Depth Estimation Based on Single-Lens Point Spread Function Regulation

Fig. 8. Comparison of depth estimation performance of models at different depths. (a) Prediction depth of each model at different depths; (b) MAE between predicted and estimated values at different depths

Download full size

Fig. 9. Visualization comparison of ablation experiment results of proposed method

Download full size

Fig. 10. Depth estimation performance of model on FlyingThings3D

Download full size

Table 1. Comparison between different imaging models and rendering method
View table
Table 1. Comparison between different imaging models and rendering method
Method PSNR SSIM
Thin lens model 25.19 0.811
Our model 28.68 0.899

Table 2. Quantitative comparison of proposed method with other methods on NYU Depth V2 dataset

View table

Table 2. Quantitative comparison of proposed method with other methods on NYU Depth V2 dataset

Method	MAE	RMSE	E_Silog	$δ$ <1.25	$δ$ <1.25²	$δ$ <1.25³
DFD^［12］	0.206	0.299	0.019	0.914	0.936	0.965
Wu et al.^［20］	0.197	0.278	0.017	0.924	0.945	0.962
Ikoma et al.^［26］	0.171	0.259	0.016	0.938	0.963	0.990
Qian et al.^［37］	0.287	0.396	0.033	0.866	0.973	0.995
Ours	0.155	0.231	0.014	0.947	0.975	0.994

Table 3. Quantitative comparison of ablation experimental results of proposed methods

View table

Table 3. Quantitative comparison of ablation experimental results of proposed methods

Method	MAE	RMSE	E_Silog	$δ$ <1.25	$δ$ <1.25²	$δ$ <1.25³
With initial single lens	0.242	0.316	0.020	0.915	0.941	0.966
Without preprocessing method	0.157	0.240	0.015	0.943	0.971	0.994
With preprocessing method	0.155	0.231	0.014	0.947	0.975	0.998

Table 4. Quantitative evaluation results of model on FlyingThings3D
View table
Table 4. Quantitative evaluation results of model on FlyingThings3D
Method MAE RMSE E_Silog $δ$ <1.25 $δ$ <1.25² $δ$ <1.25³
Without preprocessing method 0.226 0.383 0.023 0.927 0.977 0.985
With preprocessing method 0.221 0.377 0.024 0.929 0.974 0.987

Tools

Get Citation

Copy Citation Text

Zaiwu Sun, Fanjiao Tan, Pengliang Yu, Zongling Li, Rongshuai Zhang, Changjian Yang, Qingyu Hou. Depth Estimation Based on Single-Lens Point Spread Function Regulation[J]. Acta Optica Sinica, 2025, 45(11): 1111003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites