betaFeedback

WorksComputer Science

Correspondence-Free Camera Pose Estimation by Matching Moments

Abstract

The blind perspective-nn-point (PnP) problem is to estimate the pose of a calibrated camera from 2D and 3D point sets with unknown correspondences. We show that, for complete, noiseless observations, fourteen permutation-invariant image moments up to order four identify a fixed camera pose for almost every 3D point set containing at least twenty points. Our learning-free method, MomentPnP, matches these moments between the observed 2D points and the projected 3D points to initialize a camera pose without correspondences and refines the pose and correspondence weights using expectation–maximization. Across the test sets of three benchmarks, our method reduces median rotation error by 55–170×170\times relative to the best learned baseline, both with and without outliers. We recover accurate poses even with as many outliers as inliers. On dense ETH3D scans, pose accuracy improves as the point sets grow, even as individual correspondences become harder to resolve.

Introduction

The blind PnP problem is to estimate a calibrated camera’s pose from unordered 2D points and 3D model points, without given correspondences [54] (Figure 1). It enables localization from geometry alone when compatible visual descriptors are deliberately omitted [82] or unavailable, as for laser scans [7]. The sets must still share scene structure: inlier 2D points are projections of model points under a common camera pose, while either set may contain unmatched points. Known pairs would give the pose by PnP [44], but the number of possible pairings grows superpolynomially with the set sizes [24].

Figure 1. Blind perspective-nn-point with MomentPnP. The inputs are (a) an unordered set of 2D SIFT points [49] in a Sacre-Cœur image from the Phototourism dataset [37, 69], tracked by its structure-from-motion reconstruction [64], (b) the unordered 3D point set of those tracks, and known camera intrinsics KK. The outputs are (c) the camera pose, with the matched model points and their rays in pink, and (d) the correspondences between projected model points (blue crosses) and observed 2D points (orange).

Local methods alternate pose estimation and correspondence updates, making their success sensitive to initialization [11, 54]. Global methods search pose space by branch and bound [4, 5, 7], with computational cost driven by the regions that must be examined and excluded. Learned methods predict correspondences from coordinates, using information acquired from training data [6, 48, 82]. These limitations motivate estimating the pose from the whole point sets before resolving matches.

Because a known 3D model limits how its points can appear in an image, feature averages of the projected model constrain the pose without any pairing of points [60, 72]. Mean and covariance describe position and spread, but distinct projections can share them (Figure 2). Higher-order moments, averages of products of point coordinates [19, 34], capture shape beyond the mean and covariance [31] and can give the pose of planar point sets. Whether a fixed number of moments determines a perspective camera pose without pairing points was, to our knowledge, open.

Figure 2. Higher-order moments and Fourier features. Pairs (a, b) and (c, d) share centroid and covariance, and bars compare their moments Mg\mathcal{M}_g. Third-order moments separate (a) and (b), fourth-order moments the noisy versions (c, d). (e) An outlier at distance rr among n=100n=100 points changes fourth-order moments in proportion to r4r^4 and Fourier feature averages by at most 2/n2/n.

Contributions. We introduce MomentPnP (§ 3), a learning-free blind PnP solver that estimates the camera pose before the correspondences. It adjusts the six pose parameters until averages of bounded Fourier features over the projected model match those over the observations, after subtracting the expected effect of outliers in both sets [23]. This comparison pairs no points. On complete inputs, it matches fourteen polynomial moments, and we prove that noiseless moments determine the pose under perspective for almost every model of at least twenty points (Theorem 1). With outliers and noise, finitely many Fourier frequencies identify the pose and bound its error by the residual plus sampling and noise terms (Theorem 2).

Empirical results. On released blind PnP benchmarks, MomentPnP has a 5 to 170 times smaller median rotation error than the best learned baseline, and 82 to 97% of frames are within 1° with as many outliers as inliers (Table 1). On rendered ETH3D laser scans [65], its median rotation error falls to 0.00007° at one million points (Figure 8). In full scene reconstructions, it localizes 35 of 39 ETH3D images within 1° (Table A18).

MegaDepth [46]ModelNet40 [79]Data61 [56]
Methodno outlierswith outliersno outlierswith outliersno outlierswith outliersTime (s) ↓\downarrow
Err. ↓\downarrowAcc. ↑\uparrowErr. ↓\downarrowAcc. ↑\uparrowErr. ↓\downarrowAcc. ↑\uparrowErr. ↓\downarrowAcc. ↑\uparrowErr. ↓\downarrowAcc. ↑\uparrowErr. ↓\downarrowAcc. ↑\uparrow
Learning-free baselines
P3P-RANSAC [18]131 / 8,4820.0127 / 7,0910.0138 / 1610.0124 / 1230.0132 / 5,5360.0137 / 4,7340.030
SoftPOSIT [11]0.063 / 0.3186.09.47 / 3634.00.27 / 0.2680.030 / 140.022 / 6741.07.63 / 2937.01.0–30
GOSMA [5]135 / 3420.078 / 2970.049 / 750.043 / 560.0121 / 5280.0128 / 4940.00.25–2.9
EM-ICP (ours) [21]0.009 / 0.01199.034 / 13031.00.16 / 0.1294.028 / 3.746.00.092 / 0.2093.0148 / 4150.030
CFPE (reproduced) [60]0.123 / 0.3491.593 / 2230.00.674 / 0.1564.4129 / 100.00.100 / 0.2296.8102 / 1990.00.47–2.02
Learned baselines
Deep Blind PnP [48]0.045 / 0.1696.80.43 / 2.0859.03.35 / 0.8315.627 / 110.0130 / 3471.8138 / 4280.00.24–0.33
BPnPNet [6]0.85 / 5.4053.414 / 741.77.50 / 1.916.029 / 110.00.76 / 2.9055.416 / 710.70.16–0.22
GoMatch [82]0.14 / 0.5986.00.33 / 1.6862.3100 / 160.170 / 140.0143 / 2680.0148 / 3180.00.02–0.04
MomentPnP (ours)0.009 / 0.013100.00.022 / 0.05191.00.16 / 0.09697.30.34 / 0.2082.30.092 / 0.18100.00.09 / 0.2297.00.03–0.06

Table 1. Blind PnP on released MegaDepth, ModelNet40 and Data61 test sets. On each test set, we sample 1,000 outlier-free frames and 300 frames with outliers. Searches use a 30-second limit per frame. MomentPnP treats inputs as complete only when they are known to be outlier-free. Err.: median rotation error (∘^\circ) / translation error (%) over returned poses. Acc.: accuracy (%). Time: range of the per-dataset median times without outliers (s). EM-ICP (ours): our EM refinement (Step 3) from random poses, without moment matching. Bold: best per column.

Figure 8. Rotation error against point count. Medians over the views of three ETH3D scans [65].

ScenePointsRatioDBPBPnPGoMTuningTest
Kicker15,8972.70/30/3OOM3/33/3
Office3,4613.42/30/30/33/33/3
Facade85,0964.8OOM0/3OOM3/33/3
Terrace10,6765.30/30/3OOM3/33/3
Pipes2,4735.40/30/30/33/33/3
Relief26,5165.60/30/3OOM3/33/3
Courtyard33,4877.10/30/3OOM3/33/3
Relief 220,5149.30/30/3OOM3/33/3
Meadow2,19410.80/30/30/33/32/3
Delivery Area31,97813.20/30/3OOM3/33/3
Electro20,26814.60/30/3OOM3/31/3
Terrains18,55317.90/30/3OOM3/33/3
Playground15,66340.30/30/3OOM3/32/3
All2/390/390/3939/3935/39

Table A18. Recovery from full scene reconstructions on ETH3D. Each query contains the tracked 2D points of one image and every finite SfM model point of the scene, without the visible subset or a pose. Points is the number of model points, and Ratio is the scene extent divided by the viewing distance. Recovery columns count queries with rotation error below 1∘1^\circ. The learned baselines and Tuning use the three queries on which we chose the settings, and Test uses three held-out queries with these settings fixed. DBP, BPnP and GoM denote Deep Blind PnP, BPnPNet and GoMatch, and OOM means out of memory. Bold marks the highest recovery on the tuning queries in each scene.

Related work

Blind PnP. Methods differ in how they couple pose and correspondences. Hypothesize-and-test methods such as P3P-RANSAC [18] sample triples of candidate pairs, but correct triples become unlikely as the sets grow. Local methods such as SoftPOSIT [11] and BlindPnP [54] jointly estimate pose and soft correspondences from an initial pose or pose prior, as do recent refinements [3, 70]. Global methods such as GOPAC [7] and GOSMA [5] search pose space by branch and bound [4, 29, 77], at a cost growing with the search domain. Learned methods such as Deep Blind PnP [48] and BPnPNet [6] predict correspondences from coordinates and then solve PnP. GoMatch [82] does so for localization against retrieved views, and MinCD-PnP [1] adds color images and surface normals. All rely on correspondences, a pose prior or pose search.

Point-set registration. Registering point sets of equal dimension also couples transformations and correspondences. Softassign [20], EM-ICP [21] and coherent point drift [55] alternate between soft correspondences and a transformation, which we adopt for refinement under perspective projection. Moment-based methods avoid correspondences by comparing moments. They have been used for planar object pose [72], affine registration [31] and planar shapes under homographies [16]. Sketching methods also average sines and cosines over whole datasets [23]. Because these features are bounded, estimators that match their averages remain accurate when a fraction of the data are outliers [8]. Correspondence-free pose estimation (CFPE) [60] comes closest to our setting by matching averages of nonlinear features under perspective projection, but its planar experiments start near the true pose. We estimate a perspective camera pose from feature averages of whole point sets, without a pose prior, with outliers in both sets, before matching points.

Kernel mean embeddings. Averages of feature maps represent a distribution as a kernel mean embedding, and maximum mean discrepancy measures the distance between two embeddings [22]. For characteristic kernels, equal embeddings imply equal distributions [71]. Our feature averages are embeddings of this kind, but we need identification only among the projections of one known model. § 3.3 proves it from fourteen moments on complete inputs and from finitely many Fourier features with outliers.

Pose verification. A fitted pose can be wrong in two ways. For planar targets, two poses can give nearly the same image even with known correspondences [58, 66]. Without correspondences, a wrong pose can also place projections near many observations by chance. Desolneux et al. [14] judge a structure by how often it would arise at random, and Moisan and Stival [53] apply this test to two-view matching. We check alignment against randomly placed projections and compare each accepted pose with its half-turn alternatives (§ 3.1).

Problem statement

Input & camera. We are given nn 2D points (observations) U={ui}i=1n⊂R2\mathcal{U} = \{u_i\}_{i=1}^{n} \subset\mathbb{R}^{2}, NN model points Y={yj}j=1N⊂R3\mathcal{Y} = \{y_j\}_{j=1}^{N} \subset\mathbb{R}^{3} and the intrinsic matrix KK of the camera, with zero skew and positive focal lengths. No correspondences between the two point sets are given [6, 54]. A pose ξ=(R,t)∈SO(3)×R3\xi= (R,t) \in\mathrm{SO}(3) \times\mathbb{R}^{3} maps a model point to camera coordinates x=Ry+tx = Ry + t. If its depth x3x_3 is positive, its image proj⁡(x)\operatorname{proj}(x) is the first two entries of Kx/x3Kx/x_3 [27]. We write pj(ξ)=proj⁡(Ryj+t)p_j(\xi) = \operatorname{proj}(Ry_j + t).

Correspondences, outliers & noise. The correspondences form an unknown map σ:{1,…,n}→{1,…,N}∪{∅}\sigma: \{1,\ldots,n\} \to\{1,\ldots,N\} \cup\{\varnothing\}. We set σ(i)=j\sigma(i)=j if observation ii is the image of model point jj, and σ(i)=∅\sigma(i)=\varnothing when it is the image of no model point. Distinct observations correspond to distinct model points. The mm corresponding pairs are inliers, and the remaining n−mn-m observations and N−mN-m model points are outliers [6, 82]. The number mm is not known. Complete inputs have n=N=mn=N=m. At the true pose ξ∗\xi^*, every inlier model point has positive depth, and we model each inlier observation as

ui=pσ(i)(ξ∗)+wi,wi∼N(0,s2I2),(1)u_i = p_{\sigma(i)}(\xi^*) + w_i,\qquad w_i \sim\mathcal{N}(0,s^2 I_2), \tag*{(1)}

with i.i.d. noise wiw_i of unknown standard deviation ss pixels per coordinate. This noise also approximately absorbs model errors [27].

Task. Blind PnP is the task of estimating, from U\mathcal{U}, Y\mathcal{Y} and KK, the true pose ξ∗\xi^* and the correspondences σ\sigma of Eq. (1).

MomentPnP

In blind PnP (§ 2), MomentPnP estimates the camera pose from observations, a 3D model and the intrinsics, without correspondences. In general, both point sets contain outliers. Complete inputs, without outliers, are an exception that MomentPnP treats separately when this is known.

Preliminaries

Polynomial Moments. Moments are statistical averages of products of point coordinates, and their order gg is the number of factors [19, 34]. Averages need no pairing of points, so earlier correspondence-free methods use moments [16, 31, 72], and we use orders one to four, MgM_g. Orders three and four capture asymmetry and concentration, which separate shapes with equal centroid and covariance (Figure 2a–Figure 2d).

For an observation uu in pixels, the calibrated point ucalu_{\mathrm{cal}} consists of the first two coordinates of K−1(u⊤,1)⊤K^{-1}(u^{\top},1)^{\top}, where KK is the intrinsic matrix [27]. The affine map h(u)=(ucal−uˉcal)/rcalh(u)=(u_{\mathrm{cal}}-\bar{u}_{\mathrm{cal}})/r_{\mathrm{cal}} centers calibrated points at the mean uˉcal\bar{u}_{\mathrm{cal}} of the calibrated observations and divides by their root-mean-square radius rcalr_{\mathrm{cal}}, which keeps the coordinates well conditioned [28]. We write z=h(u)z=h(u) for the normalized point, and the same map normalizes the projection pj(ξ)p_j(\xi) of each model point at a pose ξ\xi (§ 2).

Stacking the fourteen monomials of orders one to four of a normalized point zz gives the feature vector ϕ4(z)=(z1,z2,z12,z1z2,z22,…,z24)∈R14\phi_4(z)=(z_1,z_2,z_1^2,z_1z_2,z_2^2,\ldots,z_2^4)\in\mathbb{R}^{14}. Its averages μU=1N∑iϕ4(h(ui))\mu_U=\frac{1}{N}\sum_i\phi_4(h(u_i)) over the observations and μY(ξ)=1N∑jϕ4(h(pj(ξ)))\mu_Y(\xi)=\frac{1}{N}\sum_j\phi_4(h(p_j(\xi))) over the projections are the moments of both point sets, and on complete inputs they determine the pose (Theorem 1).

Trigonometric moments. Monomial features fail with outliers, because one distant outlier can dominate the moments (Figure 2e). Like sketching methods [23], MomentPnP therefore uses weighted cosines and sines ϕF(z)=(cos⁡k⊤z,sin⁡k⊤z)k∈F\phi_F(z)=(\cos k^{\top}z,\sin k^{\top}z)_{k\in F} over a finite set FF of frequencies.1 The Fourier feature averages μU,F\mu_{U,F} and μY,F(ξ)\mu_{Y,F}(\xi) are defined like μU\mu_U and μY(ξ)\mu_Y(\xi), with ϕF\phi_F and hFh_F in place of ϕ4\phi_4 and hh. They are the trigonometric moments [51] of the point distribution. Because the features are bounded, one outlier has a bounded effect, and low frequencies describe the coarse layout as low orders do [35, 38].

As in the benchmark protocol [6], we consider image outliers are spread uniformly over the rectangle that bounds the observations, and model outliers through the box that bounds the 3D model. On average, an image outlier adds μU,0\mu_{U,0}, the average of ϕF\phi_F over the rectangle, and a model outlier adds μY,0(ξ)\mu_{Y,0}(\xi), the average of ϕF\phi_F over the projections of points spread through the box. With a fraction η\eta of outliers in each point set, μU,F\mu_{U,F} and μY,F(ξ)\mu_{Y,F}(\xi) mix the inlier averages with μU,0\mu_{U,0} and μY,0(ξ)\mu_{Y,0}(\xi). The inlier averages agree at the true pose, so the expected difference there is η(μU,0−μY,0(ξ))\eta(\mu_{U,0}-\mu_{Y,0}(\xi)).

Definition 1 (Moment matching). Moment matching minimizes

LF(ξ,η)=∥μU,F−μY,F(ξ)−η(μU,0−μY,0(ξ))∥2(2)\mathcal{L}_F(\xi,\eta)=\lVert\mu_{U,F}-\mu_{Y,F}(\xi)-\eta(\mu_{U,0}-\mu_{Y,0}(\xi))\rVert^2 \tag*{(2)}

over the pose ξ\xi and η\eta. The objective is quadratic in the unknown η\eta, so the best η\eta at each pose has a closed form. On complete inputs, η=0\eta=0 and no outlier can dominate the moments, so we replace the Fourier feature averages, and we minimize L4(ξ)=∥μU−μY(ξ)∥2\mathcal{L}_4(\xi)=\lVert\mu_U-\mu_Y(\xi)\rVert^2 (Theorem 1).

Expectation–maximization (EM). EM [13, 21, 55] then refines the pose with every observation, each explained as a uniform outlier over the rectangle or a noisy projection of a model point in front of the camera (Figure 3c). A refined pose is correct only if the observation lie much closer to their nearest projections than to the next, which we test against randomly placed projections [14, 53].

Figure 3. Overview of MomentPnP. (a) Cameras are searched over centers and orientations, and observed features averaged once. (b) Moment matching refines the best cameras, raising the cutoff from 4 to 20. (c) EM refines fits with soft correspondences. (d) The alignment check accepts poses whose observations lie much closer to their nearest projections than to the next, and the most likely accepted pose or alternative is returned. Orange: observations. Blue: model points and projections. Gray: unmatched points, unselected poses.

