Abstract | ||
---|---|---|
Stereo matching is a challenging yet important task to various computer vision applications, e.g. 3D reconstruction, augmented reality, and autonomous vehicles. In this paper, we present a novel image-based convolutional neural network (CNN) for dense disparity estimation using stereo image pairs. In order to achieve precise and robust stereo matching, we introduce a feature extraction module that learns both local and global information. These features are then passed through an hour-glass structure to generate disparity maps from lower resolution to full resolution. We test the proposed method in several datasets including indoor scenes and synthetic scenes. Experimental results demonstrate that the proposed method outperforms the state-of-the-art methods in several datasets. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1109/VCIP47243.2019.8965761 | 2019 IEEE Visual Communications and Image Processing (VCIP) |
Keywords | Field | DocType |
stereo matching,convolutional neural network,feature extraction,dense depth map | Stereo matching,Computer vision,Convolutional neural network,End-to-end principle,Computer science,Image based,Feature extraction,Augmented reality,Artificial intelligence,Artificial neural network,3D reconstruction | Conference |
ISSN | ISBN | Citations |
1018-8770 | 978-1-7281-3724-7 | 0 |
PageRank | References | Authors |
0.34 | 3 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shuqiao Sun | 1 | 0 | 0.68 |
Rongke Liu | 2 | 127 | 35.79 |
Qiuchen Du | 3 | 2 | 1.76 |
Shantong Sun | 4 | 1 | 2.40 |
Shaoli Kang | 5 | 0 | 0.34 |