Method for Removal of Rain and Fog in Single Image

Fig. 3. Dehazing map and its color histogram. (a) Foggy image; (b) manifold particle filtering for fog removal; (c) color histogram of Fig. 3(a); (d) color histogram of Fig. 3(b)

Download full size

Fig. 4. Architecture of proposed method in this paper

Download full size

Fig. 5. Original images and test results of each algorithm for synthetic uniform fog and fog in natural scene. (a) Input hazy images; (b) soft matting; (c) GF algorithm; (d) method proposed by He et al.; (e) Retinex algorithm; (f) HF; (g) NBPC+PA; (h) proposed method

Download full size

Fig. 6. Original images and test results of each algorithm for rain and fog removal. (a) Rain image; (b) Gaussian curvature algorithm; (c) DRN; (d) proposed method

Download full size

Table 1. Structural parameter settings of recursive network

View table

Table 1. Structural parameter settings of recursive network

Layer		Feature
	0	Detail layer
	1	Conv(1×1, 32); stride:1; ReLU
	2	Conv(1×1, 32); stride:1; ReLU
	3	Conv(1×1, 32); stride: 1; ReLU
Residual	4	Concatenate (2,3)
	5	Conv(1×1, 32); stride:1; ReLU
	6	Conv (1×1, 32); stride:1; ReLU
	7	Concatenate (4,5)
	Input gate	Conv (3×3, 32); stride:1; sigmoid
LSTM	Forget gate	Conv (3×3, 32); stride:1; sigmoid
	Cell state	Conv (3×3, 32); stride:1; sigmoid
	Output gate	Conv (3×3, 32); stride:1; sigmoid

Table 2. Discriminator structural parameter setting

View table

Table 2. Discriminator structural parameter setting

Layer	Feature
Layer 0	Output image
Layer 1	Conv (5×5, 8); stride:1; ReLU
Layer 2	Conv (5×5, 16); stride:1; ReLU
Layer 3	Conv (5×5, 32); stride:1; ReLU
Layer 4	Conv (5×5, 64); stride:1; ReLU
Layer 5	Conv (5×5, 128); stride:1; ReLU
Layer 6	Conv (5×5, 128); stride:1; ReLU
Layer 7	Conv (5×5, 64); stride:4; ReLU
Layer 8	Conv (5×5, 64); stride:4; ReLU
Layer 9	Conv (5×5, 32); stride:4; ReLU

Table 3. Data obtained by objective evaluation methods for testing images in Fig. 5

View table

Table 3. Data obtained by objective evaluation methods for testing images in Fig. 5

SettingNo.		Softmatting	GF algorithm	Methodproposedby He et al.	Retinexalgorithm	HF	NBPC+PA	MPF (manifoldparticlefiltering)
	MSD /10⁵	0.832	0.894	1.310	0.995	0.857	1.212	2.011
1	PSNR	62.454	61.272	60.832	60.582	62.658	61.536	62.720
	SSIM	0.418	0.403	0.416	0.251	0.690	0.534	0.746
	MSD /10⁵	0.954	0.980	1.538	1.058	0.976	0.779	1.310
2	PSNR	59.081	60.718	59.930	58.373	58.545	60.681	61.346
	SSIM	0.450	0.484	0.389	0.373	0.473	0.515	0.815
	MSD /10⁵	2.590	3.416	4.235	3.220	2.319	1.960	5.099
3	PSNR	63.074	58.258	61.474	54.307	60.505	63.591	65.186
	SSIM	0.812	8.516	0.539	0.325	0.756	0.818	0.871
	MSD /10⁵	2.122	2.747	3.462	1.656	2.703	1.886	4.990
4	PSNR	61.747	56.114	60.249	55.532	61.319	63.498	61.275
	SSIM	0.589	0.474	0.483	0.351	0.595	0.684	0.743
	MSD /10⁵	2.494	3.283	3.700	3.076	1.973	1.783	4.495
5	PSNR	57.721	56.292	57.987	54.654	61.396	60.756	60.127
	SSIM	0.530	0.460	0.538	0.333	0.699	0.809	0.828
	MSD /10⁵	3.634	5.388	5.745	4.688	4.052	3.478	7.636
6	PSNR	65.729	57.372	58.644	54.456	60.794	65.207	67.130
	SSIM	0.726	0.588	0.616	0.334	0.701	0.820	0.905

Table 4. Data obtained by objective evaluation methods for natural rain and fog scene for experiments (Fig. 6)

View table

Table 4. Data obtained by objective evaluation methods for natural rain and fog scene for experiments (Fig. 6)

Scene No.	FSIM			SSIM			PSNR
Scene No.	GC	DRN	IA-GAN	GC	DRN	IA-GAN	GC	DRN	IA-GAN
1	0.741	0.778	0.967	0.740	0.906	0.937	64.038	37.482	71.495
2	0.812	0.794	0.842	0.903	0.896	0.956	67.368	70.048	70.544
3	0.703	0.833	0.895	0.900	0.912	0.947	66.611	70.681	74.159
4	0.804	0.864	0.847	0.907	0.887	0.946	70.811	69.243	69.879
5	0.718	0.873	0.893	0.354	0.915	0.939	68.337	71.086	72.815
注:GC: Gaussian curvature; IA-GAN: improved attentive generative adversarial network

Tools

Get Citation

Copy Citation Text

Bingyuan Wang, Fang Zheng, Jian Jiang, Bo Yang. Method for Removal of Rain and Fog in Single Image[J]. Laser & Optoelectronics Progress, 2020, 57(14): 141027

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Oct. 21, 2019

Accepted: Dec. 31, 2019

Published Online: Jul. 28, 2020

The Author Email: Fang Zheng (1067229919@qq.com)

DOI:10.3788/LOP57.141027

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology