Cross-Scale Local Stereo Matching Based on Edge Weighting

Fig. 4. Experimental results of proposed algorithm. (a) Left images; (b) ground truth maps; (c) disparity maps of proposed algorithm (without disparity refinement); (d) disparity maps of proposed algorithm (with disparity refinement); (e) mismatched maps of proposed algorithm (with disparity refinement)

Download full size

Fig. 5. Experimental results of different parameter settings. (a) k₁; (b) k₂; (c) λ_n; (d) λ_s

Download full size

Table 1. Experimental parameters
View table
Table 1. Experimental parameters
Parameter τ_c τ_g α k₁ k₂ λ_n λ_s T S λ r
Value 0.02745 0.00784 0.11 0.9 0.7 0.2 0.1 20 4 0.75 9

Table 2. Error matching rates before and after the improvement%
View table
Table 2. Error matching rates before and after the improvement%
Algorithm Tsukuba Venus Teddy Cones Average
GF 2.62 1.72 8.25 3.58 4.04
S+GF 2.30 1.09 6.99 3.23 3.40
Proposedalgorithm 2.12 0.91 6.59 3.06 3.17

Table 3. Comparison of run time of different algorithmss
View table
Table 3. Comparison of run time of different algorithmss
Algorithm Tsukuba Venus Teddy Cones Average
S+GF 0.81 1.56 5.08 5.43 3.22
S+BF 37.60 70.84 232.09 231.78 143.08
Proposedalgorithm 0.93 1.72 5.44 5.74 3.46

Table 4. Average matching error rates of different algorithms on non-occlusion regions%

View table

Table 4. Average matching error rates of different algorithms on non-occlusion regions%

Stereo pairs	CT-GF	CGA-GF	GA-MST	GA-ST	Proposed algorithm
Tsukuba	3.54	2.92	1.76	2.04	2.12
Venus	1.99	2.50	1.24	1.41	0.91
Teddy	8.56	8.23	5.73	6.22	6.59
Cones	4.87	3.75	4.42	4.76	3.06
Aloe	7.04	5.80	4.88	4.83	5.57
Art	12.15	9.79	10.69	10.38	8.56
Baby1	3.31	3.31	8.21	4.49	3.56
Baby2	4.03	3.79	13.54	15.12	2.76
Baby3	4.97	5.28	5.59	3.96	3.85
Books	9.96	9.13	10.66	10.04	8.51
Bowling1	6.18	7.46	19.56	18.88	8.75
Bowling2	8.47	6.90	10.11	10.53	5.27
Cloth1	1.96	1.16	0.63	0.69	1.29
Cloth2	4.33	3.56	4.35	4.35	3.53
Cloth3	2.80	2.05	2.90	2.96	2.28
Cloth4	2.38	1.95	1.88	1.84	1.68
Dolls	7.27	5.71	5.89	5.49	4.80
Flowerpots	9.67	12.92	16.79	12.50	8.54
Lampshade1	10.64	11.63	9.81	9.13	7.10
Lampshade2	10.52	15.97	12.08	9.89	12.46
Laundry	18.45	18.50	11.92	11.90	12.03
Midd1	33.45	37.61	24.43	22.01	32.90
Midd2	32.96	35.53	20.57	18.90	26.61
Moebius	10.92	10.75	7.57	7.30	8.11
Monopoly	18.99	23.43	21.03	20.75	21.68
Plastic	22.23	29.39	45.02	37.39	25.74
Reindeer	8.15	9.36	9.79	7.83	6.05
Rocks1	4.03	3.94	3.35	3.06	3.00
Rocks2	2.74	2.20	2.28	1.99	1.53
Wood1	4.93	4.99	10.18	5.46	3.66
Wood2	2.95	2.96	3.17	4.84	1.78
Average matching error rate	9.18	9.76	10.00	9.06	7.88
Average rank	3.61	3.42	3.39	2.68	1.93

Table 5. Percentage of mismatching pixels in different regions for different algorithms%

View table

Table 5. Percentage of mismatching pixels in different regions for different algorithms%

Algorithm	Tsukuba			Venus			Teddy			Cones			Average
Algorithm	Non-occ	All	Disc		Non-occ	All	Disc	Non-occ	All	Disc	Non-occ	All	Disc
GC+occ^[5]	1.19	2.01	6.24	1.64	2.19	6.75	11.2	17.4	19.8	5.36	12.4	13.0	8.26
Adapt Weight^[10]	1.38	1.85	6.90	0.71	1.19	6.13	7.88	13.3	18.6	3.97	9.79	8.26	6.67
TiwceGF^[17]	1.82	2.99	7.35	0.34	1.53	3.45	6.93	14.74	16.08	3.2	12.29	11.45	6.84
ASSW^[21]	1.81	2.17	7.85	0.32	0.51	3.73	7.02	12.5	17.4	3.21	8.40	8.99	6.16
iFBS^[22]	1.78	2.10	7.57	0.31	0.50	2.17	7.94	12.8	17.1	3.07	8.73	8.46	6.05
SDDS^[23]	3.31	3.62	10.4	0.39	0.76	2.85	7.65	13.0	19.4	3.99	10.0	10.8	7.19
VSW^[24]	1.62	1.88	6.98	0.47	0.81	3.40	8.67	13.3	18.0	3.37	8.85	8.12	6.29
Proposed algorithm	2.02	2.29	8.66	0.26	0.48	2.88	6.11	11.4	15.77	2.94	8.80	8.07	5.81

Tools

Get Citation

Copy Citation Text

Deqiang Cheng, Huandong Zhuang, Wenjie Yu, Chunmeng Bai, Xiaoshun Wen. Cross-Scale Local Stereo Matching Based on Edge Weighting[J]. Laser & Optoelectronics Progress, 2019, 56(21): 211504

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Mar. 20, 2019

Accepted: Apr. 30, 2019

Published Online: Nov. 2, 2019

The Author Email: Deqiang Cheng (chengdq@cumt.edu.cn), Huandong Zhuang (hdzhuang@cumt.edu.cn)

DOI:10.3788/LOP56.211504

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Experimental parameters

Table 1. Experimental parameters

Table 2. Error matching rates before and after the improvement%

Table 2. Error matching rates before and after the improvement%

Table 3. Comparison of run time of different algorithmss

Table 3. Comparison of run time of different algorithmss

Table 4. Average matching error rates of different algorithms on non-occlusion regions%

Table 4. Average matching error rates of different algorithms on non-occlusion regions%

Table 5. Percentage of mismatching pixels in different regions for different algorithms%

Table 5. Percentage of mismatching pixels in different regions for different algorithms%