Weight-Adaptive Cross-Scale Algorithm for Stereo Matching

Peixuan Li; Pengfei Liu; Feidao Cao; Huaici Zhao

doi:10.3788/AOS201838.1215006

Acta Optica Sinica, Volume. 38, Issue 12, 1215006(2018)

Weight-Adaptive Cross-Scale Algorithm for Stereo Matching

Peixuan Li^1,2,3,4、*, Pengfei Liu^1,2,3,4, Feidao Cao^1,2,3,4, and Huaici Zhao^1,3,4、*

Author Affiliations

¹ Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, Liaoning 110016, China

² University of Chinese Academy of Sciences, Beijing 100049, China

³ Laboratory of Opto-Electronic Information Processing, Chinese Academy of Sciences, Shenyang, Liaoning 110016, China

⁴ Key Laboratory of Image Understanding and Computer Vision, Shenyang, Liaoning 110016, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(6)

Fig. 1. Image (blue box) and mismatching points (black box) of certain pixel window when cost aggregation performed at different scales. (a) Original image; (b) second-scale image

Download full size

Fig. 2. Different images and corresponding information entropy. (a) E=0 for pure white image information; (b) E=0 for pure black image; (c) E=1.0413 for 4-grid image; (d) E=1.3476 for 16-grid image; (e) E=5.0542 for image in blue box of Fig. 1(b); (f) E=5.6215 for image in blue box of Fig. 1(a)

Download full size

Fig. 3. Parallax maps of Teddy images by different methods

Download full size

Table 1. Parameters for proposed cross-scale cost matching algorithm
View table
Table 1. Parameters for proposed cross-scale cost matching algorithm
Parameter τ₁ τ₂ α η S λ
Value 0.02745 0.00784 0.11 2 4 0.27

Table 2. Matching error comparison between proposed method and that in Ref. [12] when dead pixel rate is 2%

View table

Table 2. Matching error comparison between proposed method and that in Ref. [12] when dead pixel rate is 2%

Stereo Pairs	S+BOX		AS+BOX
Stereo Pairs	Non-occluded	All	Non-occluded	All
Bowling2	10.07	22.46	9.43	21.64
Baby1	10.92	15.36	8.35	12.68
Cloth3	7.42	11.66	7.01	11.00
Flowerpots	15.59	30.78	14.43	29.75
Lampshade2	30.41	36.58	24.72	31.33
Midd1	46.53	49.43	42.00	45.16
Monopoly	37.08	42.48	25.44	31.78
Plastic	56.12	57.38	55.47	56.77
Rocks1	12.04	17.42	11.94	17.31
Wood1	13.11	19.25	12.75	18.92
Books	15.51	22.22	14.91	21.68
Moebius	16.82	22.64	16.52	22.36
Dolls	7.51	14.72	7.51	14.75
Baby2	6.88	12.12	6.93	12.18
Wood2	16.77	18.22	16.80	18.25
Rocks2	8.79	15.14	8.82	15.19
Teddy	7.18	16.13	7.13	16.01
Cones	3.99	13.56	3.86	13.52
Average	17.93	24.31	16.34	22.80

Table 3. Running time comparison for cost aggregation by guiding filterss

View table

Table 3. Running time comparison for cost aggregation by guiding filterss

Stereo pairs	S+GF	AS+GF
Bowling2	6.75220	7.32107
Baby1	5.43631	5.54525
Cloth3	6.28890	6.33252
Flowerpots	6.56278	6.68801
Lampshade2	7.53322	7.98634
Midd1	6.96293	7.36999
Monopoly	6.57855	6.82937
Plastic	6.94070	7.30089
Rocks1	6.17969	6.27993
Wood1	6.43060	6.46233
Books	6.30772	6.76049
Moebius	6.03519	6.98040
Dolls	5.91785	6.18341
Baby2	5.18777	5.40952
Wood2	6.26264	6.80027
Rocks2	5.78323	5.95239
Teddy	5.55470	6.00646
Cones	5.56898	5.85602
Average	6.237998	6.559148

Tools

Get Citation

Copy Citation Text

Peixuan Li, Pengfei Liu, Feidao Cao, Huaici Zhao. Weight-Adaptive Cross-Scale Algorithm for Stereo Matching[J]. Acta Optica Sinica, 2018, 38(12): 1215006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites