Underwater Image Enhancement Model Based on Deep Multi-Prior Learning

Fig. 2. Visual quality of UIEB_10223 image. (a) Original image; (b) $z_{1}^{c} [m, n]$ ; (c) $z_{2}^{c} [m, n]$ ; (d) $z_{3}^{c} [m, n]$ ; (e) $z_{4}^{c} [m, n]$

Download full size

Fig. 3. Enhancement results of the test_p114 image in the EUVP dataset. (a) Original image; (b) RoWS^[3]; (c) MIP^[2]; (d) ULAP^[11]; (e) FUnIE-GAN^[7]; (f) UIESS^[8]; (g) UWNet^[21]; (h) PUIENet-MC^[10]; (i) PUIENet-MP^[10]; (j) TOPAL^[9]; (k) proposed model; (l) reference truth map

Download full size

Fig. 4. Enhancement results of the test_p317 image in the EUVP dataset. (a) Original image; (b) RoWS^[3]; (c) MIP^[2]; (d) ULAP^[11]; (e) FUnIE-GAN^[7]; (f) UIESS^[8]; (g) UWNet^[21]; (h) PUIENet-MC^[10]; (i) PUIENet-MP^[10]; (j) TOPAL^[9]; (k) proposed model; (l) reference truth map

Download full size

Table 1. Impact of different models on network performance
View table
Table 1. Impact of different models on network performance
Model EUVP
PSNR /dB SSIM UIQM
V1 26.02 0.78 2.83
V2 26.21 0.79 2.85
V3 26.98 0.79 2.93
DMPL 27.21 0.81 3.01

Table 2. Performance comparison on EUVP dataset

View table

Table 2. Performance comparison on EUVP dataset

Model	PSNR	SSIM	UIQM	BRISQUE
RoWS^［2］	24.86	0.72	2.14	36.81
MIP^［3］	19.51	0.63	1.93	35.31
ULAP^［12］	21.89	0.72	2.20	35.39
FUnIE-GAN^［8］	26.22	0.79	2.97	30.69
UIESS^［9］	24.30	0.79	3.01	37.56
UWNet^［20］	27.10	0.79	2.96	40.34
PUIENet-MC^［11］	21.35	0.78	3.17	36.75
PUIENet-MP^［11］	21.45	0.77	3.19	36.92
TOPAL^［10］	22.14	0.80	2.98	35.10
DMPL	27.21	0.81	3.01	32.03

Table 3. Performance comparison on UIEB dataset

View table

Table 3. Performance comparison on UIEB dataset

Model	NIQE	PIQE	UCIQE	CCF
RoWS^［2］	5.09	44.58	0.52	35.52
MIP^［3］	5.54	46.18	0.54	38.54
ULAP^［12］	5.21	45.16	0.56	42.10
FUnIE-GAN^［8］	5.09	15.99	0.54	18.99
UIESS^［9］	5.71	42.63	0.57	19.26
UWNet^［20］	5.80	50.67	0.50	16.85
PUIENet-MC^［11］	4.73	42.15	0.56	21.60
PUIENet-MP^［11］	4.76	43.28	0.54	19.64
TOPAL^［10］	4.99	36.97	0.55	23.11
DMPL	4.70	15.47	0.56	18.25

Table 4. Complexity of different underwater image enhancement algorithms
View table
Table 4. Complexity of different underwater image enhancement algorithms
Model Inference time /s
FUnIE-GAN^［8］ 0.024
UIESS^［9］ 0.046
PUIENet-MC^［11］ 0.170
PUIENet-MP^［11］ 0.030
DMPL 0.201

Tools

Get Citation

Copy Citation Text

Yang Ou, Jianfeng Huang, Rong Yuan. Underwater Image Enhancement Model Based on Deep Multi-Prior Learning[J]. Laser & Optoelectronics Progress, 2024, 61(22): 2237005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites