Guided neighborhood affine subspace embedding for feature matching
作者：
Highlights：
• We present a general yet powerful robust feature matching algorithm.
• The intrinsic manifold with affine subspace approximation is proposed for feature matching.
• A density-based seed point selection strategy is designed for neighborhood refinement.
• A multi-scale neighborhood is constructed to deal with various kinds of degradation.
• A closed-form solution of our model is derived with O(NlogN) time complexity.
摘要
•We present a general yet powerful robust feature matching algorithm.•The intrinsic manifold with affine subspace approximation is proposed for feature matching.•A density-based seed point selection strategy is designed for neighborhood refinement.•A multi-scale neighborhood is constructed to deal with various kinds of degradation.•A closed-form solution of our model is derived with O(NlogN) time complexity.
论文关键词：Feature matching,Image correspondence,Neighborhood affine subspace,Multi-scale,Outlier,Mismatch removal
论文评审过程：Received 8 June 2021, Revised 27 November 2021, Accepted 4 December 2021, Available online 7 December 2021, Version of Record 13 December 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108489
