Binocular-Separated Modeling for Efficient Binocular Stereo Matching

Binocular stereo matching is a crucial task in autonomous driving for accurately estimating the depth information of objects and scenes. This task, however, is challenging due to various ill-posed regions within binocular image pairs, such as repeated textures and weak textures which present complex correspondences between the points. Existing methods extract features from binocular input images mainly by relying on deep convolutional neural networks with a substantial number of convolutional layers, which may incur high memory and computation costs, thus making it hard to deploy in real-world applications. Additionally, previous methods do not consider the correlation between view unary features during the construction of the cost volume, thus leading to inferior results. To address these issues, a novel lightweight binocular-separated feature extraction module is proposed that includes a view-shared multi-dilation fusion module and a view-specific feature extractor. Our method leverages a shallow neural network with a multi-dilation modeling module to provide similar receptive fields as deep neural networks but with fewer parameters and better computational efficiency. Furthermore, we propose incorporating the correlations of view-shared features to dynamically select view-specific features during the construction of the cost volume. Extensive experiments conducted on two public benchmark datasets show that our proposed method outperforms the deep model-based baseline method (i.e., 13.6% improvement on Scene Flow and 2.0% on KITTI 2015) while using 29.7% fewer parameters. Ablation experiments show that our method achieves superior matching performance in weak texture and edge regions. The source code will be made publicly available.

I. Introduction

Binocular stereo matching is a vital and demanding task in computer vision that aims to estimate a disparity map from a pair of input binocular stereo images [15], as depicted in Figure 1. This task is particularly critical for autonomous driving, as it provides the vehicle with precise depth information about the objects and scenes for decision-making. Such depth information is also crucial for obstacle detection, localization, mapping, and path planning, helping to ensure the safety and efficient operation of autonomous vehicles [8], [24], [32], [33]. However, this task is highly challenging because the computational complexity of binocular stereo matching, particularly with large images, can be prohibitively high, making the process time-consuming. Fig. 1.

An illustrative example of the binocular stereo matching task. Given the left and right views, this task aims to obtain a disparity map with reference to the left view.