Journal of Applied Optics, Volume. 43, Issue 3, 436(2022)
Patch-match binocular 3D reconstruction based on deep learning
Figures & Tables(6)
Fig. 4. Disparity and heteroscedastic uncertainty of PSMNU network output PSMNU网络输出的视差 和异方差不确定度
Table 1. Quantitative analysis results of patch-match algorithm and proposed algorithm
View tableView in ArticleTable 1. Quantitative analysis results of patch-match algorithm and proposed algorithm
DataSet Algorithm Thresh/ pixel Nonocc/ % Disc/ % All/ % Time/ s Tsukuba Patch-match[ 1 ]1.0 4.21 10.56 4.44 286.795 Ours 1.0 4.16 10.46 4.43 221.337 CSCA[ 20 ]1.0 5.03 10.02 5.80 0.24 Venus Patch-match[ 1 ]1.0 1.53 9.44 2.12 458.980 Ours 1.0 1.40 8.64 1.94 383.256 CSCA[ 20 ]1.0 1.49 8.53 2.28 0.41 Teddy Patch-match[ 1 ]1.0 6.60 14.64 12.60 502.908 Ours 1.0 6.33 13.81 12.19 406.529 CSCA[ 20 ]1.0 7.21 17.39 14.77 0.39 Cones Patch-match[ 1 ]1.0 2.75 7.68 7.77 466.620 Ours 1.0 2.58 7.31 7.58 297.364 CSCA[ 20 ]1.0 5.98 15.32 14.96 0.47
Tools
Get Citation
Copy Citation Text
Lizheng SONG, Dongyun LIN, Xiafu PENG, Tengfei LIU. Patch-match binocular 3D reconstruction based on deep learning[J]. Journal of Applied Optics, 2022, 43(3): 436
Paper Information
Category: OE INFORMATION ACQUISITION AND PROCESSING
Received: Nov. 19, 2021
Accepted: --
Published Online: Jun. 7, 2022
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20