Underwater Image Restoration Based on Scene Depth Estimation and Background Segmentation

Jingyi Li; Guojia Hou; Xiaojia Zhang; Ting Lu; Yongfang Wang

doi:10.3788/LOP212986

Laser & Optoelectronics Progress, Volume. 60, Issue 2, 0210010(2023)

Underwater Image Restoration Based on Scene Depth Estimation and Background Segmentation

Jingyi Li¹, Guojia Hou^1、*, Xiaojia Zhang¹, Ting Lu¹, and Yongfang Wang²

Author Affiliations

¹College of Computer Science & Technology, Qingdao University, Qingdao 266071, Shandong, China

²School of Computer Science & Engineering, Linyi University, Linyi 276000, Shandong, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Flowchart of proposed method

Download full size

Fig. 2. Depth map estimation. (a) Original image; (b) gradient map; (c) estimated depth map based on gradient; (d) estimated depth map based on color difference

Download full size

Fig. 3. Results of background region segmentation. (a) Original image; (b1) gradient map; (b2) enhanced gradient map; (b3) edge information; (b4) initial detection only using gradient information; (c1) maximum value of G-B channel; (c2) R channel; (c3) difference between two channels; (c4) initial detection only using color difference; (d) candidate region of background light

Download full size

Fig. 4. Comparison of different background light estimation strategies. (a) Original underwater images; (b)-(d) estimated background light and corresponding restored results using method of reference [7], method of reference [9] , and proposed method

Download full size

Fig. 5. Comparison of restored results using different methods. (a) Original images; (b)-(f) restored results by RCP, IBLA, ULAP, UWCNN, and proposed method

Download full size

Fig. 6. Failure cases of proposed method. (a) Scenes with non-uniform illumination; (b) scene depth and background light estimation; (c) restored results

Download full size

Table 1. Comparison of different quality evaluation metrics

View table

Table 1. Comparison of different quality evaluation metrics

Metric	Method	Image 1	Image 2	Image 3	Image 4	Image 5	Image 6
UIQM	RCP	1.3302	1.4185	1.1620	0.9088	1.5060	1.5306
	IBLA	1.3382	1.3450	1.1763	0.9417	1.6645	1.4470
	ULAP	1.4922	1.3578	1.2203	1.0191	1.6680	1.6725
	UWCNN	1.0780	1.1509	0.9204	0.7838	1.3863	1.4114
	Proposed method	1.5166	1.4809	1.4404	1.1794	1.7633	1.5853
UCIQE	RCP	0.5019	0.5806	0.6068	0.5001	0.5225	0.4888
	IBLA	0.5197	0.5787	0.5884	0.4916	0.5861	0.4843
	ULAP	0.5770	0.5953	0.5917	0.4922	0.6163	0.4932
	UWCNN	0.4357	0.4631	0.5034	0.4595	0.4907	0.5036
	Proposed method	0.6643	0.6753	0.7182	0.6124	0.6393	0.6604
FDUM	RCP	0.5021	0.5345	0.5055	0.3165	0.5552	0.3278
	IBLA	0.5895	0.5811	0.5083	0.3541	0.8788	0.3372
	ULAP	0.7749	0.6142	0.5310	0.3953	0.9367	0.4126
	UWCNN	0.2967	0.2862	0.2913	0.2466	0.5453	0.3209
	Proposed method	0.9132	0.7820	0.7346	0.5223	1.1151	0.5506
FADE	RCP	0.6511	0.3969	0.4806	1.1197	0.3915	0.3208
	IBLA	0.7155	0.4913	1.3956	0.9377	0.3255	0.3194
	ULAP	0.4921	0.4629	0.4245	0.7758	0.3664	0.2438
	UWCNN	1.2032	0.8691	0.5235	0.9187	0.7373	0.2824
	Proposed method	0.3741	0.3061	0.2714	0.5971	0.2870	0.2006

Table 2. Comparison of average values of different quality evaluation metrics
View table
Table 2. Comparison of average values of different quality evaluation metrics
Method RCP IBLA ULAP UWCNN Proposed method
UIQM 1.3416 1.2180 1.2238 1.0473 1.3952
UCIQE 0.5660 0.5577 0.5520 0.4690 0.6272
FDUM 0.5713 0.5148 0.5206 0.3052 0.7023
FADE 0.4508 0.5871 0.5596 0.7486 0.4092

Tools

Get Citation

Copy Citation Text

Jingyi Li, Guojia Hou, Xiaojia Zhang, Ting Lu, Yongfang Wang. Underwater Image Restoration Based on Scene Depth Estimation and Background Segmentation[J]. Laser & Optoelectronics Progress, 2023, 60(2): 0210010

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites