Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images
Deep convolutional neural networks (CNNs) have been successfully applied to spectral reconstruction (SR) and acquired superior performance. Nevertheless, the existing CNN-based SR approaches integrate hierarchical features from different layers indiscriminately, lacking an investigation of the relat...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-12-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/13/1/115 |
| _version_ | 1797542788133289984 |
|---|---|
| author | Jiaojiao Li Chaoxiong Wu Rui Song Yunsong Li Weiying Xie |
| author_facet | Jiaojiao Li Chaoxiong Wu Rui Song Yunsong Li Weiying Xie |
| author_sort | Jiaojiao Li |
| collection | DOAJ |
| description | Deep convolutional neural networks (CNNs) have been successfully applied to spectral reconstruction (SR) and acquired superior performance. Nevertheless, the existing CNN-based SR approaches integrate hierarchical features from different layers indiscriminately, lacking an investigation of the relationships of intermediate feature maps, which limits the learning power of CNNs. To tackle this problem, we propose a deep residual augmented attentional u-shape network (RA<sup>2</sup>UN) with several double improved residual blocks (DIRB) instead of paired plain convolutional units. Specifically, a trainable spatial augmented attention (SAA) module is developed to bridge the encoder and decoder to emphasize the features in the informative regions. Furthermore, we present a novel channel augmented attention (CAA) module embedded in the DIRB to rescale adaptively and enhance residual learning by using first-order and second-order statistics for stronger feature representations. Finally, a boundary-aware constraint is employed to focus on the salient edge information and recover more accurate high-frequency details. Experimental results on four benchmark datasets demonstrate that the proposed RA<sup>2</sup>UN network outperforms the state-of-the-art SR methods under quantitative measurements and perceptual comparison. |
| first_indexed | 2024-03-10T13:35:29Z |
| format | Article |
| id | doaj.art-13b39436b2b345bfa49a6720977f06d7 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T13:35:29Z |
| publishDate | 2020-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-13b39436b2b345bfa49a6720977f06d72023-11-21T07:31:19ZengMDPI AGRemote Sensing2072-42922020-12-0113111510.3390/rs13010115Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB ImagesJiaojiao Li0Chaoxiong Wu1Rui Song2Yunsong Li3Weiying Xie4The State Key Laboratory of Integrated Service Networks, Xidian University, Xi’an 710000, ChinaThe State Key Laboratory of Integrated Service Networks, Xidian University, Xi’an 710000, ChinaThe State Key Laboratory of Integrated Service Networks, Xidian University, Xi’an 710000, ChinaThe State Key Laboratory of Integrated Service Networks, Xidian University, Xi’an 710000, ChinaThe State Key Laboratory of Integrated Service Networks, Xidian University, Xi’an 710000, ChinaDeep convolutional neural networks (CNNs) have been successfully applied to spectral reconstruction (SR) and acquired superior performance. Nevertheless, the existing CNN-based SR approaches integrate hierarchical features from different layers indiscriminately, lacking an investigation of the relationships of intermediate feature maps, which limits the learning power of CNNs. To tackle this problem, we propose a deep residual augmented attentional u-shape network (RA<sup>2</sup>UN) with several double improved residual blocks (DIRB) instead of paired plain convolutional units. Specifically, a trainable spatial augmented attention (SAA) module is developed to bridge the encoder and decoder to emphasize the features in the informative regions. Furthermore, we present a novel channel augmented attention (CAA) module embedded in the DIRB to rescale adaptively and enhance residual learning by using first-order and second-order statistics for stronger feature representations. Finally, a boundary-aware constraint is employed to focus on the salient edge information and recover more accurate high-frequency details. Experimental results on four benchmark datasets demonstrate that the proposed RA<sup>2</sup>UN network outperforms the state-of-the-art SR methods under quantitative measurements and perceptual comparison.https://www.mdpi.com/2072-4292/13/1/115spectral reconstructionresidual augmented attentional u-shape networkspatial augmented attentionchannel augmented attentionboundary-aware constraint |
| spellingShingle | Jiaojiao Li Chaoxiong Wu Rui Song Yunsong Li Weiying Xie Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images Remote Sensing spectral reconstruction residual augmented attentional u-shape network spatial augmented attention channel augmented attention boundary-aware constraint |
| title | Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images |
| title_full | Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images |
| title_fullStr | Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images |
| title_full_unstemmed | Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images |
| title_short | Residual Augmented Attentional U-Shaped Network for Spectral Reconstruction from RGB Images |
| title_sort | residual augmented attentional u shaped network for spectral reconstruction from rgb images |
| topic | spectral reconstruction residual augmented attentional u-shape network spatial augmented attention channel augmented attention boundary-aware constraint |
| url | https://www.mdpi.com/2072-4292/13/1/115 |
| work_keys_str_mv | AT jiaojiaoli residualaugmentedattentionalushapednetworkforspectralreconstructionfromrgbimages AT chaoxiongwu residualaugmentedattentionalushapednetworkforspectralreconstructionfromrgbimages AT ruisong residualaugmentedattentionalushapednetworkforspectralreconstructionfromrgbimages AT yunsongli residualaugmentedattentionalushapednetworkforspectralreconstructionfromrgbimages AT weiyingxie residualaugmentedattentionalushapednetworkforspectralreconstructionfromrgbimages |