SURE: Semi-Dense Uncertainty-REfined Feature Matching
Sicheng Li, Zaiwang Gu, Jie Zhang, Qing Guo, Xudong Jiang, Jun Cheng
AI summary
Problem
Existing semi-dense feature matching methods lack explicit reliability estimation, causing overconfident errors in challenging scenarios like textureless regions or large viewpoint changes, while often prioritizing accuracy over computational efficiency.
Approach
SURE employs an evidential regression head to jointly predict correspondence coordinates and both aleatoric and epistemic uncertainties using lightweight 1D heatmaps, alongside a spatial fusion module that enhances local feature precision with minimal overhead.
Key results
- Surpasses state-of-the-art semi-dense matchers in relative pose estimation accuracy on MegaDepth and ScanNet benchmarks
- Achieves superior computational efficiency with a 62.8 ms runtime on standard hardware
- Provides principled uncertainty estimates that effectively filter unreliable matches in low-texture and high-viewpoint-change scenarios
- Introduces a novel evidential regression head and spatial fusion module that reduce computational overhead while improving sub-pixel refinement
Why it matters
Enables more robust and reliable visual localization and 3D reconstruction in real-time robotic vision applications by providing trustworthy match confidence without sacrificing speed.
Abstract
Establishing reliable image correspondences is essential for many robotic vision problems. However, existing methods often struggle in challenging scenarios with large viewpoint changes or textureless regions, where incorrect cor- respondences may still receive high similarity scores. This is mainly because conventional models rely solely on fea- ture similarity, lacking an explicit mechanism to estimate the reliability of predicted matches, leading to overconfident errors. To address this issue, we propose SURE, a Semi- dense Uncertainty-REfined matching framework that jointly predicts correspondences and their confidence by modeling both aleatoric and epistemic uncertainties. Our approach in- troduces a novel evidential head for trustworthy coordinate regression, along with a lightweight spatial fusion module that enhances local feature precision with minimal overhead. We evaluated our method on multiple standard benchmarks, where it consistently outperforms existing state-of-the-art semi-dense matching models in both accuracy and efficiency. our code will be available on https://github.com/LSC-ALAN/SURE.