Dense Match Summarization for Faster Two-view Estimation

2025-06-03

AI Summary

While dense matching improves accuracy and robustness in two-view relative pose estimation, it severely slows down RANSAC-based robust estimation. To address this, we propose Matching Summarization—a geometric consistency-driven framework that compresses and filters dense matches while preserving critical geometric constraints. The method is agnostic to underlying matchers and seamlessly integrates with state-of-the-art dense matchers such as LoFTR and SuperPoint+SuperGlue. On standard benchmarks, summarized matches retain only 5–10% of the original correspondences yet achieve over 98% of the pose accuracy attained by full matching. Crucially, RANSAC runtime is reduced by one to two orders of magnitude (10×–100× speedup). This work marks the first approach to jointly optimize the high accuracy of dense matching and the computational efficiency of robust estimation algorithms.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: State Estimation
• Natural Language Processing: Summarization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Security and Privacy: Large-scale security measurements

Abstract

In this paper, we speed up robust two-view relative pose from dense correspondences. Previous work has shown that dense matchers can significantly improve both accuracy and robustness in the resulting pose. However, the large number of matches comes with a significantly increased runtime during robust estimation in RANSAC. To avoid this, we propose an efficient match summarization scheme which provides comparable accuracy to using the full set of dense matches, while having 10-100x faster runtime. We validate our approach on standard benchmark datasets together with multiple state-of-the-art dense matchers.

Problem

• Speed up robust two-view pose estimation
• Reduce runtime of dense match processing
• Maintain accuracy with summarized matches

Innovation

• Efficient match summarization scheme
• 10-100x faster runtime
• Comparable accuracy to dense matches