Adaptive Stereo Matching Based on Local Information Entropy and Improved AD-Census Transform

Jingfa Lei; Zihan Wei; Yongling Li; Ruhai Zhao; Miao Zhang

doi:10.3788/LOP240968

Journals >Laser & Optoelectronics Progress >Volume 62 >Issue 2 >Page 0215004 > Article

Laser & Optoelectronics Progress
Vol. 62, Issue 2, 0215004 (2025)

Adaptive Stereo Matching Based on Local Information Entropy and Improved AD-Census Transform

Jingfa Lei^{1、2、3、*}, Zihan Wei^1、2, Yongling Li^1、2、3, Ruhai Zhao^1、2, and Miao Zhang^1、2

Author Affiliations

¹School of Mechanical and Electrical Engineering, Anhui Jianzhu University, Hefei 230601, Anhui , China

²Key Laboratory of Intelligent Manufacturing of Construction Machinery, Anhui Education Department, Hefei 230601, Anhui , China

³Sichuan Provincial Key Laboratory of Process Equipment and Control Engineering, Zigong 643000, Sichuan , China

show less

DOI: 10.3788/LOP240968 Cite this Article Set citation alerts

Jingfa Lei, Zihan Wei, Yongling Li, Ruhai Zhao, Miao Zhang. Adaptive Stereo Matching Based on Local Information Entropy and Improved AD-Census Transform[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0215004 Copy Citation Text

show less

Fig. 1. Flow chart of proposed algorithm

Download full size

Fig. 2. Census transformation process. (a) The pixel grayscale value of the original image window; (b) Census transform results

Download full size

Fig. 3. Original image and local entropy image. (a) Original image; (b) local entropy image

Download full size

Fig. 4. Grayscale histogram and information entropy value in three regions. (a) Grayscale histogram of part 1; (b) grayscale histogram of part 2; (c) grayscale histogram of part 3; (d) the entropy of the three regions

Download full size

Fig. 5. Curves of weight

Download full size

Fig. 6. Schematic diagram of path aggregation

Download full size

Fig. 7. Matching results. (a) Original images; (b) standard parallax maps; (c) Traditional Census; (d) AD+Census; (e) proposed algorithm

Download full size

Fig. 8. Average mismatch rates for different noises. (a) Average mismatch rates for salt-and-pepper noise; (b) average mismatch rates for Gaussian noise

Download full size

Parameter	$T_{1}$	$T_{2}$	$λ_{C e n}$	$λ_{A D}$	$θ$	$P_{1}$	$P_{2}^{'}$
Value	1.5	5.5	10	30	2.0	10	150

Table 1. Parameter setting

Noise density	CT	SAD	AD-Census	GRD	Proposed algorithm
0	7.09	7.42	7.20	8.09	6.12
0.03	7.94	10.28	8.32	10.41	6.61
0.06	9.05	12.45	9.27	12.44	7.61
0.09	10.49	14.62	10.23	13.12	8.67
0.12	12.21	17.73	11.96	16.18	10.28

Table 2. Average mismatch rate under salt and pepper noise

Standard deviation	CT	SAD	AD-Census	GRD	Proposed algorithm
0	7.09	7.42	7.20	8.09	6.12
2	7.91	8.04	8.31	9.77	7.27
4	10.01	10.20	9.64	12.49	9.59
6	11.88	12.73	11.63	14.09	11.39
8	14.98	15.53	13.85	17.63	13.54

Table 3. Average mismatch rate under Gaussian noise

Image	Mismatch rate /%
Image	SGM	AD-Census	Ada_SGM	MCT	Proposed algorithm
Average	8.17	6.67	7.20	7.49	5.94
Adirondack	7.19	6.91	5.68	6.85	5.33
Teddy	7.75	6.15	6.14	6.94	5.76
Pipes	13.40	10.00	11.48	12.88	9.25
Recycle	9.89	7.70	8.46	8.24	6.85
Wood2	4.02	3.51	4.89	3.96	2.68
Cloth2	6.74	5.72	6.54	6.07	5.75

Table 4. Mismatch rates of no-occluded region

Image	Mismatch rate /%
Image	SGM	AD-Census	Ada_SGM	MCT	Proposed algorithm
Average	10.54	9.09	9.62	9.93	8.37
Adirondack	7.73	7.54	6.27	7.35	5.84
Teddy	12.55	10.85	10.82	11.90	10.60
Pipes	18.81	15.80	17.01	18.28	14.86
Recycle	11.07	8.83	9.92	9.70	8.16
Wood2	5.76	5.26	6.61	5.70	4.44
Cloth2	7.29	6.27	7.09	6.62	6.30

Table 5. Mismatch rates of all regions

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information