Acta Optica Sinica, Volume. 40, Issue 9, 0915001(2020)
Stereo Matching Based on Guidance Image and Adaptive Support Region
Figures & Tables(13)
Fig. 3. Disparity maps of three cost computation methods. (a) C1; (b) C2; (c) proposed gradient calculation method
Fig. 4. Weighted averages for all regions and non-occluded regions. (a) Avgerr; (b) RMSE
Fig. 5. Weighted average after disparity refinement on each step. (a)(b) Avgerr; (c)(d) RMSE
Fig. 6. Comparison of disparity results. (a) Adirondack; (b) Jadeplant; (c) Piano; (d) Motorcycle; (e) Recycle
Table 1. Weighted average of errors before and after disparity refinement and reduced percentage (disparity map of proposed gradient calculation method)
View tableTable 1. Weighted average of errors before and after disparity refinement and reduced percentage (disparity map of proposed gradient calculation method)
Parameter Weighted average /pixel Reduced percentage / % Before disparity refinement After disparity refinement Avgerr(all) 21.1 11.30 46.4 Avgerr(nonocc) 12.1 7.81 35.5 RMSE(all) 47.2 27.70 41.3 RMSE(nonocc) 31.6 20.90 33.9
Table 2. Weighted average of errors before and after disparity refinement and reduced percentage (disparity map of C1)
View tableTable 2. Weighted average of errors before and after disparity refinement and reduced percentage (disparity map of C1)
Parameter Weighted average /pixel Reduced percentage / % Before disparity refinement After disparity refinement Avgerr(all) 22.8 12.40 45.6 Avgerr(nonocc) 13.6 8.63 36.5 RMSE(all) 49.6 29.80 39.9 RMSE(nonocc) 34.5 22.90 33.6
Table 3. Weighted average of errors before and after disparity refinement and reduced percentage (disparity map of C2)
View tableTable 3. Weighted average of errors before and after disparity refinement and reduced percentage (disparity map of C2)
Parameter Weighted average /pixel Reduced percentage / % Before disparity refinement After disparity refinement Avgerr(all) 24.5 14.9 39.2 Avgerr(nonocc) 15.1 10.7 29.1 RMSE(all) 51.5 34.6 32.8 RMSE(nonocc) 36.3 27.1 25.3
Table 4. Comparison of Avgerr in all regionspixel
View tableTable 4. Comparison of Avgerr in all regionspixel
Image name LE-ELAS IEBIMst SPS SM-AWP DSGCA DoGGuided Proposed algorithm Adirondack 9.31 27.30 6.51 10.50 7.68 20.10 6.40 ArtL 5.90 15.10 15.20 19.90 21.70 28.00 9.00 Jadeplant 64.50 55.60 40.00 62.70 45.00 56.50 26.10 Motorcycle 7.24 5.54 8.35 11.00 10.60 13.80 8.11 MotorcycleE 7.65 8.21 8.45 12.50 10.40 16.80 11.40 Piano 6.25 6.40 12.00 9.08 11.50 13.40 6.15 PianoL 9.69 18.90 25.00 29.70 24.50 37.30 34.00 Pipes 12.80 11.80 16.10 21.10 19.90 23.80 14.90 Playroom 10.10 18.00 25.20 20.70 24.60 30.30 10.50 Playtable 23.90 17.90 15.70 9.50 34.50 30.80 16.70 PlaytableP 4.27 4.95 12.40 9.75 14.80 13.00 10.00 Recycle 7.39 5.29 8.81 7.18 7.56 9.13 4.20 Shelves 8.48 17.10 23.70 11.40 17.30 19.00 9.97 Teddy 2.98 5.31 8.01 9.44 12.20 13.40 3.35 Vintage 14.00 10.90 53.70 16.80 43.80 13.60 10.90 Australia 15.20 9.17 8.64 19.10 16.60 18.20 12.00 AustraliaP 6.94 5.54 8.77 18.20 12.40 12.60 8.31 Bicycle2 6.68 7.54 11.40 16.00 12.90 17.60 13.70 Classroom2 24.60 27.90 20.20 29.30 32.60 34.90 9.09 Classroom2E 69.60 55.00 27.00 51.10 39.30 76.30 67.10 Computer 12.40 13.80 22.20 22.50 20.60 22.10 13.20 Crusade 21.70 74.30 50.80 91.80 49.50 73.40 36.30 CrusadeP 21.00 74.60 50.20 94.90 50.50 71.30 35.60 Djembe 2.73 2.10 3.65 7.33 5.71 6.64 2.98 DjembeL 13.80 29.10 17.20 31.80 24.50 39.00 19.50 Hoops 22.80 45.00 38.70 37.70 36.30 56.60 23.00 Livingroom 10.30 9.49 30.40 16.80 22.90 25.90 7.18 Newkuba 16.20 13.30 20.30 28.50 23.20 28.70 11.30 Plants 43.30 23.30 26.20 32.20 27.70 33.90 25.80 Staircase 21.30 30.90 39.40 36.40 39.80 57.50 29.80 Weighted average 15.95 20.95 20.70 26.50 22.80 29.20 15.20
Table 5. Comparison of Avgerr in non-occluded regionspixel
View tableTable 5. Comparison of Avgerr in non-occluded regionspixel
Image name LE-ELAS IEBIMst SPS SM-AWP DSGCA DoGGuided Proposed algorithm Adirondack 8.46 26.10 3.57 6.31 3.25 15.20 4.84 ArtL 3.83 4.67 5.34 9.65 5.95 9.57 4.62 Jadeplant 41.10 41.90 22.80 31.80 18.90 27.10 16.10 Motorcycle 5.12 2.72 3.11 4.71 3.60 5.64 4.58 MotorcycleE 5.80 4.99 3.15 6.39 3.41 8.31 7.72 Piano 5.54 5.69 9.34 6.68 7.17 8.09 5.20 PianoL 8.97 17.50 22.90 28.40 21.10 32.40 34.40 Pipes 7.44 5.47 6.78 10.60 7.23 9.67 7.53 Playroom 8.76 12.90 12.50 9.08 9.36 14.00 5.05 Playtable 22.40 14.80 9.70 5.09 29.40 24.50 13.00 PlaytableP 3.47 3.26 7.64 5.18 7.94 5.32 5.67 Recycle 6.93 4.99 6.27 3.86 3.80 5.56 3.37 Shelves 8.26 16.40 22.30 9.73 14.70 16.20 9.49 Teddy 2.29 2.64 1.52 3.64 3.51 4.15 2.15 Vintage 13.10 10.40 52.60 10.70 39.70 15.00 9.64 Australia 13.40 6.53 5.32 13.50 11.00 12.30 8.48 AustraliaP 5.27 3.36 5.48 12.70 6.75 6.62 5.70 Bicycle2 4.88 5.04 7.70 11.00 7.01 11.20 10.50 Classroom2 19.30 19.30 5.60 17.50 13.70 16.30 5.35 Classroom2E 66.50 45.70 12.50 41.80 21.50 62.60 64.80 Computer 6.06 3.41 8.05 11.00 5.90 6.83 3.92 Crusade 15.60 51.30 15.10 70.10 6.72 34.00 19.20 CrusadeP 13.70 46.40 13.10 72.30 5.85 30.60 15.30 Djembe 1.94 1.52 1.84 4.01 2.78 3.65 2.23 DjembeL 13.30 29.20 16.10 28.30 22.20 37.00 19.20 Hoops 18.20 39.20 22.80 25.80 17.20 35.00 16.90 Livingroom 9.62 8.77 19.40 7.88 11.90 13.40 6.19 Newkuba 12.80 8.19 12.60 21.50 11.10 14.20 7.60 Plants 35.40 16.90 14.50 19.70 14.10 19.10 17.90 Staircase 19.10 27.60 27.50 21.70 23.80 34.40 24.40 Weighted average 12.43 15.10 11.00 17.36 10.23 15.85 10.36
Table 6. Comparison of RMSE in all regionspixel
View tableTable 6. Comparison of RMSE in all regionspixel
Image name LE-ELAS IEBIMst SPS SM-AWP DSGCA DoGGuided Proposed algorithm Adirondack 25.40 58.90 23.00 30.30 25.10 48.90 19.20 ArtL 16.80 38.80 39.90 38.70 48.60 59.50 22.80 Jadeplant 120.00 128.00 95.80 119.00 102.00 118.00 62.60 Motorcycle 23.30 20.10 29.10 33.80 32.50 39.40 23.20 MotorcycleE 24.00 27.40 29.80 36.90 32.40 44.60 28.90 Piano 13.90 13.90 32.10 19.30 29.20 31.60 12.40 PianoL 18.70 39.00 53.80 56.30 50.80 65.80 65.30 Pipes 30.70 30.20 41.00 47.30 47.40 52.60 35.10 Playroom 25.90 38.00 59.50 47.80 58.30 66.70 27.60 Playtable 52.30 38.60 43.30 25.00 68.30 59.50 40.10 PlaytableP 12.60 14.20 37.20 25.50 38.40 34.40 28.00 Recycle 17.60 17.50 26.50 22.50 23.30 25.70 10.60 Shelves 15.30 30.40 44.40 23.80 33.50 33.90 19.40 Teddy 8.32 18.60 27.80 25.90 34.90 36.70 13.60 Vintage 27.40 28.80 131.00 45.80 105.00 66.10 27.70 Australia 34.90 27.90 30.00 46.60 40.80 45.20 32.60 AustraliaP 26.60 22.90 30.90 45.80 35.90 38.30 27.50 Bicycle2 21.00 23.50 32.10 37.70 32.60 40.20 32.30 Classroom2 55.80 68.70 60.90 64.10 82.60 85.40 27.30 Classroom2E 112.00 106.00 76.40 93.10 88.30 148.00 127.00 Computer 28.20 34.10 53.70 42.30 46.70 48.00 31.70 Crusade 58.70 156.00 141.00 152.00 131.00 151.00 79.40 CrusadeP 59.00 160.00 140.00 156.00 134.00 150.00 80.10 Djembe 8.10 7.30 15.00 27.80 19.70 21.50 9.28 DjembeL 31.00 58.60 42.80 61.30 51.40 66.40 42.70 Hoops 51.70 77.90 78.20 72.60 73.70 99.90 50.10 Livingroom 23.00 24.90 67.90 39.80 52.80 58.60 17.60 Newkuba 53.30 38.10 65.20 78.20 68.00 81.90 33.70 Plants 72.60 54.60 59.30 63.90 62.90 69.40 55.70 Staircase 46.00 48.20 85.80 73.60 78.70 102.00 51.20 Weighted average 36.25 48.15 54.85 55.20 56.15 64.50 35.70
Table 7. Comparison of RMSE in non-occluded regionspixel
View tableTable 7. Comparison of RMSE in non-occluded regionspixel
Image name LE-ELAS IEBIMst SPS SM-AWP DSGCA DoGGuided Proposed algorithm Adirondack 24.60 58.00 16.40 21.30 13.20 42.20 15.80 ArtL 13.80 15.60 17.20 22.70 17.90 28.30 14.10 Jadeplant 91.90 121.00 75.70 81.50 63.60 75.70 46.80 Motorcycle 18.30 11.40 15.50 19.10 14.90 21.40 14.80 MotorcycleE 20.00 20.90 15.50 24.00 14.50 28.80 22.30 Piano 13.10 12.70 27.50 14.50 20.20 19.40 10.50 PianoL 17.90 37.40 52.00 56.20 47.30 60.90 66.80 Pipes 23.30 19.20 24.50 32.20 24.60 29.60 24.50 Playroom 25.50 31.50 35.80 27.20 25.80 37.50 13.90 Playtable 51.20 32.40 33.10 14.30 63.90 52.20 33.70 PlaytableP 11.30 9.20 29.10 15.30 25.60 15.50 15.30 Recycle 17.10 16.90 21.90 12.90 13.60 18.10 8.36 Shelves 15.10 29.40 43.50 20.20 29.80 29.70 18.60 Teddy 6.86 10.70 6.63 12.10 14.40 14.20 9.72 Vintage 26.30 29.00 134.00 26.20 104.00 50.70 25.00 Australia 31.50 21.30 22.40 37.80 31.00 35.40 25.20 AustraliaP 22.80 16.40 23.50 36.70 24.90 25.90 22.40 Bicycle2 17.10 17.40 26.00 29.90 22.00 30.70 27.30 Classroom2 49.80 57.50 26.10 48.50 47.10 52.60 19.50 Classroom2E 112.00 99.20 53.30 86.30 59.10 136.00 126.00 Computer 16.60 9.47 30.00 22.50 17.50 20.50 11.30 Crusade 49.80 131.00 76.70 127.00 32.90 84.20 51.00 CrusadeP 48.60 123.00 71.20 131.00 31.10 79.20 37.90 Djembe 5.73 5.54 9.13 17.90 11.80 14.00 7.17 DjembeL 30.80 59.20 41.70 55.00 49.20 65.10 43.00 Hoops 47.50 72.70 59.70 60.10 46.30 73.20 43.40 Livingroom 22.70 24.30 51.80 21.30 32.60 35.30 16.00 Newkuba 52.00 26.50 54.00 73.80 42.50 50.50 27.90 Plants 62.40 45.00 42.40 46.60 42.90 47.90 44.20 Staircase 42.00 41.80 75.40 56.20 60.60 66.20 43.50 Weighted average 31.45 39.15 37.45 41.20 31.45 41.50 26.65
Tools
Get Citation
Copy Citation Text
Lingyin Kong, Jiangping Zhu, Sancong Ying. Stereo Matching Based on Guidance Image and Adaptive Support Region[J]. Acta Optica Sinica, 2020, 40(9): 0915001
Paper Information
Category: Machine Vision
Received: Nov. 25, 2019
Accepted: Jan. 17, 2020
Published Online: May. 6, 2020
The Author Email: Sancong Ying (yingsancong@scu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20