References

  1. [1]Pei An, Jiaqi Yang, Muyao Peng, You Yang, Qiong Liu, Xiaolin Wu, and Liangliang Nan. MinCD-PnP: Learning 2D-3D correspondences with approximate blind PnP. In 2025 IEEE/CVF International Conference on Computer Vision (ICCV), pages 26519–26528, 2025. https://doi.org/10.1109/ICCV51070.2025.02461. 2
  2. [2]Jacek Bochnak, Michel Coste, and Marie-Françoise Roy. Real algebraic geometry. Springer, Berlin, Heidelberg, 1998. https://doi.org/10.1007/978-3-662-03718-8. 17, 21
  3. [3]Younes Boutiyarnzist, Samy Labsir, Jean-Yves Tourneter, François Vincent, and Philippe Salmon. A new EM algorithm for 2D-3D point cloud registration with probabilistic data association. In 2025 IEEE Statistical Signal Processing Workshop (SSP), pages 11–15, 2025. https://doi.org/10.1109/SSP64130.2025.11073421. 2
  4. [4]Mark Brown, David Windridge, and Jean-Yves Guillau­mat. Globally optimal 2D-3D registration from points or lines without correspondences. In 2015 IEEE International Conference on Computer Vision (ICCV), pages 2111–2119, 2015. https://doi.org/10.1109/ICCV.2015.244. 1, 2
  5. [5]Dylan Campbell, Lars Pettersson, Laurent Kneip, Hongdong Li, and Stephen Gould. The alignment of the spheres: Globally-optimal spherical mixture alignment for camera pose estimation. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11788–11798, 2019. https://doi.org/10.1109/CVPR.2019.01207. 1, 2, 4, 6, 37, 51
  6. [6]Dylan Campbell, Liu Liu, and Stephen Gould. Solving the blind perspective-n-point problem end-to-end with robust differentiable geometric optimization. In Computer Vision – ECCV 2020, pages 244–261. Springer International Publishing, 2020. https://doi.org/10.1007/978-3-030-58536-5_15. 1, 2, 3, 5, 6, 47, 48, 49
  7. [7]Dylan Campbell, Lars Pettersson, Laurent Kneip, and Hongdong Li. Globally-optimal inlier set maximisation for camera pose and correspondence estimation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(2):328–342, 2020. https://doi.org/10.1109/TPAMI.2018.2848650. 1, 2, 51
  8. [8]Badr-Eddine Chérief-Abdellatif and Pierre Alquier. Finite sample properties of parametric MMD estimation: Robustness to misspecification and dependence. Bernoulli, 28(1):181–213, 2022. https://doi.org/10.3150/21-BEJ1338. 2
  9. [9]Sung Nok Chiu, Dietrich Stoyan, Wilfrid S. Kendall, and Joseph Mecke. Stochastic Geometry and Its Applications. Wiley, 3rd edition, 2013. https://doi.org/10.1002/9781118658222. 3, 4
  10. [10]Lénaïc Chizat, Gabriel Peyré, Bernhard Schmitzer, and François-Xavier Vialard. Scaling algorithms for unbalanced optimal transport problems. Mathematics of Computation, 87(314):2563–2609, 2018. https://doi.org/10.1090/mcom/3303. 8
  11. [11]Philip David, Daniel DeMenthon, Ramani Duraiswami, and Hanan Samet. SoftPOSIT: Simultaneous pose and correspondence determination. International Journal of Computer Vision, 59(3):259–284, 2004. https://doi.org/10.1023/B:VISI.000002580.010423.1f. 1, 2, 6
  12. [12]Daniel F. DeMenthon and Larry S. Davis. Model-based object pose in 25 lines of code. International Journal of Computer Vision, 15(1–2):123–141, 1995. https://doi.org/10.1007/BF01450852. 5
  13. [13]A. P. Dempster, N. M. Laird, and D. B. Rubin. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B (Methodological), 39(1):1–22, 1977. https://doi.org/10.1111/j.2517-6161.1977.tb01600.x. 3
  14. [14]Agnès Desolneux, Lionel Moisan, and Jean-Michel Morel. Meaningful alignments. International Journal of Computer Vision, 40(1):7–23, 2000. https://doi.org/10.1023/A:1026593302236. 2, 3
  15. [15]Yaqing Ding, Viktor Kocur, Václav Vávra, Zuzana Berger Ha­ladová, Jian Yang, Torsten Sattler, and Zuzana Kukelova. RePoseD: Efficient relative pose estimation with known depth information. In 2025 IEEE/CVF International Conference on Computer Vision (ICCV), pages 14876–14886, 2025. https://doi.org/10.1109/ICCV51070.2025.021380. 8
  16. [16]Csaba Domokos, József Németh, and Zoltan Kato. Nonlinear shape registration without correspondences. IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(5):943–958, 2012. https://doi.org/10.1109/TPAMI.2011.200. 2, 3
  17. [17]Matthias Faessler, Elias Mueggler, Karl Schwabe, and Davide Scaramuzza. A monocular pose estimation system based on infrared LEDs. In IEEE International Conference on Robotics and Automation (ICRA), pages 907–913, 2014. https://doi.org/10.1109/icra.2014.6906962. 65
  18. [18]Martin A. Fischler and Robert C. Bolles. Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography. Communications of the ACM, 24(6):381–395, 1981. https://doi.org/10.1145/358669.358692. 2, 6
  19. [19]Jan Flusser, Tomáš Suk, and Barbara Zitová. 2D and 3D Image Analysis by Moments. Wiley, 2016. https://doi.org/10.1002/9781118940422. 1, 3
  20. [20]Steven Gold, Anand Rangarajan, Chien-Ping Lu, Suguna Pappu, and Eric Mjolsness. New algorithms for 2D and 3D point matching: Pose estimation and correspondence. Pattern Recognition, 31(8):1019–1031, 1998. https://doi.org/10.1016/S0031-3203(98)80010-1. 2
  21. [21]Sébastien Granger and Xavier Pennec. Multi-scale EM-ICP: A fast and robust approach for surface registration. In Computer Vision – ECCV 2002, pages 418–432. Springer Berlin Heidelberg, 2002. https://doi.org/10.1007/3-540-47979-1_28. 2, 3, 4, 6
  22. [22]Arthur Gretton, Karsten M. Borgwardt, Malte J. Rasch, Bernhard Schölkopf, and Alexander Smola. A kernel two-sample test. Journal of Machine Learning Research, 13(25):723–773, 2012. https://www.jmlr.org/papers/v13/gretton12a.html. 2DOI
  23. [23]Rémi Gribonval, Antoine Chatalic, Nicolas Keriven, Vincent Schellekens, Laurent Jacques, and Philip Schniter. Sketching data sets for large-scale learning: Keeping only what you need. IEEE Signal Processing Magazine, 38(5):12–36, 2021. https://doi.org/10.1109/MSP.2021.3092574. 1, 2, 3
  24. [24]W. Eric L. Grimson and Tomás Lozano-Pérez. Localizing overlapping parts by searching the interpretation tree. IEEE Transactions on Pattern Analysis and Machine Intelligence, 9(4):469–482, 1987. https://doi.org/10.1109/TPAMI.1987.4767935. 1
  25. [25]Rafael Grompone von Gioi, Jérémie Jakubowicz, Jean-Michel Morel, and Gregory Randall. LSD: A line segment detector. Image Processing On Line, 2:35–55, 2012. https://doi.org/10.5201/ipol.2012.2.1. 64
  26. [26]Fredric J. Harris. On the use of windows for harmonic analysis with the discrete Fourier transform. Proceedings of the IEEE, 66(1):51–83, 1978. https://doi.org/10.1109/PROC.1978.10837. 4
  27. [27]Richard Hartley and Andrew Zisserman. Multiple view geometry in computer vision. Cambridge University Press, 2nd edition, 2004. https://doi.org/10.1017/CBO9780511811685. 2, 3, 40
  28. [28]Richard I. Hartley. In defense of the eight-point algorithm. IEEE Transactions on Pattern Analysis and Machine Intelligence, 19(6):580–593, 1997. https://doi.org/10.1109/34.601246. 3
  29. [29]Richard I. Hartley and Fredrik Kahl. Global optimization through rotation space search. International Journal of Computer Vision, 82(1):64–79, 2009. https://doi.org/10.1007/s11263-008-0186-9. 2
  30. [30]Stefan Hinterstoisser, Vincent Lepetit, Slobodan Ilic, Stefan Holzer, Gary Bradski, Kurt Konolige, and Nassir Navab. Model based training, detection and pose estimation of texture-less 3D objects in heavily cluttered scenes. In Computer Vision – ACCV 2012, pages 548–562. Springer Berlin Heidelberg, 2013. https://doi.org/10.1007/978-3-642-37331-2_42. 48
  31. [31]Jeffrey Ho, Adrian Peter, Anand Rangarajan, and Ming-Hsuan Yang. An algebraic approach to affine registration of point sets. In 2009 IEEE 12th International Conference on Computer Vision, pages 1335–1340, 2009. https://doi.org/10.1109/ICCV.2009.5459309. 1, 2, 3
  32. [32]Tomáš Hodaň, Pavel Haluza, Štěpán Obdřžálek, Jiří Matas, Manolis Lourakis, and Xenophon Zabulis. T-LESS: An RGB-D dataset for 6D pose estimation of texture-less objects. In 2017 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 880–888, 2017. https://doi.org/10.1109/WACV.2017.103. 64
  33. [33]Wassily Hoeffding. Probability inequalities for sums of bounded random variables. Journal of the American Statistical Association, 58(301):13–30, 1963. https://doi.org/10.1080/01621459.1963.10500830. 27
  34. [34]Ming-Kuei Hu. Visual pattern recognition by moment invariants. IRE Transactions on Information Theory, 8(2):179–187, 1962. https://doi.org/10.1109/TIT.1962.1057692. 1, 3
  35. [35]Peter J. Huber and Elvezio M. Ronchetti. Robust Statistics. Wiley, Hoboken, NJ, 2 edition, 2009. https://doi.org/10.1002/9780470434569. 3
  36. [36]Aapo Hyvärinen and Erkki Oja. Independent component analysis: Algorithms and applications. Neural Networks, 13(4–5):411–430, 2000. https://doi.org/10.1016/S0893-6080(00)00026-5. 18
  37. [37]Yuhe Jin, Dmytro Mishkin, Anastasiia Mishchuk, Jiri Matas, Pascal Fua, Kwang Moo Yi, and Eduard Trulls. Image matching across wide baselines: From paper to practice. International Journal of Computer Vision, 129(2):517–547, 2021. https://doi.org/10.1007/s11263-020-01385-0. 1, 6, 7, 53, 57
  38. [38]Yitzhak Katznelson. An Introduction to Harmonic Analysis. Cambridge University Press, 3 edition, 2004. https://doi.org/10.1017/CBO9781139165372. 3, 5, 26, 27
  39. [39]Tong Ke and Stergios I. Roumeliotis. An efficient algebraic solution to the perspective-three-point problem. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 4618–4626, 2017. https://doi.org/10.1109/CVPR.2017.491. 38
  40. [40]Alex Kendall, Matthew Grimes, and Roberto Cipolla. PoseNet: A convolutional network for real-time 6-DOF camera relocalization. In 2015 IEEE International Conference on Computer Vision (ICCV), pages 2938–2946, 2015. https://doi.org/10.1109/ICCV.2015.336. 6, 7, 53, 57
  41. [41]Dmitry Kunisky and Jonathan Niles-Weed. Strong recovery of geometric planted matchings. In Proceedings of the 2022 Annual ACM-SIAM Symposium on Discrete Algorithms (SODA), pages 834–876. Society for Industrial and Applied Mathematics, 2022. https://doi.org/10.1137/1.9781611977073.36. 30
