cs.CVOct 24, 2026

E3tereo: Improving Stereo Matching at Depth Boundaries with Learned Geometric Edges

Authors: Qi Xiong, Qinghua Zhao, Yongqiang Xi

Organizations: Xidian University · Hefei University

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Depth boundaries remain a major source of error in stereo matching. For IGEV on Scene Flow, geometric edge pixels account for only 6% of evaluated pixels but about half of the total absolute disparity error. Photometric edges can respond to texture changes without corresponding depth discontinuities, making them unreliable cues for boundary refinement. We propose E3Stereo, which learns geometric edges from disparity supervision and combines these boundary cues with the backbone disparity and image details to predict full-resolution disparity corrections. The same edge map also emphasizes boundary errors during training, while joint optimization allows the edge predictor to adapt to disparity refinement. On the Scene Flow dataset, E3Stereo reduces end-point error (EPE) at depth boundaries by 9.5%-14.3% across RAFT-Stereo, IGEV, and MonSter, while also reducing overall EPE. After fine-tuning, our IGEV-based model improves upon IGEV on all eleven reported metrics across ETH3D, KITTI 2012, and KITTI 2015. The source code will be made publicly available upon acceptance.

Figures & tables

Figure 1: Disparity errors at depth boundaries.
Figure 2: Overview of E 3 Stereo.
Figure 3: Architecture of GeoEdgeNet.
ConfigurationEPE ↓\downarrowEPE Edge ↓\downarrowParams (M)
Pretrained, frozen GeoEdgeNet0.463.5314.90
Jointly trained full model0.433.2312.78
Table 1: Frozen-predictor and jointly trained configurations on Scene Flow. Results are reproduced from Tab. 5 .
MethodETH3DMiddleburyKITTI 2015KITTI 2012Runtime (ms) ↓\downarrow
Bad1.0 Noc ↓\downarrowBad1.0 All ↓\downarrowRMSE Noc ↓\downarrowBad2.0 Noc ↓\downarrowBad2.0 All ↓\downarrowBad4.0 All ↓\downarrowD1-bg Noc ↓\downarrowD1-all Noc ↓\downarrowD1-bg All ↓\downarrowD1-all All ↓\downarrowOut-2 Noc ↓\downarrowOut-2 All ↓\downarrowOut-3 Noc ↓\downarrowOut-3 All ↓\downarrow
Cost-volume methods
GwcNet [ 16 ]6.426.950.69–––1.611.921.742.112.162.711.321.70320
GANet [ 47 ]6.226.860.75–––1.341.631.481.811.892.501.191.601800
AdaStereo [ 35 ]3.093.340.4413.719.810.92.392.832.593.08––––410
CasStereo [ 11 ]–––18.826.018.5–––––––––
Table 2: Results on four real-world datasets. All metrics are reported as percentages (%), except RMSE, which is reported in pixels. “All” denotes all evaluated pixels, while “Noc” denotes non-occluded pixels. Results for competing methods are taken from official leaderboards or the corresponding publications. † marks methods using large-scale pre-trained foundation models, which are included for reference. Among methods without a foundation prior, the best and second-best results are highlighted in red and yellow, respectively.
Figure 4: Overall EPE and boundary EPE Edge on the Scene Flow test set before and after adding our components.
BackboneEPE ↓\downarrowEPE Edge ↓\downarrowParams (M)Time (ms)
RAFT-Stereo [ 22 ]0.72 →\rightarrow 0.675.14 →\rightarrow 4.4511.12 →\rightarrow 11.30380 →\rightarrow 390
IGEV [ 41 ]0.47 →\rightarrow 0.433.77 →\rightarrow 3.2312.60 →\rightarrow 12.78180 →\rightarrow 190
MonSter [ 7 ]0.37 →\rightarrow 0.353.27 →\rightarrow 2.96388.69 →\rightarrow 388.87450 →\rightarrow 460
Table 3: Results before and after adding our components to three stereo backbones on Scene Flow.
DatasetThreshold (px)Bad- τ\tau (%) ↓\downarrow
IGEVE 3 Stereo
KITTI 201235.145.59
KITTI 201536.046.25
Middlebury Quarter26.216.80
Middlebury Half27.087.43
ETH3D13.613.97
Table 4: Zero-shot stereo matching results on real-world datasets.
ConfigurationEPE ↓\downarrowEPE Edge ↓\downarrowParams (M)Time (ms)
Progressive component addition
IGEV baseline0.473.7712.60180
+ edge-weighted loss (GT edges)0.463.6812.60180
+ GeoEdgeNet0.453.4212.74186
+ GeoRestore0.433.2312.78190
Additional ablations
Table 5: Ablation results on the Scene Flow test set.
Figure 5: Qualitative comparison of edge sources on Scene Flow, ETH3D, and Middlebury. The Scene Flow green boxes highlight appearance contours absent from the reference geometric edge map. GeoEdgeNet suppresses several of these responses.
Figure 6: Qualitative disparity comparisons on Scene Flow in the top two rows and ETH3D in the bottom two rows.
Figure 7: EPE versus distance to depth boundaries across three stereo backbones.
More in this topic

Stereo Matching

Stereo matching estimates depth by finding corresponding points in paired images and measuring their disparity.

See all 57 papers

Explore similar work

CardsList

  1. StereoDiffuer: Diffusion-based Progressive Geometry Modeling with Saliency Attention Perception for Stereo Matching

    Aug 22, 2026Bohan LiStereo Matching

    Shanghai Jiaotong University, No.800 Dongchuan Road, Shanghai, 200240, China · Ningbo Institute of Digital Twin, Eastern Institute of Technology, No. 568, Tongxin Road, Ningbo, 315000, China

  2. Rethinking Monocular Depth Embedding for Generalized Stereo Matching

    Jul 10, 2026Libo Lin, Shuangli Du, Minghua Zhao +3Stereo Matching

    Shaanxi Key Laboratory for Network Computing and Security Technology, Xi’an University of Technology, No. 5, Jinhua South Road, Xi’an, 710048, Shaanxi, China · College of Computer Science, Sichuan University, , Chengdu, 610065, Sichuan, China

  3. STEREOFLOW: Progressive Stereo Matching with StereoDiT and Transition Flow Matching

    Jul 22, 2026Hao Wang, Haoran Geng, Xiaotong Yang +7Flow MatchingDiffusion Transformer

    School of Computer Science and Engineering, Beihang University, Beijing 100191, China · Google · State Key Laboratory of Intelligent Manufacturing Equipment and Technology, Huazhong University of Science and Technology, Wuhan 430074, China +3