We present minimal solvers for joint homography and radial distortion estimation using affine- or orientation-covariant features. The core contribution is a set of novel constraints that exploit local feature geometry inherent to affine descriptors, enabling solvers that require fewer correspondences than point-based methods. Using modern robust estimation frameworks, we demonstrate that the proposed solvers are advantageous to the current state-of-the-art with a significant speed-up in the presence of outliers while maintaining comparable or superior estimation accuracy. The proposed solvers are tested on both synthetic and real images containing challenging fisheye distortions.
Figures & tables
Figure 5: Grossmünster Church: ( Left: ) Location of extracted SIFT correspondences (blue) of a fisheye-distorted input image. ( Right :) Corresponding 3D points (red) showing the dominant plane including the two church towers. Note that the 3D model is not notably affected by distortion artifacts, as the rectification has been done during the reconstruction phase.
Method
Hom. error
Inlier ratio
Exec. time (ms)
Median
Mean
Max
Median
Mean
Max
Median
Mean
Max
Fitzgibbon [ 15 ]
0.047
0.053
0.085
0.48
0.53
0.62
105.7
88.19
121.4
Nakano [ 30 ]
0.048
0.053
0.085
0.48
0.53
0.62
90.31
80.19
99.98
Wadenbäck et al . [ 49 ]
0.048
0.053
0.085
0.47
0.52
0.62
88.24
78.18
105.4
Proposed (ORI)
0.047
0.053
0.085
0.58
0.59
0.63
60.70
47.47
61.52
Proposed (Affine)
0.046
0.053
0.085
0.73
0.73
0.78
20.39
22.89
35.84
Table 1: Grossmünster Church: Results of the solvers in a LOMSAC framework. Best marked in bold and second best in italic .
Figure 6: Grossmünster Church: Rectified images using the estimated distortion parameter obtained by the proposed affine solver. Note that the edges of the church appear straight. This is a strong indication that the method performs well for real-life images, despite the severe fisheye distortion.
While homographies are fundamental to many computer vision tasks, the majority of conventional estimation techniques provide only point estimates without directly quantifying uncertainty introduced by noisy observations. Uncertainty, though, propagates to subsequent processing steps such as camera calibration and 3D reconstruction and is particularly relevant in safety-critical and socially relevant fields including medical imaging, autonomous driving, and defense. We present a fast Bayesian formulation for homography estimation from point correspondences that explicitly incorporates measurement uncertainty and prior knowledge while providing a posterior distribution over the homography parameters. A closed-form solution of the posterior mean of the homography is derived in homogeneous coordinates and supplemented by an iterative Bayesian approach to handle non-linearities. Synthetic experiments demonstrate the applicability to projective transformations and show improved estimation accuracy over DLT under varying noise conditions. Image stitching experiments further demonstrate applicability to real image correspondences while additionally providing uncertainty information.
Affine correspondences (ACs) enable robust and efficient solvers for multiple-view geometry by constraining the local geometry of point correspondences (PCs). Despite their utility in minimal solvers, AC estimation remains understudied. We revisit AC estimation beyond local image patches, extending it to the image and image-pair levels. Rather than estimating the components of the affine transformation separately, we learn the full 4-DoF matrix jointly. Building on this formulation, we introduce ALoMa, an AC estimator using LoMa keypoint descriptions, demonstrating consistent improvements across a wide range of downstream tasks. Code at https://github.com/davnords/loma.
David Nordström, Georg Bökman, Viktor Larsson +2
Chalmers University of Technology · University of Amsterdam · Lund University +1
Omnidirectional (360∘) cameras capture the full surrounding scene in every frame, providing a strong advantage for visual odometry (VO) over pinhole sensors with a limited field of view. However, their equirectangular projection (ERP) introduces latitude-dependent distortion, an artificial seam at the image border, and a sampling bias toward the poles. Existing 360∘ VO methods address these challenges only partially. DEFDVO closes this gap with a patch-based deep VO method that operates directly on ERP imagery, combining spherically initialized deformable feature extraction, end-to-end seam-aware estimation, and spherical-gradient patch selection. The deformable feature extractor combines spherical geometric priors with learned sampling offsets, while seam-aware processing preserves ERP continuity, and spherical-gradient selection reduces polar sampling bias. Trained exclusively on synthetic TartanAirV2 ERP imagery, DEFDVO improves the best prior average absolute trajectory error (ATE) by 4.8× (0.201 m to 0.042 m) on the TartanAirV2 evaluation set defined by 360DVO and achieves the lowest average ATE on both splits of the real-world 360DVO benchmark. Code is available at https://github.com/BurakSevsay/DEFDVO
Burak Sevsay, Egor Bondarev, Erkut Akdag
AIMS Group, Department of Electrical Engineering, Eindhoven University of Technology