Image Restoration Based on Haze Feature Supervision and Super-Resolution Reconstruction in Nighttime-Driving Scenes

Table 1. Comparison of metrics for dehazing results of different algorithms
View table
Table 1. Comparison of metrics for dehazing results of different algorithms
Method PSNR /dB SSIM Params /10⁶ FLOPs /10⁹
NDIM 13.5891 0.5003 5.4 5.2
GMLC 16.2001 0.5506 7.3 4.3
PWAB 22.5658 0.6758 6.6 4.2
MRP 20.4592 0.7315 7.3 3.9
OSFD 21.1073 0.6432 7.9 5.2
HSRDN 27.4500 0.9416 4.7 2.8

Table 2. Comparison of metrics for residual and HSAM ablation experiments
View table
Table 2. Comparison of metrics for residual and HSAM ablation experiments
Method PSNR /dB SSIM
SRCNN 17.57 0.7979
SRCNN+Res 18.18 0.8167
SRCNN+Res+HSAM 27.45 0.9416

Table 3. Comparison of metrics for ablation experiments between smooth L1 loss function and MSE loss function
View table
Table 3. Comparison of metrics for ablation experiments between smooth L1 loss function and MSE loss function
Loss function PSNR /dB SSIM
MSE 22.87 0.8234
L1 27.45 0.9416

Table 4. Comparison of metrics for ablation experiments between WaveletPooling, AvgPooling, and MaxPooling
View table
Table 4. Comparison of metrics for ablation experiments between WaveletPooling, AvgPooling, and MaxPooling
Method PSNR /dB SSIM
AvgPooling 27.40 0.9410
MaxPooling 27.42 0.9415
WaveletPooling 27.45 0.9416

Tools

Get Citation

Copy Citation Text

Guangye Wu, Fang Liu, Honggang Qu, Lingyu Lei, Qian Ren. Image Restoration Based on Haze Feature Supervision and Super-Resolution Reconstruction in Nighttime-Driving Scenes[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1237007

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites