GMS: Grid-Based Motion Statistics for Fast, Ultra-Robust Feature Correspondence

By Jiawang Bian, Wen-Yan Lin, Yasuyuki Matsushita, Sai-Kit Yeung, Tan Nguyen, Ming-Ming Cheng. Computer Vision and Pattern Recognition, 2017

628 citations recorded by the source corpus.

Incorporating smoothness constraints into feature matching is known to enable ultra-robust matching. However, such formulations are both complex and slow, making them unsuitable for video applications. This paper proposes GMS (Grid-based Motion Statistics), a simple means of encapsulating motion smoothness as the statistical likelihood of a certain number of matches in a region. GMS enables translation of high match numbers into high match quality. This provides a real-time, ultra-robust correspondence system. Evaluation on videos, with low textures, blurs and wide-baselines show GMS consistently out-performs other real-time matchers and can achieve parity with more sophisticated, much slower techniques.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/0af769f3f0e02487eecd744c7ac71ee6d6021af9

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Neighbourhood Consensus Networks (2018, neutral): “Furthermore, the neighbourhood consensus has been evaluated by manually engineered criteria, such as a certain number of locally consistent matches [33, 38, 5], or consistency in geometric parameters including distances and angles between matches [46, 34].”
ACNe: Attentive Context Normalization for Robust Permutation-Equivariant Learning (2020, background): “[55] demonstrated that they can also be highly effective for robust matching in stereo, showing a drastic leap in performance against hand-crafted methods [16, 45, 5].”
A review of multimodal image matching: Methods and applications (2021, methodology): “These relaxation rules are typically based on the assumption of locality or piecewise consistency [328], such as grid-based motion statistics (GMS) [329], locality preserving matching (LPM) [330] and its varieties [331,332], feature matching using spatial clustering with heavy outliers [333], and co”
Enforcing Geometric Constraints of Virtual Normal for Depth Prediction (2019, background): “The former ones use the assistant optical information for prediction, such as coded patterns [47], while the latter ones completely focus on image matching [2, 3].”
Image Matching from Handcrafted to Deep Features: A Survey (2020, methodology): “(7) The mismatch removal methods include: RANSAC (Fischler and Bolles 1981) (abbreviated as RS), SM (Leordeanu and Hebert 2005), ICF (Li and Hu 2010), GS (Liu and Yan 2010), LO-RANSAC (Lebeda et al. 2012) (abbreviated as LRS), VFC (Ma et al. 2014), LPM (Ma et al. 2019d), GMS (Bian et al. 2017), and ”
PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency (2021, background): “Specifically, in a 2D domain, the spatial consistency only provides a weak relation between points and epipolar lines [13, 9, 73].”
Learning Two-View Correspondences and Geometry Using Order-Aware Network (2019, methodology): “‖Ep1‖2[1] + ‖Ep1‖ 2 [2] + ‖Ep2‖ 2 [1] + ‖Ep2‖ 2 [2] ,”
Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video (2019, methodology): “Traditional methods are usually hand-crafted stage-wise systems, which rely on correspondence search [1, 2] and multi-view geometry [3, 4] for estimation.”
AiATrack: Attention in Attention for Transformer Visual Tracking (2022, methodology): “Several efforts have been made to recalibrate the raw correlations by processing them as features through hand-crafted algorithms [49,5] or learnable blocks [48,43,36,4,47,9].”