Degradation-adaptive neural network for jointly single image dehazing and desnowing

doi:10.1007/s11704-023-2764-y

Front. Comput. Sci.

2024, Vol. 18

Issue (2) : 182707 https://doi.org/10.1007/s11704-023-2764-y

Image and Graphics

Degradation-adaptive neural network for jointly single image dehazing and desnowing

Erkang CHEN¹, Sixiang CHEN¹, Tian YE¹, Yun LIU²(

)

¹. School of Ocean Information Engineering, Jimei University, Xiamen 361021 China
². College of Artificial Intelligence, Southwest University, Chongqing 400715 China

Download: PDF(1829 KB) HTML
Export: BibTeX | EndNote | Reference Manager | ProCite | RefWorks

Corresponding Author(s): Yun LIU

Just Accepted Date: 13 October 2023 Issue Date: 08 January 2024

Cite this article:

Erkang CHEN,Sixiang CHEN,Tian YE, et al. Degradation-adaptive neural network for jointly single image dehazing and desnowing[J]. Front. Comput. Sci., 2024, 18(2): 182707.

URL:

https://academic.hep.com.cn/fcs/EN/10.1007/s11704-023-2764-y
https://academic.hep.com.cn/fcs/EN/Y2024/V18/I2/182707

Fig.1 The schematic overview of the proposed degradation-adaptive neural network

Fig.2 The detailed architecture of the proposed expert network

Fig.3 Details of our proposed modules. (a) Multi-branch spectral transform block; (b) dual-pool attention module; (c) cross-layer activation gated module; (d) adaptive gated neural

Method	Haze4k [3]		SOTS Outdoor [10]		#Param	#GMacs
Method	PSNR $↑$	SSIM $↑$	PSNR $↑$	SSIM $↑$	#Param	#GMacs
(AAAI'20) FFA-Net [1]	26.97	0.95	33.57	0.98	4.6 M	150.94 G
(CVPR'21) PSD-FFA [2]	16.56	0.73	15.29	0.72	4.6 M	150.94 G
(ACMMM'21) DMT-Net [3]	28.53	0.96	?	?	54.9 M	80.71 G
Dehazing Expert-Net	29.12	0.97	33.59	0.99	1.14 M	12.00 G
Dehazing Expert-Net-Tiny	28.18	0.96	32.09	0.98	288 K	3.06 G
DAN-Net	29.24	0.97	33.69	0.99	2.73 M	31.36 G
DAN-Net-Tiny	28.56	0.96	32.24	0.98	1.02 M	13.47 G

Tab.1 Quantitative comparisons of our framework with state-of-the-art dehazing methods on Haze4k and SOTS outdoor dataset

Method	CSD (2000) [7]		SRRS (2000) [6]		Snow 100K (2000) [5]		#Param	#GMacs
Method	PSNR $↑$	SSIM $↑$	PSNR $↑$	SSIM $↑$	PSNR $↑$	SSIM $↑$	#Param	#GMacs
(TIP'18) Desnow-Net [5]	20.13	0.81	20.38	0.84	30.50	0.94	15.6 M	?
(CVPR'20) All-in-One [8]	26.31	0.87	24.98	0.88	26.07	0.88	44 M	12.26 G
(ECCV'20) JSTASR [6]	27.96	0.88	25.82	0.89	23.12	0.86	65 M	?
(ICCV'21) HDCW-Net [7]	29.06	0.91	27.78	0.92	31.54	0.95	6.99 M	9.78 G
Desnowing Expert-Net	30.56	0.95	29.07	0.95	32.14	0.96	1.1 M	12.00 G
Desnowing Expert-Net-Tiny	29.06	0.92	28.20	0.94	31.67	0.95	288 K	3.06 G
DAN-Net	30.82	0.95	29.34	0.95	32.48	0.96	2.73 M	31.36 G
DAN-Net-Tiny	29.12	0.92	28.32	0.94	31.93	0.95	1.02 M	13.47 G

Tab.2 Quantitative comparisons of our framework with state-of-the-art desnowing methods on CSD, SRRS, and Snow 100K datasets

1	X, Qin Z, Wang Y, Bai X, Xie H Jia . FFA-Net: feature fusion attention network for single image dehazing. In: Proceedings of the 34th AAAI Conference on Artificial Intelligence. 2020, 11908–11915
2	Z, Chen Y, Wang Y, Yang D Liu . PSD: principled synthetic-to-real dehazing guided by physical priors. In: Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021, 7180–7189
3	Y, Liu L, Zhu S, Pei H, Fu J, Qin Q, Zhang L, Wan W Feng . From synthetic to real: image dehazing collaborating with unlabeled real data. In: Proceedings of the 29th ACM International Conference on Multimedia. 2021, 50–58
4	Y, Liu Z, Yan J, Tan Y Li . Multi-purpose oriented single nighttime image haze removal based on unified variational retinex model. IEEE Transactions on Circuits and Systems for Video Technology, 2023, 33( 4): 1643–1657
5	Y F, Liu D W, Jaw S C, Huang J N Hwang . DesnowNet: context-aware deep network for snow removal. IEEE Transactions on Image Processing, 2018, 27( 6): 3064–3073
6	W T, Chen H Y, Fang J J, Ding C C, Tsai S Y Kuo . JSTASR: joint size and transparency-aware snow removal algorithm based on modified partial convolution and veiling effect removal. In: Proceedings of the 16th European Conference on Computer Vision. 2020, 754–770
7	W T, Chen H Y, Fang C L, Hsieh C C, Tsai I H, Chen J J, Ding S Y Kuo . ALL snow removed: single image desnowing algorithm using hierarchical dual-tree complex wavelet representation and contradict channel loss. In: Proceedings of 2021 IEEE/CVF International Conference on Computer Vision. 2021, 4196–4205
8	R, Li R T, Tan L F Cheong . All in one bad weather removal using architectural search. In: Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020, 3175–3185
9	S, Pavlitskaya C, Hubschneider M, Weber R, Moritz F, Hüger P, Schlicht J M Zöllner . Using mixture of expert models to gain insights into semantic segmentation. In: Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. 2020, 342–343
10	B, Li W, Ren D, Fu D, Tao D, Feng W, Zeng Z Wang . Benchmarking single-image dehazing and beyond. IEEE Transactions on Image Processing, 2019, 28( 1): 492–505

[1]	FCS-22764-OF-EC_suppl_1	Download
[2]	FCS-22764-OF-EC_suppl_2	Download

Viewed

Full text

Abstract

Cited

Shared

Discussed