Local Binary Description Combined with Superpixel Segmentation Refinement for Stereo Matching

Yan Liu; Qingwu Li; Guanying Huo; Jun Xing

doi:10.3788/AOS201838.0615003

Acta Optica Sinica, Volume. 38, Issue 6, 0615003(2018)

Local Binary Description Combined with Superpixel Segmentation Refinement for Stereo Matching

Yan Liu^1,2, Qingwu Li^1,2、*, Guanying Huo^1,2, and Jun Xing¹

Author Affiliations

¹ College of Internet of Things Engineering, Hohai University, Changzhou, Jiangsu 213022, China

² Changzhou Key Laboratory of Sensor Networks and Environmental Sensing, Changzhou, Jiangsu 213022, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(10)

Fig. 1. Disparity images of Cones. (a) Left view of Cones; (b) true disparity image; (c) initial disparity image; (d) mismatching test result

Download full size

Fig. 2. Segmentation results for various numbers of superpixel. (a) M=800; (b) M=1600; (c) M=2400; (d) M=3200

Download full size

Fig. 3. Comparison of initial disparity results. (a) Reference image; (b) Cencus method; (c) GD-Cencus method; (d) proposed method

Download full size

Fig. 4. Effect of numbers of superpixel on the (a) speed and (b) quality

Download full size

Fig. 5. Mismatching test result for Cones after disparity refinement. (a) MST method; (b) WCPSP method; (c) DR method; (d) proposed method

Download full size

Fig. 6. Disparity images of non-overlapping region between reference image and match image. (a) Ground truth disparity image; (b) initial disparity image; disparity images using (c) MST method, (d) WCPSP method, (e) DR method and (f) proposed method

Download full size

Fig. 7. Refinement results for Tsukuba, Venus, Teddy, and Cones. (a) Ground truth disparity; (b) MST method; (b) WCPSP method; (c) DR method; (d) proposed method

Download full size

Fig. 8. Comparisons of error rates using WCPSP, BSM, ARW, and proposed method. (a) Percentage of bad pixel for non-occluded; (b) percentage of bad pixel for all

Download full size

Table 1. Running time of four methods

View table

Table 1. Running time of four methods

Item	Tsukuba	Venus	Teddy	Cones
Image resolution /(pixel×pixel)	384×288	434×383	450×375	450×375
Disparity search range /pixel	0-15	0-19	0-59	0-59
MST /s	1.201	2.215	1.123	1.232
WCPSP /s	1.513	2.355	2.012	2.356
DR /s	1.529	1.388	2.855	2.620
Proposed method /s	0.202	0.296	0.405	0.445

Table 2. Comparison of error rate of optimization algorithm%

View table

Table 2. Comparison of error rate of optimization algorithm%

Item	Tsukuba			Venus			Teddy			Cones
Item	E_nonocc	E_all	E_disc	E_nonocc	E_all	E_disc	E_nonocc	E_all	E_disc	E_nonocc	E_all	E_disc
NonPost	11.24	9.06	18.40	4.01	5.61	19.90	9.99	20.81	23.79	4.47	15.64	14.19
MST	11.59	13.11	19.22	4.05	5.66	20.18	10.71	19.91	25.28	5.39	15.92	14.84
WCPSP	3.01	4.00	12.29	0.86	1.54	9.29	3.74	9.69	11.16	3.99	11.18	11.80
DR	5.17	5.61	12.86	1.03	1.69	7.83	6.48	14.20	16.56	3.46	8.90	9.99
Proposed method	4.39	4.66	9.77	0.81	1.27	5.24	6.94	11.77	17.50	2.86	7.97	8.09

Tools

Get Citation

Copy Citation Text

Yan Liu, Qingwu Li, Guanying Huo, Jun Xing. Local Binary Description Combined with Superpixel Segmentation Refinement for Stereo Matching[J]. Acta Optica Sinica, 2018, 38(6): 0615003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Nov. 3, 2017

Accepted: --

Published Online: Jul. 9, 2018

The Author Email: Li Qingwu (li_qingwu@163.com)

DOI:10.3788/AOS201838.0615003

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology