Depth Estimation Based on Spatial Geometry in Focal Stacks

Fig. 2. Profile of the focal stack and focus measure. (a) The 13th layer of the focal stack; (b) the 13th layer of the focus measure; (c) the 13th layer of the mutual structure; (d) horizontal profile; (e) gradient of profile; (f) structural value

Download full size

Fig. 3. Profile along Z-axis direction. (a) Scene; (b) pixel; (c) FM

Download full size

Fig. 4. Flow chart of Split Bregman method

Download full size

Fig. 5. Effect of parameters $N_{i t e r}$ , $λ$ , $μ$ , and $α$ , the default values have been set to $μ = 50, λ = 0.8, α = 10^{4}, N_{i t e r} = 8$

Download full size

Fig. 6. Comparison of optimization result. (a) Initial depth; (b) RFV optimization; (c) proposed optimization method

Download full size

Fig. 7. Experimental results on part scenes of HCI dataset. (a) Scene; (b) initial depth; (c) MuGIF; (d) RFV; (e) Ours; (f) GT

Download full size

Fig. 8. Experimental results on real dataset. (a) Scene; (b) initial depth; (c) MuGIF; (d) RFV; (e) Ours

Download full size

Table 1. Description of experimental dataset

View table

Table 1. Description of experimental dataset

Dataset	Scene	Number of layers	Dimension
Synthetic	HCI	28	$512 \times 512$
Real	Motion	14	$518 \times 518$
	Balls	25	$320 \times 320$
	Kitchen	12	$518 \times 518$
	Keyboard	32	$360 \times 360$
	Home1-4	49	$320 \times 320$
	Home2-3	49	$320 \times 320$
	Microkitchen1_44	49	$320 \times 320$
	Pillowroom1_8	49	$320 \times 320$

Table 2. Quantitative index on part scenes of HCI dataset

View table

Table 2. Quantitative index on part scenes of HCI dataset

Scene	RMSE			Corr			SSIM
Scene	MuGIF	RFV	Ours	MuGIF	RFV	Ours	MuGIF	RFV	Ours
Antinous	0.9981	0.7779	0.7628	0.3507	0.5174	0.5253	0.2640	0.2881	0.3640
Boardgames	0.3036	0.2306	0.2575	0.7852	0.8902	0.9327	0.6183	0.6820	0.6183
Boxes	0.0943	0.1577	0.1737	0.6640	0.4302	0.3627	0.5746	0.7017	0.7034
Cotton	0.6051	0.3849	0.3672	0.3421	0.7213	0.7907	0.7051	0.5301	0.6057
Dino	0.2499	0.2010	0.2002	0.8628	0.8857	0.9274	0.6749	0.6014	0.6783
Dishes	2.3292	2.0300	1.8100	0.6836	0.7036	0.7687	0.7126	0.4253	0.4626
Greek	0.8296	0.7202	0.7166	0.4506	0.5329	0.5428	0.3882	0.2796	0.2761
Medieval2	0.6085	0.5880	0.6286	0.8896	0.8789	0.8917	0.6033	0.5628	0.6709
Museum	0.5417	0.4584	0.4057	0.7386	0.7890	0.8432	0.5827	0.5265	0.5922
Pillows	0.2563	0.1893	0.1744	0.6718	0.7819	0.8571	0.8314	0.8049	0.8375
Platonic	1.2301	1.2540	1.2267	0.8754	0.8752	0.8766	0.4702	0.4709	0.4375
Sideboard	0.9688	0.6437	0.6744	0.7906	0.8926	0.8936	0.3920	0.4388	0.4760
Table	0.3741	0.2569	0.2521	0.5684	0.7714	0.8176	0.5599	0.5794	0.6046
Tomb	3.5769	1.0681	1.0512	0.9805	0.9922	0.9931	0.3973	0.3652	0.3979
Town	2.7154	1.5993	1.4801	0.4825	0.9001	0.9220	0.2512	0.2395	0.2755
Vinyl	0.1993	0.1828	0.1800	0.5931	0.6539	0.7203	0.7517	0.7072	0.7517

Tools

Get Citation

Copy Citation Text

Tianqi Luo, Xiaojuan Deng, Chang Liu, Jun Qiu. Depth Estimation Based on Spatial Geometry in Focal Stacks[J]. Acta Optica Sinica, 2023, 43(20): 2010001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Mar. 8, 2023

Accepted: May. 6, 2023

Published Online: Oct. 13, 2023

The Author Email: Qiu Jun (qiujun@bistu.edu.cn)

DOI:10.3788/AOS230645

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology