Abstract
Wide-baseline image registration under out-of-plane rotation and larger viewpoint change is still challenging. Most of the commonly used matching algorithms are not invariant to affine transformation. They heavily rely on the local features of image patches and ignore global information, the mismatch is inevitable and greatly affect the accuracy of image registration. To address this issue, we propose a feature point matching pair filter based on global spatial position correspondences of feature points, coined Descriptor Net Filter (DNF). We put forward two criteria to evaluate matching quality. One is the local matching quality computed by independent local feature, the other is the global matching quality relying on geometric network constraint. Combining the advantages of both local feature and large-scale geometric constraint, our method removes mismatches effectively. The experiments on both planar scenes and 3D objects from several standard datasets show that the DNF significantly enhances the matching precision and retains more correct matches as well.
Highlights
Accurate image registration is a fundamental preprocessing step in computer vision applications, which provides an important technique for manufacturing [1], camera calibration [2], 3D reconstruction [3] and visual SLAM [4]
For the Descriptor Net Filter (DNF) algorithm, we evaluate the validity of matches with both the local and the global matching quality criteria
In accordance with the local matching quality, reliable matches are selected from the initial matching set as the basic elements of the topology network
Summary
Accurate image registration is a fundamental preprocessing step in computer vision applications, which provides an important technique for manufacturing [1], camera calibration [2], 3D reconstruction [3] and visual SLAM [4]. To improve the invariance and repeatability, the line descriptors [10], [11] are constructed by keypoint correspondences, which are based on well-developed point feature. To address the reliability issue of wide-baseline feature matching, we propose a feature point matching pair filter for image registration, termed DNF (Descriptor Net Filter). F. Ye et al.: DNF: Feature Point Matching Pairs Filter Based on Descriptor Net of keypoints and build up the topological constraint to filter the correct matches. The process of D-Nets includes steps of: keypoints detection, d-tokens (descriptor tokens) construction, hash and vote for matching, ranking matches It shows a large advantage in terms of precision and recall, but computes with a marked increase of time. Through the experiment on planar scene and 3D object dataset, it is proved that our proposed method enables SIFT and ASIFT (Affine-SIFT) to enhance the affine invariance and preserves more correct matches than ratio-test [5], LRC (Left-Right Consistency) [28] and RANSAC (Random Sample Consensus) [29]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
