Deep Stereo Matching with Superpixel Based Feature and Cost.

Kai Zeng,Hui Zhang,Wei Wang,Yaonan Wang,Jianxu Mao

DOI: https://doi.org/10.1007/978-981-99-8432-9_1

2024-01-01

Abstract:Previous stereo methods achieved state-of-the-art performances but are still difficult to handle the well-known edge-fattening issue at depth discontinuity regions. In this paper, we proposed differentiable superpixel-based feature and cost aggregation (DSFCA) networks for stereo matching. More specifically, we generated the superpixel maps using the simple linear iterative clustering (SLIC) method and used the efficient feature extraction network to extract the dynamic scale unary features of stereo pair images. Next, we exploited the edge or contour feature information of superpixel maps to aggregate the unary features using the differentiable superpixel-based feature aggregation (DSFA) module. The aggregated features better represent the similarity of stereo pair image feature maps and produce high-quality initial cost volumes. Furthermore, the matching cost is also aggregated by the proposed differentiable superpixel-based cost aggregation (DSCA) module. The experimental results demonstrate that the proposed method outperforms previous methods on the KITTI 2012, KITTI 2015, and SceneFlow datasets.