  42. [42]B. Laurent and P. Massart. Adaptive estimation of a quadratic functional by model selection. The Annals of Statistics, 28(5):1302–1338, 2000. https://doi.org/10.1214/aos/1015957395. 35
  43. [43]John M. Lee. Introduction to smooth manifolds. Springer New York, 2nd edition, 2012. https://doi.org/10.1007/978-1-4419-9982-5. 23
  44. [44]Vincent Lepetit, Francesc Moreno-Noguer, and Pascal Fua. EPnP: An accurate O(n) solution to the PnP problem. International Journal of Computer Vision, 81(2):155–166, 2009. https://doi.org/10.1007/s11263-008-0152-6. 1, 5, 65
  45. [45]Kenneth Levenberg. A method for the solution of certain non-linear problems in least squares. Quarterly of Applied Mathematics, 2(2):164–168, 1944. https://doi.org/10.1090/qam/10666. 4
  46. [46]Zhengqi Li and Noah Snavely. MegaDepth: Learning single-view depth prediction from Internet photos. In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2041–2050, 2018. https://doi.org/10.1109/CVPR.2018.00218. 5, 6, 47, 48, 49, 50, 51, 52, 54, 62
  47. [47]Hao Liang, Jingyu Lu, Manolis C. Tsakiris, and Lihong Zhi. A field-theoretic view of unlabeled sensing. Journal of Symbolic Computation, 135:102542, 2026. https://doi.org/10.1016/j.jsc.2025.102542. 30
  48. [48]Liu Liu, Dylan Campbell, Hongdong Li, Dingfu Zhou, Xibin Song, and Ruigang Yang. Learning 2D-3D correspondences to solve the blind perspective-n-point problem, 2020. https://doi.org/10.48550/arXiv.2003.06752. 1, 2, 6, 48, 49
  49. [49]David G. Lowe. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 60(2):91–110, 2004. https://doi.org/10.1023/B:VISI.0000029664.99615.94. 1, 8, 63
  50. [50]Leon Lufkin, Yihong Wu, and Jiaming Xu. Sharp information-theoretic thresholds for shuffled linear regression. In 2024 IEEE International Symposium on Information Theory (ISIT), pages 363–367, 2024. https://doi.org/10.1109/ISIT57864.2024.10619515. 30
  51. [51]Kanti V. Mardia and Peter E. Jupp. Directional Statistics. Wiley, 1999. https://doi.org/10.1002/9780470316979. 3
  52. [52]Donald W. Marquardt. An algorithm for least-squares estimation of nonlinear parameters. Journal of the Society for Industrial and Applied Mathematics, 11(2):431–441, 1963. https://doi.org/10.1137/0111030. 4
  53. [53]Lionel Moisan and Bérenger Stival. A probabilistic criterion to detect rigid point matches between two images and estimate the fundamental matrix. International Journal of Computer Vision, 57(3):201–218, 2004. https://doi.org/10.1023/B:VISI.0000013094.38752.54. 2, 3
  54. [54]Francesc Moreno-Noguer, Vincent Lepetit, and Pascal Fua. Pose priors for simultaneously solving alignment and correspondence. In Computer Vision – ECCV 2008, pages 405–418. Springer Berlin Heidelberg, 2008. https://doi.org/10.1007/978-3-540-88688-4_30. 1, 2
  55. [55]Andriy Myronenko and Xubo Song. Point set registration: Coherent point drift. IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(12):2262–2275, 2010. https://doi.org/10.1109/TPAMI.2010.46. 2, 3, 4
  56. [56]Sarah Taghavi Namin, Mohammad Najafi, Mathieu Salzmann, and Lars Petersson. A multi-modal graphical model for scene analysis. In 2015 IEEE Winter Conference on Applications of Computer Vision, pages 1006–1013, 2015. https://doi.org/10.1109/WACV.2015.139. 5, 6, 47, 48, 49, 50, 51, 52, 54, 62
  57. [57]Jorge Nocedal and Stephen J. Wright. Numerical optimization. Springer, 2nd edition, 2006. https://doi.org/10.1007/978-0-387-40065-5. 4, 28, 38, 40
  58. [58]Denis Oberkampf, Daniel F. DeMenthon, and Larry S. Davis. Iterative pose estimation using coplanar feature points. Computer Vision and Image Understanding, 63(3):495–511, 1996. https://doi.org/10.1006/cviu.1996.0037. 2, 5
  59. [59]Edwin Olson. AprilTag: A robust and flexible visual fiducial system. In 2011 IEEE International Conference on Robotics and Automation, pages 3400–3407, 2011. https://doi.org/10.1109/ICRA.2011.5979561. 65
  60. [60]Quan Quan and Dun Dai. Correspondence-free pose estimation with patterns: A unified approach for multi-dimensional vision. In 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 20203–20209, 2025. https://doi.org/10.1109/IROS60139.2025.11246324. 1, 2, 5, 6
  61. [61]Inder K. Rana. An introduction to measure and integration. American Mathematical Society, 2nd edition, 2002. https://doi.org/10.1090/gsm/045. 33, 36
  62. [62]Paul-Edouard Sarlin, Ajaykumar Unagar, Måns Larsson, Hugo Germain, Carl Toft, Viktor Larsson, Marc Pollefeys, Vincent Lepetit, Lars Hammarstrand, Fredrik Kahl, and Torsten Sattler. Back to the feature: Learning robust camera localization from pixels to pose. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3246–3256, 2021. https://doi.org/10.1109/CVPR46437.2021.00326. 6, 7, 53
  63. [63]Torsten Sattler, Qunjie Zhou, Marc Pollefeys, and Laura Leal-Taixé. Understanding the limitations of CNN-based absolute camera pose regression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3297–3307, 2019. https://doi.org/10.1109/CVPR.2019.00342. 8
