Performance Improvement of Laser Interference Image Restoration Based on Multi-Scale Feature Fusion

Haoqian Wang; Ju Liu; Teng Li; Zhongjie Xu; Xiang'ai Cheng; Zhongyang Xing

doi:10.3788/LOP241476

Laser & Optoelectronics Progress, Volume. 62, Issue 4, 0437004(2025)

Performance Improvement of Laser Interference Image Restoration Based on Multi-Scale Feature Fusion

Haoqian Wang^1,2,3、*, Ju Liu^1,2,3, Teng Li¹, Zhongjie Xu^1,2,3, Xiang'ai Cheng^1,2,3, and Zhongyang Xing^1,2,3

¹College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha 410073, Hunan , China

²State Key Laboratory of Pulsed Power Laser Technology, Changsha 410073, Hunan , China

³Hunan Provincial Key Laboratory of High Energy Laser Technology, Changsha 410073, Hunan , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Overall network architecture

Download full size

Fig. 2. Structure of LFEM

Download full size

Fig. 3. Structure of GFEM

Download full size

Fig. 4. Comparison of laser interference restoration results by different algorithms

Download full size

Fig. 5. Restoration effect of actual application scenes. (a) Original images; (b) laser interference images; (c) restored images

Download full size

Fig. 6. Comparison of network convergence during training. (a) Celeba; (b) Building_facade; (c) Airfield; (d) Auto_showroom; (e) Tower

Download full size

Table 1. Datasets setting

View table

Table 1. Datasets setting

Parameter	CelebA dataset	Places365-Standard dataset
Parameter	CelebA dataset	Airfield	Auto_showroom	Building_facade	Tower
Number	5000	5000	5000	5000	5000
Laser example
Laser size /（pixel×pixel）	96×96	96×96	96×96	96×96	96×96
Synthetic image example
Synthetic image size /（pixel×pixel）	256×256	256×256	256×256	256×256	256×256

Table 2. Quantitative evaluation results of laser interference restoration by different algorithms

View table

Table 2. Quantitative evaluation results of laser interference restoration by different algorithms

Algorithm	PSNR /dB					SSIM
Algorithm	CelebA	Airfield	Auto_showroom	Building_facade	Tower	CelebA	Airfield	Auto_showroom	Building_facade	Tower
Restormer^［7］	10.06	29.04	10.17	11.02	10.77	0.316	0.969	0.250	0.258	0.315
MISF^［14］	30.26	28.74	28.43	28.06	26.48	0.976	0.975	0.974	0.974	0.970
ZGAN^［21］	34.46	33.63	31.79	31.93	31.05	0.971	0.974	0.967	0.972	0.957
Proposed	38.31	34.02	35.07	35.53	34.90	0.992	0.983	0.983	0.986	0.983

Table 3. Comparison of model parameters and time
View table
Table 3. Comparison of model parameters and time
Algorithm Params /10⁶ Train time /h Test time /s
Restormer^［7］ 26.12 11.0 0.069
MISF^［14］ 12.28 6.0 0.052
ZGAN^［21］ 51.07 4.0 0.067
Proposed 39.31 2.5 0.062

Table 4. Results of ablation experiments for LFEM and GFEM
View table
Table 4. Results of ablation experiments for LFEM and GFEM
LFEM GFEM PSNR /dB SSIM
√ 33.04 0.977
√ 32.42 0.969
√ √ 38.31 0.992

Table 5. Results of ablation experiments for CAM and SAM
View table
Table 5. Results of ablation experiments for CAM and SAM
CAM SAM PSNR /dB SSIM
32.42 0.969
√ 33.05 0.981
√ 33.20 0.978
√ √ 38.31 0.992

Tools

Get Citation

Copy Citation Text

Haoqian Wang, Ju Liu, Teng Li, Zhongjie Xu, Xiang'ai Cheng, Zhongyang Xing. Performance Improvement of Laser Interference Image Restoration Based on Multi-Scale Feature Fusion[J]. Laser & Optoelectronics Progress, 2025, 62(4): 0437004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites