Pith. sign in

REVIEW 5 minor 64 references

Numerically Computing Galois Groups of Minimal Problems

T0 review · 0 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The geometric Galois group of a minimal problem is its true complexity measure, and numerical monodromy can compute it well enough to build optimal solvers.

desk verdict A clear, honest tutorial that makes a good case for Galois groups as a lens on minimal problems; the five-point code is reproducible, but the RANSAC formula has a sign error and the monodromy-based widths are lower bounds unless certified. read the letter →

arxiv 2507.10407 v1 pith:WI55V77R submitted 2025-07-14 cs.CV cs.SCmath.AG

Timothy Duff This is my paper · ORCID
classification cs.CVcs.SCmath.AG MSC 14Q2012F1065H1068T45
keywords Galoisgroupmonodromyminimalproblemsnumericalalgebraicgeometryhomotopycontinuationwidthcomputervisionbranchedcovers
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This tutorial argues that the geometric Galois group of a minimal problem—a parametric family of polynomial equations with finitely many solutions—is the overarching invariant of how hard that family is to solve. It reports a numerical recipe: sample monodromy loops by parameter homotopy, collect the induced permutations of the solution set, and read off the group's “Galois width,” a minimax measure of the most expensive algebraic step needed to reach the solutions. If this recipe is right, the Galois width tells you the optimal algebraic solving strategy: the five-point relative pose problem has width 10, meaning tracking 10 homotopy paths suffices to obtain all 20 solutions, matching and explaining why the classical algorithm is algebraically optimal. The paper also surveys cases where Galois groups predicted decompositions that led to faster solvers, including a radial-distortion reconstruction problem whose naive degree 3584 collapses to a 28-path solver.

What carries the argument

The machinery is the numerical monodromy action: starting from a fabricated generic problem-solution pair $(z_0, x_0)$, one builds two parameter homotopies $z_0 \to z_1$ and $z_1 \to z_0$, tracks $x_0$ along them, and records the permutation of the $d$ solutions of the fiber; many such loops generate a permutation group modelling $G$. The Galois width is the minimax index of an unrefinable subgroup chain $\mathrm{id} = H_m \le \cdots \le H_0 = G$, i.e. the smallest possible largest step cost in any exact computation of a solution, and a theorem gives its values for symmetric, alternating, cyclic, and simple groups. An equivalence action on the solution set (for five-point, the 20 solutions paired into 10 essential matrices) reduces the number of tracked paths.

What would settle it

Run the same monodromy computation with certified path tracking on the five-point problem; if the certified permutation group is strictly larger than the computed $(C_2)^9 \rtimes S_{10}$, or if a certified 10-path homotopy fails to reach all 20 solutions for a generic real instance, then the claim that the numerical Galois width predicts the optimal solver is refuted.

Watch

Extended reading notes

Core claim

The paper's central claim is that every minimal problem is a branched cover $X \to \mathbb{C}^m$, and the Galois group $G = \operatorname{Gal}(K/F)$ of the corresponding function-field extension is the intrinsic complexity measure of the problem: its degree is the number of complex solutions, its deck transformations are the problem's symmetries, its imprimitivity detects decomposability, and its Galois width is the minimax cost of the cheapest tower of field extensions needed to compute any solution. The paper further claims that this group can be computed numerically: tracking parameter homotopies along loops in the problem space produces permutations of the fiber, and the generated group exposes the structure. For the five-point problem the computed group is $(C_2)^9 \rtimes S_{10}$ acting on 20 solutions, with Galois width 10, and the paper demonstrates a parameter homotopy tracking 10 paths whose outputs, augmented by the twisted-pair symmetry, recover all 20 solutions.

Load-bearing premise

The method assumes that the random loops it tracks produce enough of the true solution-symmetry group to reveal its structure; missed generators or silently wrong paths would break the guarantee that the computed path count is optimal.

Editorial extensions

If this is right

  • For the five-point relative pose problem, Galois width 10 means the optimal solver tracks 10 paths to recover 20 solutions; the Nistér–Stewénius algorithm achieves this optimum.
  • P3P has Galois width 3, concretely explaining the classical cubic-plus-quadratics solution; P3L has width 8, so no such low-degree reduction is possible for the pure-line case.
  • For radial camera reconstruction, Galois structure reduces the naive 3584 solutions to 28 tracked paths via quotient layers ($S_3$-orbits, uncalibrated reconstructions, radial quadrifocal tensors).
  • Galois groups expose deck symmetries (e.g., the twisted-pair symmetry of the five-point problem) that let a solver recover the remaining solutions from half the paths.
  • Decomposability of a minimal problem is equivalent to imprimitivity of its Galois action, so group computation detects when a problem splits into subproblems.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the numerically generated group is a subgroup of the true Galois group, the computed Galois width is a lower bound on the true width; a solver is certified only when this bound is known to saturate, for instance through a known quotient structure.
  • Swapping the heuristic path tracker for a certified one would turn the five-point computation into a proof, and the same recipe could certify widths for other moderate-size minimal problems.
  • The Galois-width viewpoint suggests a practical ranking of minimal problems by algebraic difficulty, which could guide which solver a RANSAC loop should call, though the paper does not test this directly.
  • A natural testbed is the atlas of pinhole-camera minimal problems: compute monodromy groups for every entry and compare predicted path counts with actual homotopy track counts.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

0 major / 5 minor

Summary. This invited ISSAC tutorial argues that the Galois group of a minimal problem is the key intrinsic invariant of solver difficulty, and that its Galois width predicts the optimal algebraic solving strategy. It surveys minimal problems from algebraic vision (P3P, point-line resectioning, five-point relative pose, radial four-view reconstruction), defines the Galois group and Galois width, reviews known width values, and describes numerical monodromy as a computational tool. The paper includes a reproducible Macaulay2/GAP example for the five-point problem in which monodromy produces the group (C2)^9 : S10 acting on the 20 solutions, the Galois width is verified to be 10 via the 10 two-point blocks, and a parameter homotopy tracking 10 paths recovers all 20 solutions with residuals on the order of 1e-12. The paper is explicit that the numerical calculation is heuristic and produces at most a subgroup of the true Galois group.

Significance. If the numerical heuristic is trusted or later certified, the tutorial makes a strong case that Galois width is a practically meaningful measure of solver complexity, and the five-point example concretely demonstrates the path-count reduction from 20 to 10. The paper's strengths are its transparent treatment of the subgroup caveat, reproducible code, and the independent block argument that certifies the five-point width. The Galois-width theorems and most width values are cited from prior work rather than proved here, which is appropriate for a tutorial but means the reader must look to the cited literature for full justification. The tutorial should be useful to both the symbolic-computation and computer-vision communities.

minor comments (5)
  1. [Section 2, Eq. (4)] With p defined as the fraction of erroneous correspondences, the all-inlier probability should be binom((1-p)n,k)/binom(n,k), not binom(pn,k)/binom(n,k). The two expressions coincide at p=0.5, so Figure 2 is not numerically affected, but the displayed formula is incorrect for general p and should be fixed.
  2. [Theorem 3.1(6)] The statement gw(S_n)=gw(A_n)=n for all n≥1 fails for n=2, since A_2 is trivial and hence has width 1 under the convention used in the GAP code. Please either restrict the statement to n≥3 or spell out the trivial-group convention.
  3. [Section 3, Example 3.2 / Section 5] The width 28 quoted for radial four-view reconstruction is presented as a fact in Example 3.2, but the only evidence described is numerical monodromy, and Section 5 states that such computations yield a subgroup of the true Galois group, hence a lower bound for the true width by Theorem 3.1(1). I recommend adding a qualifier such as "as computed numerically" and citing any certification in [38], so that the heuristic nature of this value is consistent across sections.
  4. [Example 2.3] The Nistér-Stewénius five-point algorithm is described as reconstructing "five points in three views," but the surrounding setup has m=2 cameras (two views); please correct "three" to "two" or clarify the intended statement.
  5. [Abstract and text] There are several small copyediting issues: "multiples instances" in the abstract, the missing closing parenthesis after "RanSaC" in the abstract, and "we we may assume" in Example 2.4.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; the monodromy-subgroup caveat is a heuristic completeness limitation, not a circular reduction.

full rationale

This paper is a tutorial/survey rather than a derivation in which a fitted quantity is renamed a prediction. The central object, the geometric Galois group, is defined independently as Gal(K/F) of the branched cover, and the monodromy computation tracks actual loops in parameter space; no parameter is fit to a subset of solutions and then reused as the predicted quantity. The five-point example is cross-checked against an independent structural argument: Section 5 states 'we know that the Galois width must equal 10, since 20 solutions divide into 10 blocks with constant essential matrices,' so the numerical monodromy output is validated by a block/deck-symmetry argument rather than being taken as its own evidence. The stated limitation that the numerically constructed group is only a subgroup of the true Galois group ('At best, it tells us only that the group G constructed in this example is a subgroup of the true Galois group') is a completeness and path-tracking-correctness caveat, not a circular reduction: it weakens the conclusion but does not make the width result equivalent to its input. The remaining headline values, such as the radial-camera width 28, are imported from prior published work, including the author's own; self-citation is present, but those cited computations are externally checkable numerical/algebraic results and are not used as a uniqueness theorem or as the sole justification of a claim that reduces to itself. I find no equation or fitted quantity that is equal by construction to the claimed prediction, so the paper has no significant circularity.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

This tutorial introduces no free parameters fitted to data; the Galois widths and degrees are quoted from prior literature or computed from group theory. The only new quantity, Galois width, is a definition from [16] with an independent GAP implementation, not an entity invoked to force a conclusion.

assumptions (5)
  • domain assumption The incidence variety X of a minimal problem is irreducible (Section 3, Assumption (2)).
    The Galois group and monodromy action on a full generic fiber are only defined as stated when X is irreducible. The paper asserts this holds for nearly all minimal problems in practice.
  • domain assumption Numerically tracked monodromy paths stay on X and generate the true geometric Galois group.
    Section 4 remarks and Section 5 acknowledge the computation is heuristic and uncertified. The five-point Galois width and the 10-path solver depend on the sampled permutations being faithful.
  • standard math Properties of Galois width in Theorem 3.1 are correct.
    Stated without proof in this tutorial and attributed to the author's preprint [16]; they are finite-group facts used to justify the Galois width values quoted later.
  • standard math The Cayley parameterization (28) produces generic rotation matrices for fabrication.
    Standard rational parameterization of SO(3); used in the Macaulay2 fabrication routine to generate a problem-solution pair.
  • domain assumption The five-point problem's 20 solutions split into 10 pairs with the same essential matrix.
    This block structure is taken from the Nister-Stewenius algorithm [61] and is used to conclude Galois width 10 in Section 5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Numerically Computing Galois Groups of Minimal Problems." pith.science (2026). https://pith.science/paper/WI55V77R

@misc{pith2026250710407,
  author       = {Pith},
  title        = {Pith review of: Numerically Computing Galois Groups of Minimal Problems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WI55V77R}},
  note         = {Machine review of arXiv:2507.10407}
}
read the original abstract

I discuss a seemingly unlikely confluence of topics in algebra, numerical computation, and computer vision. The motivating problem is that of solving multiples instances of a parametric family of systems of algebraic (polynomial or rational function) equations. No doubt already of interest to ISSAC attendees, this problem arises in the context of robust model-fitting paradigms currently utilized by the computer vision community (namely "Random Sampling and Consensus", aka "RanSaC".) This talk will give an overview of work in the last 5+ years that aspires to measure the intrinsic difficulty of solving such parametric systems, and makes strides towards practical solutions.

Figures

Figures reproduced from arXiv: 2507.10407 by the authors.

Figure 1
Figure 1. Illustration of Perspective-3-Point: p𝑖 denotes the normalized homogeneous coordinates of a 2D point y𝑖 ∈ R 2 , and 𝜆𝑖 denotes the projective depth of the 3D point q𝑖 (in affine terms, the distance from q𝑖 to the center of projection.) Example 2.1. The classic “Perspective-𝑛-Point" problem (P𝑛P) can be formulated as follows: fix q1, . . . , q𝑛 ∈ P [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Based on (6), the number of RanSaC trials 𝑁 needed to find an outlier-free subsample of size 𝑘 with 95% confidence from 𝑛 ∈ [10, 100] total correspondences, with 50% outliers. To cope with erroneous correspondences, we could try multiple trials in which we solve (2) for different sub-samples of the data; then, using the solutions obtained in each trial, we may attempt to discern which data are erroneous. This is the… view at source ↗
Figure 3
Figure 3. Geometry of the calibrated stereo pair (18). for an illustration of the 𝑚 = 2 “stereo vision" problem. To simplify the exposition, we assume each camera is calibrated—that is, an element A ∈ SE3. Thus, inverting the map Ψ𝑚,𝑛 : SE𝑚 3 × (P 3 ) 𝑛 d  P 2 𝑚𝑛 (14) (A1, . . . , A𝑚, q1, . . . , q𝑛) ↦→ (A1q1, . . . , A𝑚q𝑛). (15) seems to be a natural formulation for the reconstruction problem. However, for any 𝑚 and 𝑛, the… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Two subgroup chains in the Galois group of P3P. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 51 canonical work pages

  1. [16]

    Timothy Duff. 2025. A Galois-Theoretic Complexity Measure for Solving Systems of Algebraic Equations. arXiv preprint arXiv:2503.17884 (2025)

    arXiv 2025

  2. [1]

    Sameer Agarwal, Timothy Duff, Max Lieblich, and Rekha R Thomas. 2023. An Atlas for the Pinhole Camera. Foundations of Computational Mathematics (2023), 1–51

    work page 2023

  3. [2]

    Carlos Améndola, Julia Lindberg, and Jose Israel Rodriguez. 2016. Solving pa- rameterized polynomial systems with decomposable projections. arXiv preprint arXiv:1612.08807 (2016)

    work page Pith review arXiv 2016

  4. [3]

    Federica Arrigoni, Tomás Pajdla, and Andrea Fusiello. 2023. Viewing Graph Solvability in Practice. In IEEE/CVF International Conference on Computer Vision, ICCV 2023, Paris, France, October 1-6, 2023 . IEEE, 8113–8121. https://doi.org/10. 1109/ICCV51070.2023.00748 Numerically Computing Galois Groups of Minimal Problems ISSAC ’25, July 28–August 1, 2025, G...

    arXiv 2023

  5. [4]

    Daniel J Bates, Paul Breiding, Tianran Chen, Jonathan D Hauenstein, Anton Leykin, and Frank Sottile. 2023. Numerical nonlinear algebra. arXiv preprint arXiv:2302.08585 (2023)

    work page Pith review arXiv 2023

  6. [5]

    Nathan Bliss, Timothy Duff, Anton Leykin, and Jeff Sommars. 2018. Monodromy Solver: Sequential and Parallel. In Proceedings of the 2018 ACM on International Symposium on Symbolic and Algebraic Computation, ISSAC 2018, New York, NY, USA, July 16-19, 2018, Manuel Kauers, Alexey Ovchinnikov, and Éric Schost (Eds.). ACM, 87–94. https://doi.org/10.1145/3208976.3209007

    work page arXiv 2018

  7. [6]

    Paul Breiding and Sascha Timme. 2018. HomotopyContinuation. jl: A package for homotopy continuation in Julia. In Mathematical Software–ICMS 2018: 6th International Conference, South Bend, IN, USA, July 24-27, 2018, Proceedings 6 . Springer, 458–465

    work page 2018

  8. [7]

    Taylor Brysiewicz. 2024. Monodromy Coordinates. In International Congress on Mathematical Software. Springer, 265–274

    work page 2024

Show all 64 references
  1. [8]

    Taylor Brysiewicz, Jose Israel Rodriguez, Frank Sottile, and Thomas Yahl. 2021. Solving decomposable sparse systems.Numerical Algorithms 88, 1 (2021), 453–474. https://doi.org/10.1007/s11075-020-01045-x

    2021 doi

  2. [9]

    Tsigaridas, Stan- imire Tomov, and Benjamin B

    Chiang-Heng Chien, Hongyi Fan, Ahmad Abdelfattah, Elias P. Tsigaridas, Stan- imire Tomov, and Benjamin B. Kimia. 2022. GPU-Based Homotopy Continuation for Minimal Problems in Computer Vision. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orl...

    2022

  3. [10]

    Ondrej Chum, Jiri Matas, and Josef Kittler. 2003. Locally Optimized RANSAC. In Pattern Recognition, 25th DAGM Symposium, Magdeburg, Germany, September 10-12, 2003, Proceedings (Lecture Notes in Computer Science, Vol. 2781) , Bernd Michaelis and Gerald Krell (Eds.). Springer, 2...

    2003 doi

  4. [12]

    Erin Connelly, Timothy Duff, and Jessie Loucks-Tavitas. 2024. Algebra and geometry of camera resectioning. Math. Comp. (2024)

    2024

  5. [13]

    Bernard Deconinck and Mark van Hoeij. 2001. Computing Riemann matrices of algebraic curves. Vol. 152/153. 28–46. Advances in nonlinear mathematics and science

    2001

  6. [14]

    Michel Dhome, Marc Richetin, Jean-Thierry Lapresté, and Gérard Rives. 1989. Determination of the Attitude of 3D Objects from a Single Perspective View. IEEE Trans. Pattern Anal. Mach. Intell. 11, 12 (1989), 1265–1278. https://doi.org/ 10.1109/34.41365

    1989 doi

  7. [17]

    Timothy Duff, Cvetelina Hill, Anders Jensen, Kisun Lee, Anton Leykin, and Jeff Sommars. 2019. Solving polynomial systems via homotopy continuation and monodromy. IMA J. Numer. Anal. 39, 3 (2019), 1421–1446. https://doi.org/10. 1093/imanum/dry017

    2019

  8. [18]

    Timothy Duff, Kathlén Kohn, Anton Leykin, and Tomas Pajdla. 2023. PLMP–[oint- Line Minimal Problems in Complete Multi-view Visibility. IEEE Transactions on Pattern Analysis and Machine Intelligence 46, 1 (2023), 421–435

    2023

  9. [19]

    Timothy Duff, Kathlén Kohn, Anton Leykin, and Tomás Pajdla. 2024. PL1P: Point- Line Minimal Problems under Partial Visibility in Three Views. Int. J. Comput. Vis. 132, 8 (2024), 3302–3323. https://doi.org/10.1007/S11263-024-01992-1

    2024 doi

  10. [20]

    Timothy Duff, Viktor Korotynskiy, Tomas Pajdla, and Margaret H. Regan. 2022. Galois/monodromy groups for decomposing minimal problems in 3D reconstruc- tion. SIAM J. Appl. Algebra Geom. 6, 4 (2022), 740–772. https://doi.org/10.1137/ 21M1422872

    2022

  11. [21]

    Timothy Duff, Viktor Korotynskiy, Tomás Pajdla, and Margaret H. Regan. 2023. Using monodromy to recover symmetries of polynomial systems. InProceedings of the 2023 International Symposium on Symbolic and Algebraic Computation, ISSAC 2023, Tromsø, Norway, July 24-27, 2023, Alic...

    2023

  12. [23]

    Timothy Duff and Michael Ruddy. 2023. Signatures of algebraic curves via numerical algebraic geometry. J. Symb. Comput. 115 (2023), 452–477. https: //doi.org/10.1016/J.JSC.2022.08.003

    2023 doi

  13. [24]

    Timothy Duff and Felix Rydell. 2025. Metric Multiview Geometry–a Catalogue in Low Dimensions. (to appear) Acta Universitatis Sapientiae, Mathematica (2025)

    2025

  14. [25]

    Regan, David da Costa de Pinho, Elias P

    Ricardo Fabbri, Timothy Duff, Hongyi Fan, Margaret H. Regan, David da Costa de Pinho, Elias P. Tsigaridas, Charles W. Wampler, Jonathan D. Hauenstein, Peter J. Giblin, Benjamin B. Kimia, Anton Leykin, and Tomás Pajdla. 2023. Trifocal Relative Pose From Lines at Points. IEEE Tr...

