Global Occlusion-Aware Transformer for Robust Stereo Matching
This addresses a bottleneck in stereo matching for computer vision applications, but it is incremental as it builds on existing attention-based methods to specifically improve occlusion handling.
The paper tackled the problem of poor performance in occluded regions in stereo matching by introducing the Global Occlusion-Aware Transformer (GOAT), which uses attention mechanisms to exploit long-range dependencies and occlusion-aware global context, resulting in outstanding performance on benchmarks like SceneFlow, KITTI 2015, and Middlebury, particularly in occluded regions.
Despite the remarkable progress facilitated by learning-based stereo-matching algorithms, the performance in the ill-conditioned regions, such as the occluded regions, remains a bottleneck. Due to the limited receptive field, existing CNN-based methods struggle to handle these ill-conditioned regions effectively. To address this issue, this paper introduces a novel attention-based stereo-matching network called Global Occlusion-Aware Transformer (GOAT) to exploit long-range dependency and occlusion-awareness global context for disparity estimation. In the GOAT architecture, a parallel disparity and occlusion estimation module PDO is proposed to estimate the initial disparity map and the occlusion mask using a parallel attention mechanism. To further enhance the disparity estimates in the occluded regions, an occlusion-aware global aggregation module (OGA) is proposed. This module aims to refine the disparity in the occluded regions by leveraging restricted global correlation within the focus scope of the occluded areas. Extensive experiments were conducted on several public benchmark datasets including SceneFlow, KITTI 2015, and Middlebury. The results show that the proposed GOAT demonstrates outstanding performance among all benchmarks, particularly in the occluded regions.