APSIPA Transactions on Signal and Information Processing > Vol 9 > Issue 1

NLCA-Net: a non-local context attention network for stereo matching

Zhibo Rao, Northwestern Polytechnical University, China, Mingyi He, Northwestern Polytechnical University, China, myhe@nwpu.edu.cn , Yuchao Dai, Northwestern Polytechnical University, China, Zhidong Zhu, Northwestern Polytechnical University, China, Bo Li, Northwestern Polytechnical University, China, Renjie He, Northwestern Polytechnical University, China
 
Suggested Citation
Zhibo Rao, Mingyi He, Yuchao Dai, Zhidong Zhu, Bo Li and Renjie He (2020), "NLCA-Net: a non-local context attention network for stereo matching", APSIPA Transactions on Signal and Information Processing: Vol. 9: No. 1, e18. http://dx.doi.org/10.1017/ATSIP.2020.16

Publication Date: 07 Jul 2020
© 2020 Zhibo Rao, Mingyi He, Yuchao Dai, Zhidong Zhu, Bo Li and Renjie He
 
Subjects
 
Keywords
Stereo matchingNon-local attentionGeometry contextGeometry refine
 

Share

Open Access

This is published under the terms of the Creative Commons Attribution licence.

Downloaded: 2955 times

In this article:
I. INTRODUCTION 
II. RELATED WORK 
III. OUR METHOD 
IV. EXPERIMENTS 
V. CONCLUSIONS 

Abstract

Accurate disparity prediction is a hot spot in computer vision, and how to efficiently exploit contextual information is the key to improve the performance. In this paper, we propose a simple yet effective non-local context attention network to exploit the global context information by using attention mechanisms and semantic information for stereo matching. First, we develop a 2D geometry feature learning module to get a more discriminative representation by taking advantage of multi-scale features and form them into the variance-based cost volume. Then, we construct a non-local attention matching module by using the non-local block and hierarchical 3D convolutions, which can effectively regularize the cost volume and capture the global contextual information. Finally, we adopt a geometry refinement module to refine the disparity map to further improve the performance. Moreover, we add the warping loss function to help the model learn the matching rule of the non-occluded region. Our experiments show that (1) our approach achieves competitive results on KITTI and SceneFlow datasets in the end-point error and the fraction of erroneous pixels $({D_1})$; (2) our proposed method particularly has superior performance in the reflective regions and occluded areas.

DOI:10.1017/ATSIP.2020.16