    2023

  15. [26]

    Hongyi Fan, Joe Kileel, and Benjamin B. Kimia. 2023. Condition numbers in multiview geometry, instability in relative pose estimation, and RANSAC. CoRR abs/2310.02719 (2023). https://doi.org/10.48550/ARXIV.2310.02719 arXiv:2310.02719

    2023 doi arXiv

  16. [27]

    Jean-Charles Faugère, Guillaume Moroz, Fabrice Rouillier, and Mohab Safey El Din. 2008. Classification of the perspective-three-point problem, discriminant variety and real solving polynomial systems of inequalities. In Symbolic and Algebraic Computation, International Symposi...

    2008

  17. [28]

    Martin A Fischler and Robert C Bolles. 1981. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography. Commun. ACM 24, 6 (1981), 381–395

    1981

  18. [29]

    Fitzgibbon

    Andrew W. Fitzgibbon. 2001. Simultaneous linear estimation of multiple view geometry and lens distortion. In 2001 IEEE Computer Society Conference on Com- puter Vision and Pattern Recognition (CVPR 2001), with CD-ROM, 8-14 December 2001, Kauai, HI, USA . IEEE Computer Society,...

    2001

  19. [30]

    Louis Gaillard and Mohab Safey El Din. 2024. Solving parameter-dependent semi-algebraic systems. In Proceedings of the 2024 International Symposium on Symbolic and Algebraic Computation, ISSAC 2024, Raleigh, NC, USA, July 16-19, 2024, Jonathan D. Hauenstein, Wen-shin Lee, and ...

    2024

  20. [31]

    André Galligo and Adrien Poteaux. 2009. Continuations and Monodromy on Ran- dom Riemann Surfaces. InProceedings of the 2009 Conference on Symbolic Numeric Computation (Kyoto, Japan) (SNC ’09). Association for Computing Machinery, New York, NY, USA, 115–124. https://doi.org/10....

    2009

  21. [32]

    Grayson and Michael E

    Daniel R. Grayson and Michael E. Stillman. [n. d.]. Macaulay2, a software system for research in algebraic geometry. Available at http://www.math.uiuc.edu/ Macaulay2/

  22. [33]

    The GAP Group. [n. d.]. GAP – Groups, Algorithms, and Programming. Available at https://www.gap-system.org/

  23. [34]

    [2024] ©2024

    Alexandre Guillemot and Pierre Lairez. [2024] ©2024. Validated numerics for algebraic path tracking. In ISSAC’24—Proceedings of the 2024 International Symposium on Symbolic and Algebraic Computation . ACM, New York, 36–45. https://doi.org/10.1145/3666000.3669673

    2024

  24. [35]

    Haralick, Chung-Nan Lee, Karsten Ottenberg, and Michael Nölle

    Robert M. Haralick, Chung-Nan Lee, Karsten Ottenberg, and Michael Nölle. 1994. Review and analysis of solutions of the three point perspective pose estimation problem. Int. J. Comput. Vis. 13, 3 (1994), 331–356. https://doi.org/10.1007/ BF02028352

    1994

  25. [36]

    Petr Hruby, Timothy Duff, Anton Leykin, and Tomas Pajdla. 2023. Learning to Solve Hard Minimal Problems. IEEE Transactions on Pattern Analysis and Machine Intelligence (2023)

    2023

  26. [37]

    Petr Hruby, Timothy Duff, and Marc Pollefeys. 2024. Efficient Solution of Point- Line Absolute Pose. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 21316–21325

    2024

  27. [38]

    Petr Hruby, Viktor Korotynskiy, Timothy Duff, Luke Oeding, Marc Pollefeys, Tomás Pajdla, and Viktor Larsson. 2023. Four-view Geometry with Unknown Radial Distortion. In IEEE/CVF Conference on Computer Vision and Pattern Recog- nition, CVPR 2023, Vancouver, BC, Canada, June 17-...

