Dual-Stream Convolutional Neural Network With Residual Information Enhancement for Pansharpening

Yang, Yong; Tu, Wei; Huang, Shuying; Lu, Hangyuan; Wan, Weiguo; Liu, Gan

doi:10.1109/tgrs.2021.3098752

Cited by 26 publications

(8 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Next, the PAN image is convolved with the estimated filter, and the extracted detail components are those missing from the low-resolution MS image. Yang et al [ 37 ] proved that the filter estimation has strong robustness through experiments. Even if white noise is added to the initial fusion image, it does not affect the extraction of details.…”

Section: The Proposed Methodsmentioning

confidence: 99%

An Adaptive Injection Model for Pansharpening

Song

Chen

Ren

et al. 2023

Computational Intelligence and Neuroscience

Self Cite

View full text Add to dashboard Cite

Pansharpening technology is used to acquire a multispectral image with high spatial resolution from a panchromatic (PAN) image and a multispectral (MS) image. The detail injection model is popular for its flexibility. However, the accuracy of the injection gain and the extracted details may greatly influence the quality of the pansharpened image. This paper proposes an adaptive injection model to solve these problems. For detail extraction, we present a Gaussian filter estimation algorithm by exploring the intrinsic character of the MS sensor and convolving the PAN image with the filter to adaptively optimize the details to be consistent with the character of the MS image. For the adaptive injection coefficient, we iteratively adjust the coefficient by balancing the spectral and spatial fidelity. By multiplying the optimized details and injection gain, the final HRMS is obtained with the injection model. The performance of the proposed model is analyzed and a large number of tests are carried out on various satellite datasets. Compared to some advanced pansharpening methods, the results prove that our method can achieve the best fusion quality both subjectively and objectively.

show abstract

Section: The Proposed Methodsmentioning

confidence: 99%

An Adaptive Injection Model for Pansharpening

Song

Chen

Ren

et al. 2023

Computational Intelligence and Neuroscience

Self Cite

View full text Add to dashboard Cite

show abstract

“…This section introduces the convolutional attention mechanism [23] and the dual-stream convolutional neural network [24] used in this paper.…”

Section: Theoretical Basismentioning

confidence: 99%

Infrared image fault diagnosis based on dual-stream attention convolution network

Lu,

Yang,

Ming

et al. 2024

Eng. Res. Express

View full text Add to dashboard Cite

The development of infrared image detection technology has improved the real-time performance and safety of fault diagnosis. To address the problem of insufficient features caused by the low resolution of infrared images, a fault diagnosis model for induction motor infrared images based on Dual-Stream Attention Convolutional (DSAC) is proposed. Firstly, this model employs a dual-stream convolutional neural network to extract spatial features in the X and Y directions of the infrared images separately. This dual-stream structure allows the network to simultaneously learn and utilize information from both X and Y directions, enabling a more comprehensive capture of temperature distribution and variation trends in the infrared images. Then, a convolutional attention mechanism is introduced to assign weights to the obtained features. The attention maps generated by the convolutional attention layer enhance key features while suppressing unimportant information, thereby enhancing the focusing performance of the DSAC model on critical features. Finally, the weighted dual-stream features are fused to achieve the goal of fault diagnosis for induction motor infrared images. The DSAC model is validated using the induction motor infrared image dataset fromBabol Noshirvani University of Technology, demonstrating excellent diagnostic accuracy and speed. Moreover, it alleviates the problem of sample imbalance to some extent in small-sample datasets, providing a feasible solution for fault diagnosis of induction motor infrared images.

show abstract

“…However, simply stacking pre-interpolated MS with Pan as the input of network not only ignores individual features but also raises extra computational burden. Hence, instead of treating the two modalities equally, multi-branch networks apply different sub-networks to separately extract the modality-specific features (Shao and Cai, 2018;Zhang et al, 2019c;Liu et al, 2020b;Chen et al, 2021;Zhang and Ma, 2021;Xing et al, 2020;Yang et al, 2022b).…”

Section: Homogeneous Fusionmentioning

confidence: 99%