MatChA: Cross-Algorithm Matching with Feature Augmentation

2025-06-27

AI Summary

Existing visual localization methods assume identical keypoint detectors across devices, rendering them ill-suited for real-world scenarios where heterogeneous detectors—each optimized for distinct descriptors—are employed. This mismatch leads to low keypoint repeatability, weak descriptor discriminability, and sharp degradation in matching performance. To address this, we propose the first cross-detector sparse feature matching framework for visual localization. Our approach introduces descriptor enhancement and a latent-space cross-algorithm translation mechanism, jointly mapping features extracted by disparate detectors into a unified, comparable latent space—thereby eliminating reliance on homologous keypoints. Evaluated on multiple benchmarks, our method significantly improves cross-detector image matching accuracy and localization robustness, demonstrating both effectiveness and strong generalization across diverse detector–descriptor combinations.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Localization, Mapping, and Navigation
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

State-of-the-art methods fail to solve visual localization in scenarios where different devices use different sparse feature extraction algorithms to obtain keypoints and their corresponding descriptors. Translating feature descriptors is enough to enable matching. However, performance is drastically reduced in cross-feature detector cases, because current solutions assume common keypoints. This means that the same detector has to be used, which is rarely the case in practice when different descriptors are used. The low repeatability of keypoints, in addition to non-discriminatory and non-distinctive descriptors, make the identification of true correspondences extremely challenging. We present the first method tackling this problem, which performs feature descriptor augmentation targeting cross-detector feature matching, and then feature translation to a latent space. We show that our method significantly improves image matching and visual localization in the cross-feature scenario and evaluate the proposed method on several benchmarks.

Problem

• Enables matching across different feature extraction algorithms
• Addresses low keypoint repeatability in cross-detector cases
• Improves visual localization with feature augmentation and translation

Innovation

• Feature descriptor augmentation for cross-detector matching
• Feature translation to a latent space
• Improved image matching in cross-feature scenarios