    2023

  28. [39]

    Kim Kiehn, Albin Ahlbäck, and Kathlén Kohn. 2025. PLMP–Point-Line Minimal Problems for Projective SfM. (to appear) Proceedings of ICCV 2025 (2025)

    2025

  29. [40]

    Joe Kileel. 2017. Minimal Problems for the Calibrated Trifocal Variety. SIAM J. Appl. Algebra Geom. 1, 1 (2017), 575–598. https://doi.org/10.1137/16M1104482

    2017 doi

  30. [41]

    Joe Kileel and Kathlén Kohn. 2022. Snapshot of algebraic vision. arXiv preprint arXiv:2210.11443 (2022)

    2022 arXiv

  31. [42]

    Viktor Kocur, Daniel Kyselica, and Zuzana Kukelova. 2024. Robust Self- Calibration of Focal Lengths from the Fundamental Matrix. In IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, W A, USA, June 16-22, 2024. IEEE, 5220–5229. https://doi.or...

    2024

  32. [43]

    Zuzana Kukelova, Cenek Albl, Akihiro Sugimoto, Konrad Schindler, and Tomás Pajdla. 2020. Minimal Rolling Shutter Absolute Pose with Unknown Focal Length and Radial Distortion. InComputer Vision - ECCV 2020 - 16th European Conference, Glasgow, UK, August 23-28, 2020, Proceeding...

    2020 doi

  33. [44]

    Fitzgibbon

    Zuzana Kukelova, Jan Heller, and Andrew W. Fitzgibbon. 2016. Efficient In- tersection of Three Quadrics and Applications in Computer Vision. In 2016 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2016, Las Vegas, NV, USA, June 27-30, 2016 . IEEE Computer Soci...

    2016 doi

  34. [45]

    Zuzana Kukelova and Viktor Larsson. 2019. Radial Distortion Triangulation. In IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2019, Long Beach, CA, USA, June 16-20, 2019. Computer Vision Foundation / IEEE, 9681–9689. https://doi.org/10.1109/CVPR.2019.00991 ISS...

    2019

  35. [46]

    Zuzana Kukelova and Tomás Pajdla. 2007. A minimal solution to the autocalibra- tion of radial distortion. In 2007 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2007), 18-23 June 2007, Minneapolis, Min- nesota, USA. IEEE Computer Society. htt...

    2007

  36. [47]

    Anton Leykin. 2011. Numerical algebraic geometry. The Journal of Software for Algebra and Geometry 3 (2011), 5–10

    2011

  37. [48]

    Anton Leykin, Jose Israel Rodriguez, and Frank Sottile. 2018. Trace test. Arnold Math. J. 4, 1 (2018), 113–125. https://doi.org/10.1007/s40598-018-0084-3

    2018 doi

  38. [49]

    Anton Leykin and Frank Sottile. 2009. Galois groups of Schubert problems via homotopy computation. Math. Comp. 78, 267 (2009), 1749–1765

    2009

  39. [50]

    Anton Leykin and Jan Verschelde. 2009. Decomposing solution sets of polynomial systems: a new parallel monodromy breakup algorithm. Int. J. Comput. Sci. Eng. 4, 2 (2009), 94–101. https://doi.org/10.1504/IJCSE.2009.027001

    2009 arXiv

  40. [51]

    Maxim, Jose I

    Laurentiu G. Maxim, Jose I. Rodriguez, and Botong Wang. 2020. Euclidean distance degree of the multiview variety. SIAM Journal on Applied Algebra and Geometry 4, 1 (2020), 28–48. https://doi.org/10.1137/18M1233406

    2020 doi

  41. [52]

    Hartley, and Henrik Stewénius

    David Nistér, Richard I. Hartley, and Henrik Stewénius. 2007. Using Galois Theory to Prove Structure from Motion Algorithms are Optimal. In 2007 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2007), 18-23 June 2007, Minneapolis, Minnesota, US...

