Abstract
In this paper, we propose a Multi-Scale Boosted Dehazing Network with Dense Feature Fusion based on the U-Net architecture. The proposed method is designed based on two principles, boosting and error feedback, and we show that they are suitable for the dehazing problem. By incorporating the Strengthen-Operate-Subtract boosting strategy in the decoder of the proposed model, we develop a simple yet effective boosted decoder to progressively restore the haze-free image. To address the issue of preserving spatial information in the U-Net architecture, we design a dense feature fusion module using the back-projection feedback scheme. We show that the dense feature fusion module can simultaneously remedy the missing spatial information from high-resolution features and exploit the non-adjacent features. Extensive evaluations demonstrate that the proposed model performs favorably against the state-of-the-art approaches on the benchmark datasets as well as real-world hazy images.
| Original language | English |
|---|---|
| Article number | 9156921 |
| Pages (from-to) | 2154-2164 |
| Number of pages | 11 |
| Journal | Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition |
| DOIs | |
| Publication status | Published - 2020 |
| Event | 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2020 - Virtual, Online, United States Duration: 2020 Jun 14 → 2020 Jun 19 |
Bibliographical note
Funding Information:As shown in Figure 7(d), by remedying the spatial information and exploiting the preceding features, the MSBDN-DFF successfully removes the remaining haze in Figure 7(c) and recovers more details. 5. Conclusions We propose an effective Multi-Scale Boosted Dehazing Network with Dense Feature Fusion for image dehazing. The MSBDN is constructed on an encoder-decoder architecture, where the boosted decoder is designed based on the SOS boosting strategy. The DFF module is designed on the back-projection scheme, which can preserve the spatial information and exploit the features from non-adjacent levels. The ablation studies demonstrate that the proposed modules are effective for the dehazing problem. Extensive evaluations show that the proposed model performs favorably against state-of-the-art methods on the image dehazing datasets. Acknowledgements. H. Dong, X. Lei, X. Zhang and F. Wang are supported in part by National Major Science and Technology Projects of China (No. 2019ZX01008103), National Natural Science Foundation of China (NSFC) (No. 61603291), Natural Science Basic Research Plan in Shaanxi Province of China (No. 2018JM6057), and the Fundamental Research Funds for the Central Universities. J. Pan is supported in part by NSFC (Nos. 61872421, 61922043) and NSF of Jiangsu Province (No. BK20180471). M.-H. Yang is supported in part by NSF CAREER Grant (No. 1149783) and Gifts from Verisk, Adobe and Google.
Publisher Copyright:
© 2020 IEEE
All Science Journal Classification (ASJC) codes
- Software
- Computer Vision and Pattern Recognition