Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes
In this paper, a transmission-guided lightweight neural network called TGL-Net is proposed for efficient image dehazing. Unlike most current dehazing methods that produce simulated transmission maps from depth data and haze-free images, in the proposed work, guided transmission maps are computed aut...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-02-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/21/3/960 |
| _version_ | 1797417145660866560 |
|---|---|
| author | Zhan Li Jianhang Zhang Ruibin Zhong Bir Bhanu Yuling Chen Qingfeng Zhang Haoqing Tang |
| author_facet | Zhan Li Jianhang Zhang Ruibin Zhong Bir Bhanu Yuling Chen Qingfeng Zhang Haoqing Tang |
| author_sort | Zhan Li |
| collection | DOAJ |
| description | In this paper, a transmission-guided lightweight neural network called TGL-Net is proposed for efficient image dehazing. Unlike most current dehazing methods that produce simulated transmission maps from depth data and haze-free images, in the proposed work, guided transmission maps are computed automatically using a filter-refined dark-channel-prior (F-DCP) method from real-world hazy images as a regularizer, which facilitates network training not only on synthetic data, but also on natural images. A double-error loss function that combines the errors of a transmission map with the errors of a dehazed image is used to guide network training. The method provides a feasible solution for introducing priors obtained from traditional non-learning-based image processing techniques as a guide for training deep neural networks. Extensive experimental results demonstrate that, in terms of several reference and non-reference evaluation criteria for real-world images, the proposed method can achieve state-of-the-art performance with a much smaller network size and with significant improvements in efficiency resulting from the training guidance. |
| first_indexed | 2024-03-09T06:14:36Z |
| format | Article |
| id | doaj.art-4f605f4415334897a40e8923be990372 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T06:14:36Z |
| publishDate | 2021-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-4f605f4415334897a40e8923be9903722023-12-03T11:54:31ZengMDPI AGSensors1424-82202021-02-0121396010.3390/s21030960Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy ScenesZhan Li0Jianhang Zhang1Ruibin Zhong2Bir Bhanu3Yuling Chen4Qingfeng Zhang5Haoqing Tang6Department of Computer Science, Jinan University, Guangzhou 510632, ChinaDepartment of Computer Science, Jinan University, Guangzhou 510632, ChinaDepartment of Computer Science, Jinan University, Guangzhou 510632, ChinaDepartment of Electrical and Computer Engineering, University of California, Riverside, CA 92521, USAChina Southern Airlines Co. Ltd., Guangzhou 510080, ChinaDepartment of Computer Science, Jinan University, Guangzhou 510632, ChinaDepartment of Computer Science, Jinan University, Guangzhou 510632, ChinaIn this paper, a transmission-guided lightweight neural network called TGL-Net is proposed for efficient image dehazing. Unlike most current dehazing methods that produce simulated transmission maps from depth data and haze-free images, in the proposed work, guided transmission maps are computed automatically using a filter-refined dark-channel-prior (F-DCP) method from real-world hazy images as a regularizer, which facilitates network training not only on synthetic data, but also on natural images. A double-error loss function that combines the errors of a transmission map with the errors of a dehazed image is used to guide network training. The method provides a feasible solution for introducing priors obtained from traditional non-learning-based image processing techniques as a guide for training deep neural networks. Extensive experimental results demonstrate that, in terms of several reference and non-reference evaluation criteria for real-world images, the proposed method can achieve state-of-the-art performance with a much smaller network size and with significant improvements in efficiency resulting from the training guidance.https://www.mdpi.com/1424-8220/21/3/960single image dehazingtransmission-guidedlightweight neural networkimage restoration |
| spellingShingle | Zhan Li Jianhang Zhang Ruibin Zhong Bir Bhanu Yuling Chen Qingfeng Zhang Haoqing Tang Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes Sensors single image dehazing transmission-guided lightweight neural network image restoration |
| title | Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes |
| title_full | Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes |
| title_fullStr | Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes |
| title_full_unstemmed | Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes |
| title_short | Lightweight and Efficient Image Dehazing Network Guided by Transmission Estimation from Real-World Hazy Scenes |
| title_sort | lightweight and efficient image dehazing network guided by transmission estimation from real world hazy scenes |
| topic | single image dehazing transmission-guided lightweight neural network image restoration |
| url | https://www.mdpi.com/1424-8220/21/3/960 |
| work_keys_str_mv | AT zhanli lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes AT jianhangzhang lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes AT ruibinzhong lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes AT birbhanu lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes AT yulingchen lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes AT qingfengzhang lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes AT haoqingtang lightweightandefficientimagedehazingnetworkguidedbytransmissionestimationfromrealworldhazyscenes |