    2007

  42. [53]

    Mikael Persson and Klas Nordberg. 2018. Lambda Twist: An Accurate Fast Robust Perspective Three Point (P3P) Solver. In Computer Vision - ECCV 2018 - 15th European Conference, Munich, Germany, September 8-14, 2018, Proceedings, Part IV (Lecture Notes in Computer Science, Vol. 1...

    2018 doi

  43. [54]

    Srikumar Ramalingam, Sofien Bouaziz, and Peter F. Sturm. 2011. Pose estimation using both points and lines for geo-localization. In IEEE International Conference on Robotics and Automation, ICRA 2011, Shanghai, China, 9-13 May 2011 . IEEE, 4716–4723. https://doi.org/10.1109/IC...

    2011

  44. [55]

    Jürgen Richter-Gebert. 2011. Perspectives on projective geometry . Springer, Hei- delberg. xxii+571 pages. https://doi.org/10.1007/978-3-642-17286-1 A guided tour through real and complex geometry

    2011 doi

  45. [56]

    Maxwell Rosenlicht. 1956. Some basic theorems on algebraic groups. Amer. J. Math. 78 (1956), 401–443. https://doi.org/10.2307/2372523

    1956 doi

  46. [57]

    Schönberger, Viktor Larsson, and Marc Pollefeys

    Johannes L. Schönberger, Viktor Larsson, and Marc Pollefeys. 2025. Fixing the RANSAC Stopping Criterion. CoRR abs/2503.07829 (2025). https://doi.org/10. 48550/ARXIV.2503.07829 arXiv:2503.07829

    2025 doi arXiv

  47. [58]

    Sommese, Jan Verschelde, and Charles W

    Andrew J. Sommese, Jan Verschelde, and Charles W. Wampler. 2001. Numeri- cal decomposition of the solution sets of polynomial systems into irreducible components. SIAM J. Numer. Anal. 38, 6 (2001), 2022–2046

    2001

  48. [59]

    Sommese and Charles W

    Andrew J. Sommese and Charles W. Wampler, II. 2005. The numerical solution of systems of polynomials. World Scientific Publishing Co. Pte. Ltd., Hackensack, NJ. xxii+401 pages. https://doi.org/10.1142/9789812567727 Arising in engineering and science

    2005 doi

  49. [60]

    Frank Sottile and Thomas Yahl. 2021. Galois groups in enumerative geometry and applications. arXiv preprint arXiv:2108.07905 (2021)

    2021

  50. [61]

    Henrik Stewenius, Christopher Engels, and David Nistér. 2006. Recent devel- opments on direct relative orientation. ISPRS Journal of Photogrammetry and Remote Sensing 60, 4 (2006), 284–294

    2006

  51. [62]

    Peter Sturm. 2011. A historical survey of geometric computer vision. In In- ternational Conference on Computer Analysis of Images and Patterns . Springer, 1–8

    2011

  52. [63]

    Bernd Sturmfels and Simon Telen. 2021. Likelihood equations and scattering amplitudes. Algebraic Statistics 12, 2 (2021), 167–186. https://doi.org/10.2140/ astat.2021.12.167

    2021

  53. [64]

    Joris van Der Hoeven. 2011. Reliable homotopy continuation. Technical Report (2011)

    2011

  54. [65]

    Chi Xu, Lilian Zhang, Li Cheng, and Reinhard Koch. 2017. Pose Estimation from Line Correspondences: A Complete Analysis and a Series of Solutions. IEEE Trans. Pattern Anal. Mach. Intell. 39, 6 (2017), 1209–1222. https://doi.org/10.1109/ TPAMI.2016.2582162

    2017

  55. [66]

    Juan Xu, Michael Burr, and Chee Yap. 2018. An approach for certifying homotopy continuation paths: Univariate case. In Proceedings of the 2018 ACM International Symposium on Symbolic and Algebraic Computation . 399–406

    2018

  56. [2022]

    https://doi.org/10.1109/CVPR52688.2022.01531

    IEEE, 15744–15755. https://doi.org/10.1109/CVPR52688.2022.01531

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.