  64. [64]Johannes L. Schönberger and Jan-Michael Frahm. Structure-from-motion revisited. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 4104–4113, 2016. https://doi.org/10.1109/CVPR.2016.445. 1, 8, 53, 59, 64
  65. [65]Thomas Schöps, Johannes L. Schönberger, Silvano Galliani, Torsten Sattler, Konrad Schindler, Marc Pollefeys, and Andreas Geiger. A multi-view stereo benchmark with high-resolution images and multi-camera videos. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 2538–2547, 2017. https://doi.org/10.1109/CVPR.2017.272. 2, 6, 7, 53, 57, 59
  66. [66]Gerald Schweighofer and Axel Pinz. Robust pose estimation from a planar target. IEEE Transactions on Pattern Analysis and Machine Intelligence, 28(12):2024–2030, 2006. https://doi.org/10.1109/TPAMI.2006.252. 2, 5
  67. [67]Lucas R. Schwengber and Roberto I. Oliveira. Geometric planted matchings beyond the Gaussian model. Random Structures & Algorithms, 68(4):e70080, 2026. https://doi.org/10.1002/rsa.70080. 30
  68. [68]Jamie Shotton, Ben Glocker, Christopher Zach, Shahram Izadi, Antonio Criminisi, and Andrew Fitzgibbon. Scene coordinate regression forests for camera relocalization in RGB-D images. In 2013 IEEE Conference on Computer Vision and Pattern Recognition, pages 2930–2937, 2013. https://doi.org/10.1109/CVPR.2013.377. 7, 53, 57
  69. [69]Noah Snavely, Steven M. Seitz, and Richard Szeliski. Photo tourism: Exploring photo collections in 3D. ACM Transactions on Graphics, 25(3):835–846, 2006. https://doi.org/10.1145/1141911.1141964. 1, 7, 57
  70. [70]Jingwei Song and Maani Ghaffari. DynaWeightPnP: Toward global real-time 3D–2D solver in PnP without correspondences. Robotics and Autonomous Systems, 207:105764, 2027. https://doi.org/10.1016/j.robot.2026.105764. 2
  71. [71]Bharath K. Sriperumbudur, Arthur Gretton, Kenji Fukumizu, Bernhard Schölkopf, and Gert R. G. Lanckriet. Hilbert space embeddings and metrics on probability measures. Journal of Machine Learning Research, 11(50):1517–1561, 2010. https://www.jmlr.org/papers/v11/sriperumbudur10a.html. 2arxiv.org/abs/0907.5309
  72. [72]Omar Tahri and François Chaumette. Complex objects pose estimation based on image moment invariants. In Proceedings of the 2005 IEEE International Conference on Robotics and Automation, pages 436–441, 2005. https://doi.org/10.1109/ROBOT.2005.1570157. 1, 2, 3
  73. [73]Antonio Torralba and Alexei A. Efros. Unbiased look at dataset bias. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1521–1528, 2011. https://doi.org/10.1109/CVPR.2011.5995347. 8
  74. [74]Manolis C. Tsakiris, Liangzu Peng, Aldo Conca, Laurent Kneip, Yuanning Shi, and Hayoung Choi. An algebraic-geometric approach for linear regression without correspondences. IEEE Transactions on Information Theory, 66(8):5130–5144, 2020. https://doi.org/10.1109/TIT.2020.2977166. 30
  75. [75]Jayakrishnan Unnikrishnan, Saeid Haghighatshoar, and Martin Vetterli. Unlabeled sensing with random linear measurements. IEEE Transactions on Information Theory, 64(5):3237–3253, 2018. https://doi.org/10.1109/TIT.2018.2809002. 30
  76. [76]Roman Vershynin. High-dimensional probability: An introduction with applications in data science. Cambridge University Press, 2018. https://doi.org/10.1017/9781108231596. 33, 34
  77. [77]Chen Wang, Yinlong Liu, Yiru Wang, Xuechen Li, and Manning Wang. Efficient and outlier-robust simultaneous pose and correspondence determination by branch-and-bound and transformation decomposition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10):6924–6938, 2022. https://doi.org/10.1109/TPAMI.2021.3096842. 2
  78. [78]Changchang Wu. Towards linear-time incremental structure from motion. In 2013 International Conference on 3D Vision (3DV), pages 127–134, 2013. https://doi.org/10.1109/3DV.2013.25. 53
  79. [79]Zhirong Wu, Shuran Song, Aditya Khosla, Fisher Yu, Linguang Zhang, Xiaou Tang, and Jianxiong Xiao. 3D ShapeNets: A deep representation for volumetric shapes. In 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1912–1920, 2015. https://doi.org/10.1109/CVPR.2015.7298801. 5, 6, 47, 48, 49, 50, 51, 52, 54, 62
  80. [80]Yifan Yu, Shaohui Liu, Rémi Pautrat, Marc Pollefeys, and Viktor Larsson. Relative pose estimation through affine corrections of monocular depth priors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16706–16716, 2025. https://doi.org/10.1109/CVPR52734.2025.01557. 8
  81. [81]Hang Zhang, Martin Slawski, and Ping Li. The benefits of diversity: Permutation recovery in unlabeled sensing from multiple measurement vectors. IEEE Transactions on Information Theory, 68(4):2509–2529, 2022. https://doi.org/10.1109/TIT.2021.3127072. 30
  82. [82]Qunjie Zhou, Sérgio Agostinho, Aljoša Ošep, and Laura Leal-Taixé. Is geometry enough for matching in visual localization? In Computer Vision – ECCV 2022, pages 407–425. Springer Nature Switzerland, 2022. https://doi.org/10.1007/978-3-031-20080-9_24. 1, 2, 6, 48, 49

Paper details

vlnc:W_gm7cckce
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Codex

View statement
Looking for related papers…
Contents