Underwater Image Enhancement Based on Multiscale Generative Adversarial Network

Fig. 9. Experimental of the color restoration. (a) Original image; (b) Lab; (c) UDCP; (d) CLAHE; (e) DehazeNet and HWD; (f) Uresnet; (g) MSGAN; (h) standard color card

Download full size

Fig. 10. Reconstructed images of different algorithms. (a) Original image; (b) UDCP; (c) Lab; (d) DehazeNet and HWD; (e) DUIENet; (f) FUnIE-GAN; (g) Uresnet; (h) MSGAN

Download full size

Fig. 11. Feature matching results of different algorithms. (a) Original image; (b) DUIENet; (c) FUnIE-GAN; (d) Uresnet; (e) MSGAN

Download full size

Table 1. Parameters of the model
View table
Table 1. Parameters of the model
Parameter Image size Learning Optimizer Batch size epoch
Value 256×256 0.001 Adam 2 100

Table 2. UCIQE of different algorithms

View table

Table 2. UCIQE of different algorithms

Image	Original image	UDCP	Lab	DehazeNet and HWD	DUIENet	FUnIE-GAN	Uresnet	MSGAN
1	0.3999	0.5658	0.4826	0.5534	0.5607	0.5883	0.5713	0.6230
2	0.4734	0.6036	0.5015	0.5938	0.5955	0.5729	0.6008	0.6561
3	0.4107	0.6242	0.4989	0.5971	0.6076	0.5906	0.5800	0.6487
4	0.4869	0.6607	0.5100	0.6120	0.5952	0.5739	0.6015	0.6523
5	0.4350	0.5788	0.5015	0.6216	0.6119	0.5854	0.5868	0.6474
6	0.5804	0.6478	0.5766	0.6256	0.6466	0.6073	0.6482	0.6537
7	0.4131	0.5759	0.4322	0.5884	0.5989	0.4816	0.5829	0.6271
8	0.4626	0.5951	0.4381	0.6199	0.5789	0.5083	0.5790	0.6485
9	0.3919	0.5309	0.3725	0.5711	0.5313	0.5067	0.5928	0.5953
Average	0.4504	0.5981	0.4793	0.5981	0.5918	0.5572	0.5937	0.6391

Table 3. UIQM of different algorithms

View table

Table 3. UIQM of different algorithms

Image	Original image	UDCP	Lab	DehazeNet and HWD	DUIENet	FUnIE-GAN	Uresnet	MSGAN
1	2.7412	3.9075	4.6411	5.4308	5.3978	5.6093	5.2556	5.5662
2	4.0407	4.2090	5.4166	5.4916	5.5263	5.9190	5.7752	5.7604
3	2.4437	3.2558	4.6238	5.3570	4.9090	5.055	4.3874	5.2865
4	4.2092	5.5160	5.7056	5.3128	5.5416	5.4230	5.6801	6.0972
5	2.6370	5.3656	4.9070	5.4664	5.1776	5.6752	5.1168	6.0392
6	3.3848	4.8226	5.5025	5.2776	5.2172	5.2147	4.8218	5.4851
7	-0.1967	1.5325	3.9102	4.6253	3.7125	2.8935	3.111	4.0278
8	2.7643	4.7148	4.2933	5.2013	5.0453	5.2807	4.228	5.5883
9	1.7962	3.5674	3.7008	5.3365	4.9946	5.4482	4.7822	5.2365
Average	2.6467	4.0990	4.7445	5.2777	5.0580	5.1687	4.7953	5.4541

Table 4. SURF of different algorithms
View table
Table 4. SURF of different algorithms
Image Original image DUIENe FUnIE-GAN Uresnet MsGAN
1 31 35 36 31 54
2 15 29 20 26 34
3 21 27 30 33 46
4 39 46 46 44 68
Average 26.5 34.25 33 33.5 50.5

Tools

Get Citation

Copy Citation Text

Sen Lin, Shiben Liu. Underwater Image Enhancement Based on Multiscale Generative Adversarial Network[J]. Laser & Optoelectronics Progress, 2021, 58(16): 1610017

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites