Hyperspectral image super-resolution using cluster-based deep convolutional networks

Zou, Changzhong; Zhang, Can

doi:10.1016/j.image.2022.116884

Cited by 7 publications

(3 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep learning algorithms build models for network analysis and learning to simulate the mechanism of the human brain to interpret data information. Super-resolution technology based on deep learning can improve image resolution under existing hardware conditions [45].…”

Section: Machine Learning Methodology 261 Deep Super-resolution Convo...mentioning

confidence: 99%

Predicting the Flow Fields in Meandering Rivers with a Deep Super-Resolution Convolutional Neural Network

Yan,

Du,

Zhang

et al. 2024

Water

View full text Add to dashboard Cite

The hydrodynamic model, based on the strict conservation of momentum and continuity equations, can accurately simulate the distribution of a flow field. However, significant computing time and storage space requirements limit real-time prediction. Machine learning is well known for its fast computing speed and powerful learning ability, but its accuracy depends on an abundance of training data, hindering its wider use in locations without sufficient measurements. Application restrictions in data-deficient areas can be addressed through transfer learning, provided that two areas share common characteristics. In this study, a machine learning method based on a deep super-resolution convolutional neural network (DSRCNN) and transfer learning is proposed, validated, and applied to model two bend flows and one realistic test case. Firstly, the hydrodynamic model was established and validated against measured data. The validated model was considered to have the ability to generate real data and was used to generate a comprehensive data set for training and validating the machine learning model. Three different methods were compared and tested, with Realizable k-ε performing better than the others in predicting the outer bank flow distribution. DSRCNN was compared to a plain SRCNN (PSRCNN), as well as Bilinear, Nearest, and Bicubic methods, and the results showed that DSRCNN had the best performance. We compared Raw, RT, and TL methods, finding that the TL method performed the best overall. Therefore, the research results showed that the developed super-resolution convolutional neural network can provide more reliable predications and serve as an ideal tool for simulating flow field distribution in bends.

show abstract

Section: Machine Learning Methodology 261 Deep Super-resolution Convo...mentioning

confidence: 99%

Predicting the Flow Fields in Meandering Rivers with a Deep Super-Resolution Convolutional Neural Network

Yan,

Du,

Zhang

et al. 2024

Water

View full text Add to dashboard Cite

show abstract

“…The efficacy of CAS-CNN is reflected in various assessments, which include a few experiments on benchmark datasets as well as on real-life images. Figures show that to reach the best performance for artifact reduction, the CAS-CNN surpasses all the state-of-the-art techniques [8].…”

Section: Introductionmentioning

confidence: 99%

Enhance Image Quality with CAS-CNN: A Deep Convolutional Neural Network for Compression Artifact Suppression

Radha Seelaboyina

2024

jes

View full text Add to dashboard Cite

Objective: The goal is to outdo clumsy filtering algorithms and prevailing deep learning approaches in terms of the image quality metrics, computational efficiency, and memory usage by attaining better performance. Methodology: Two-step approach has the methodology component including the formulation of the problem and review of the literature collections to detect known techniques. CAS-CNN's architecture is designed, and the entire experiment setup requires the preparation of an original and a compressed images dataset and also defining loss function for training and improving CAS-CNN parameters as a whole. Outputs of metrics are the PSNR, SSIM, MOS values, inference time, model size, training time, and hardware usage. Results and Discussion: In terms of competitiveness, CAS-CNN is always ahead of all the latest algorithm levels and achieves the best results by standard metrics such as PSNR and SSIM, thus it is the method that gives the most undistorted recreated images with a high quality. It has shown its ability to compete closely with traditional methods by providing better performance, as well as consuming fewer computing resources, which makes it closer to the practical reality of real time applications. The CAS-CNN's model creates efficiency that enables cutting-edge design and testing into the market timelier than the existing methods do. Small progress is seen despite its quick performance; but still, important issues like artificial data and restrictions on given conditions are noticed, which means that we must not stop searching for scaling and generalizing power. Conclusion: CAS-CNN is a next-level algorithm concerning removing distortion in compression, superior to prevailing techniques in terms of both effectiveness, efficiency, and accessibility. It holds an edge over other methods due to its high performance and efficiency, and as such, it is an applicable tool for a variety of situations where one must deal with limited resources in the need for high-quality images. More research is needed to account for the existing shortcomings and apply CAS-CNN to a range of different datasets, as well as the circumstances in the real world.

show abstract

“…This provides significant potential in Earth observation missions, such as land cover mapping [1], precision agriculture [2], land cover classification [3], environmental monitoring [4], and mineral exploration [5]. Several hyperspectral image data processing techniques have been explored, such as denoising [6,7], unmixing [8,9], superresolution [10][11][12][13], target detection [14,15], change detection [16], and classification [17][18][19][20][21]. Among these techniques, HSI classification has attracted more attention.…”

Section: Introductionmentioning

confidence: 99%

S2Former: Parallel Spectral–Spatial Transformer for Hyperspectral Image Classification

Yuan,

Yu,

Qian

et al. 2023

Electronics

View full text Add to dashboard Cite

Due to their excellent representation talent in local features, the convolutional neural network (CNN) has achieved favourable performance in hyperspectral image (HSI) classification tasks. Nevertheless, current CNN models exhibit a marked flaw: they are hard to model the dependencies in long-range distanced positions. This flaw becomes more problematic for the HSI classification task, which targets extracting more discriminative features in local and global dimensions from limited samples. In this paper, we introduce a spatial–spectral transformer (S2Former), which explores spatial and spectral feature extraction in a dual-stream framework for HSI Classification. S2Former, which consists of a spatial transformer and a spectral transformer in parallel branches, extracts the discriminative feature in spatial and spectral dimensions. More specifically, we propose multi-head spatial self-attention to capture the long-range spatial dependency of non-adjacent HSI pixels in a spatial transformer. In the spectral transformer, we propose multi-head covariance spectral attention to mine and represent spectral signatures by computing covariance-based channel maps. Meanwhile, the local activation feed-forward network is developed to complement local details. Extensive experiments conducted on four publicly available datasets indicate that our S2Former achieves state-of-the-art performance for the HSI classification task.

show abstract

Hyperspectral image super-resolution using cluster-based deep convolutional networks

Cited by 7 publications

References 28 publications

Predicting the Flow Fields in Meandering Rivers with a Deep Super-Resolution Convolutional Neural Network

Predicting the Flow Fields in Meandering Rivers with a Deep Super-Resolution Convolutional Neural Network

Enhance Image Quality with CAS-CNN: A Deep Convolutional Neural Network for Compression Artifact Suppression

S2Former: Parallel Spectral–Spatial Transformer for Hyperspectral Image Classification

Contact Info

Product

Resources

About