Computer Science > Computer Vision and Pattern Recognition
[Submitted on 29 Apr 2023 (v1), revised 2 May 2023 (this version, v2), latest version 2 May 2024 (v5)]
Title:Searching from Area to Point: A Hierarchical Framework for Semantic-Geometric Combined Feature Matching
View PDFAbstract:Feature matching is a crucial technique in computer vision. Essentially, it can be considered as a searching problem to establish correspondences between images. The key challenge in this task lies in the lack of a well-defined search space, leading to inaccurate point matching of current methods. In pursuit of a reasonable matching search space, this paper introduces a hierarchical feature matching framework: Area to Point Matching (A2PM), to first find semantic area matches between images, and then perform point matching on area matches, thus setting the search space as the area matches with salient features to achieve high matching precision. This proper search space of A2PM framework also alleviates the accuracy limitation in state-of-the-art Transformer-based matching methods. To realize this framework, we further propose Semantic and Geometry Area Matching (SGAM) method, which utilizes semantic prior and geometry consistency to establish accurate area matches between images. By integrating SGAM with off-the-shelf Transformer-based matchers, our feature matching methods, adopting the A2PM framework, achieve encouraging precision improvements in massive point matching and pose estimation experiments for present arts.
Submission history
From: Yesheng Zhang [view email][v1] Sat, 29 Apr 2023 08:16:12 UTC (7,243 KB)
[v2] Tue, 2 May 2023 11:49:26 UTC (7,244 KB)
[v3] Fri, 5 May 2023 09:04:12 UTC (7,244 KB)
[v4] Sun, 2 Jul 2023 03:11:26 UTC (10,309 KB)
[v5] Thu, 2 May 2024 03:19:33 UTC (22,161 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.