art approaches and is suitable for applications such as land cover classification and vegetation index generation.
Article
Chemistry, Analytical
Yooho Lee, Dongsan Jun, Byung-Gyu Kim, Hunjoo Lee
Summary: The paper introduces a lightweight CNN-based SR method called MCDN, which extracts training images from the DIV2K dataset and explores the trade-off between SR accuracy and network complexity. Experimental results demonstrate that the proposed method can significantly reduce network complexity while maintaining slightly better or similar perceptual quality compared to previous methods.
Article
Computer Science, Artificial Intelligence
Yudong Liang, Radu Timofte, Jinjun Wang, Sanping Zhou, Yihong Gong, Nanning Zheng
Summary: Deep learning has made significant advancements in single-image super-resolution, utilizing deep network design and external prior modeling. This study introduces a specific deep model to enhance low-resolution images and improve PSNR. Moreover, the approach is particularly suitable for images with repetitive structures or high resolutions.
PATTERN RECOGNITION
(2021)
Article
Automation & Control Systems
Long Sun, Zhenbing Liu, Xiyan Sun, Licheng Liu, Rushi Lan, Xiaonan Luo
Summary: In this paper, a fast and lightweight framework named weighted multi-scale residual network (WMRN) is proposed for a better tradeoff between image super-resolution performance and computational efficiency. The network utilizes depthwise separable convolutions and weighted multi-scale residual blocks to improve efficiency and multi-scale representation capability, with Convolutional layers in the reconstruction subnetwork to filter feature maps for high-quality image reconstruction. Extensive experiments show the effectiveness of WMRN compared to several state-of-the-art algorithms.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2021)
Article
Optics
Ying Wang, Feng Qi, Jinkuan Wang
Summary: By extending the convolutional neural network from real number domain to complex number domain based on the wave nature of terahertz light, researchers have shown the successful application of complex CNN in terahertz imaging for the first time. This approach can achieve higher resolution and contrast, better restoration of phase information, and outperforms traditional CNN in image reconstruction and generalization capability.
Article
Computer Science, Interdisciplinary Applications
Defu Qiu, Yuhu Cheng, Xuesong Wang
Summary: This study presents an improved generative adversarial network (IGAN) algorithm for retinal image super-resolution reconstruction. The proposed method improves the quality of reconstruction results and enhances texture details. Experimental results show significant improvements in objective evaluation indicators and a better visual experience compared to state-of-the-art image super-resolution methods.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Engineering, Civil
Weinan Huang, Sheng Dong
COASTAL ENGINEERING
(2020)
Article
Engineering, Marine
Zhenkun Liao, Weinan Huang, Sheng Dong, Huajun Li
Summary: This study focuses on the statistical characteristics of directional wave climate in the seasonal ice zone of the Barents Sea. By constructing joint distributions using a mixture trivariate distribution model, the study provides a reasonable description of the directional wave data. The study also calculates the extreme tensions associated with different return periods based on environmental contours.
Review
Meteorology & Atmospheric Sciences
Wei Zhang, Yu Sun, Yapeng Wu, Junyu Dong, Xiaojiang Song, Zhiyi Gao, Renbo Pang, Boyu Guoan
Summary: This study employed a spatiotemporal deep-learning method to correct biases in numerical ocean wave forecasts. By using a correction model driven by both wave and wind fields and a novel pixel-switch loss function, the corrected results performed well in different seasons and improved the accuracy of the original forecasts.