Lune

CVPR Findings2026

Unleashing the Potential of Event-Based Stereo Via Coarse-to-Fine Bio-Inspired Regression

Haihao Zhang, Siwei Dong, Jianing Li, Rui Zhao, Yunjian Zhang, Geng Qin, Lin Zhu

Abstract

Event cameras provide high temporal resolution and dynamic range, making them well-suited for stereo matching under fast motion and extreme lighting conditions, which overcomes many challenges in practical applications. However, the inherent sparsity of event data makes it challenging to directly regress accurate disparities over a wide range, posing significant challenges to learning-based methods. To address this, we draw inspiration from the hierarchical perception mechanism of the human visual system, which perceives coarse structures before refining details. We explore applying this mechanism to the regression stage and propose a novel coarse-to-fine framework. Specifically, the disparity range is discretized into multiple coarse intervals, corresponding to different disparity levels. In the coarse regression, the model distinguishes which level each pixel belongs to. In the fine regression, the prior information of coarse regression makes it much easier to perform fine regression. The bottleneck caused by the wide range of disparity is overcome, which greatly unleashes the potential of event-based stereo matching. Additionally, to ensure the accuracy of prior information provided by the coarse stage, we introduce a disparity refinement & error correction module. Extensive experiments on MVSEC and DSEC demonstrate that our method consistently outperforms state-of-the-art approaches, achieving superior performance in real-world event-based stereo matching tasks.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

Questions to start from

Lune can draw on

the full paper · related work

Get the answer

Free to start. No credit card required.

Lineage

Builds onCited byLeft to right by year; the more cited, the higher.

Builds on19

  1. Learning an Event Sequence Embedding for Dense Event-Based Deep StereoStepan Tulyakov, François Fleuret, Martin Kiefel, Peter V. Gehler et al.ICCV 2019 · 122 citations
  2. Event-based Video Reconstruction via Potential-assisted Spiking Neural NetworkLin Zhu, Xiao Wang, Yi Chang, Jianing Li et al.CVPR 2022 · 109 citations
  3. Nonuniform-to-Uniform Quantization: Towards Accurate Quantization via Generalized Straight-Through EstimationZechun Liu, Kwang-Ting Cheng, Dong Huang, Eric P. Xing et al.CVPR 2022 · 108 citations
  4. Selective-Stereo: Adaptive Frequency Information Selection for Stereo MatchingXianqi Wang, Gangwei Xu, Hao Jia, Xin YangCVPR 2024 · 64 citations
  5. CF-DETR: Coarse-to-Fine Transformers for End-to-End Object DetectionXipeng Cao, Peng Yuan, Bailan Feng, Kun NiuAAAI 2022 · 59 citations

Related papers

  1. Enhanced Event-Based Dense Stereo via Cross-Sensor Knowledge DistillationHaihao Zhang, Yunjian Zhang, Jianing Li, Lin Zhu et al.ICCV 2025 · 1 citation
  2. Deep Event Stereo Leveraged by Event-to-Image TranslationSoikat Hasan Ahmed, Hae Woong Jang, S. M. Nadim Uddin, Yong Ju JungAAAI 2021 · 41 citations
  3. DERD-Net: Learning Depth from Event-based Ray DensitiesDiego de Oliveira Hitzges, Suman Ghosh, Guillermo GallegoNeurIPS 2025 · 6 citations
  4. Event-Image Fusion Stereo Using Cross-Modality Feature PropagationHoonhee Cho, Kuk-Jin YoonAAAI 2022 · 34 citations
  5. Event-Intensity Stereo: Estimating Depth by the Best of Both WorldsS. Mohammad Mostafavi I., Kuk-Jin Yoon, Jonghyun ChoiICCV 2021 · 45 citations
Dusk over the sea between two cliffs drawn in fine vertical lines