Abstract
We introduce an atlas of algebro-geometric objects associated with image formation in pinhole cameras. The nodes of the atlas are algebraic varieties or their vanishing ideals related to each other by projection or elimination and restriction or specialization, respectively. This atlas offers a unifying framework for the study of problems in 3D computer vision. We initiate the study of the atlas by completely characterizing a part of the atlas stemming from the triangulation problem. We conclude with several open problems and generalizations of the atlas.
Similar content being viewed by others
Notes
To distinguish between known and unknown quantities, we use a bar over an object to indicate specialization. For instance, A stands for a symbolic \(3 \times 4\) matrix denoting a camera, while \({\bar{A}}\) is a \(3 \times 4\) scalar matrix realizing a camera. We also use bold face letters to indicate collections. For instance, we use \(\mathbf {A}\) and \({\bar{\mathbf {A}}}\) to specify a collection of symbolic and scalar cameras, respectively.
We note that the definition of \(\Gamma _{\mathbf {A}, \mathbf {q}, \mathbf {p}}^{m,n}\) is independent of the choice of U. As such, it is insensitive to certain physical assumptions about the camera matrices (e.g., that they have full rank, or that their centers do not coincide.) In particular, although a generic point \(({\bar{\mathbf {A}}}, {\bar{\mathbf {q}}}, {\bar{\mathbf {p}}}) \in \Gamma _{\mathbf {A}, \mathbf {q}, \mathbf {p}}^{m,n}\) will be such that each \(A_i\) has full rank and all \(A_i q_j \) are defined, these conditions do not hold for an arbitrary point \(({\bar{\mathbf {A}}}, {\bar{\mathbf {q}}}, {\bar{\mathbf {p}}}) \in \Gamma _{\mathbf {A}, \mathbf {q}, \mathbf {p}}^{m,n}.\)
Not every result of successively applying projection and specialization operations to \(\Gamma _{\mathbf {A}, \mathbf {q}, \mathbf {p}}^{m,n}\) is included here. For example \(\Gamma ^{m,n}_{\mathbf {A}}\) and \(\Gamma ^{m,n}_{\mathbf {q}}\) are trivial. Similarly \(\Gamma ^{m,n}_{{\bar{\mathbf {A}}},\mathbf {q},{\bar{\mathbf {p}}}}\) and \(\Gamma ^{m,n}_{\mathbf {A},{\bar{\mathbf {q}}},{\bar{\mathbf {p}}}}\) are defined by linear equations and not interesting for projective cameras. However, this can change as the model for the camera is varied. For example, \(\Gamma ^{m,n}_{\mathbf {A},{\bar{\mathbf {q}}},{\bar{\mathbf {p}}}}\) is an interesting nonlinear variety for Euclidean cameras. See Sect. 8 for more.
References
Sameer Agarwal, Hon-leung Lee, Bernd Sturmfels, and Rekha R. Thomas, On the existence of epipolar matrices, International Journal of Computer Vision 121 (2017), no. 3, 403–415.
Sameer Agarwal, Andrew Pryhuber, and Rekha R. Thomas, Ideals of the Multiview Variety, IEEE Transactions on Pattern Analysis & Machine Intelligence 43 (2021), no. 04, 1279–1292.
Chris Aholt, Sameer Agarwal, and Rekha Thomas, A QCQP approach to triangulation, European Conference on Computer Vision, Springer-Verlag, 2012, p. 654–667.
Chris Aholt and Luke Oeding, The ideal of the trifocal variety, Mathematics of Computation 83 (2014), no. 289, 2553–2574.
Chris Aholt, Bernd Sturmfels, and Rekha Thomas, A Hilbert scheme in computer vision, Canadian Journal of Mathematics 65 (2013), no. 5, 961–988.
Michael F. Atiyah and Ian G. Macdonald, Introduction to Commutative Algebra, Addison-Wesley Publishing Co., Reading, Mass.-London-Don Mills, Ont., 1969.
Adrien Bartoli and Peter Sturm, Structure-from-motion using lines: Representation, triangulation, and bundle adjustment, Computer Vision and Image Understanding 100 (2005), no. 3, 416–441.
David Bernstein and Andrei Zelevinsky, Combinatorics of maximal minors, Journal of Algebraic Combinatorics 2 (1993), no. 2, 111–121.
Adam Boocher, Free resolutions and sparse determinantal ideals, Mathematical Research Letters 19 (2012), no. 4, 805–821.
Paul Breiding, Felix Rydell, Elima Shehu, and Angélica Torres, Line multiview varieties, 2022.
Stefan Carlsson and Daphna Weinshall, Dual computation of projective shape and camera positions from multiple images, International Journal of Computer Vision 27 (1998), no. 3, 227–241.
Dustin Cartwright and Bernd Sturmfels, The Hilbert scheme of the diagonal in a product of projective spaces, International Mathematics Research Notices 2010 (2010), no. 9, 1741–1771.
Aldo Conca, Emanuela De Negri, and Elisa Gorla, Universal Gröbner bases for maximal minors, International Mathematics Research Notices 2015 (2015), no. 11, 3245–3262.
David A. Cox, The moving curve ideal and the Rees algebra, Theoretical Computer Science 392 (2008), no. 1-3, 23–36.
David A. Cox, John Little, and Donal O’Shea, Ideals, Varieties, and Algorithms: An Introduction to Computational Algebraic Geometry and Commutative Algebra, 4 ed., Undergraduate Texts in Mathematics, Springer, Cham, 2015.
Song De Ma, Conics-based stereo, motion estimation, and pose determination, International Journal of Computer Vision 10 (1993), no. 1, 7–25.
Michel Demazure, Sur deux problemes de reconstruction, Tech. Report RR-0882, INRIA, July 1988.
Jan Draisma, Emil Horobeţ, Giorgio Ottaviani, Bernd Sturmfels, and Rekha R Thomas, The Euclidean distance degree of an algebraic variety, Foundations of Computational Mathematics 16 (2016), no. 1, 99–149.
Timothy Duff, Kathlén Kohn, Anton Leykin, and Tomas Pajdla, PL\(_1\)P-Point-Line minimal problems under partial visibility in three views, European Conference on Computer Vision, Springer, 2020, pp. 175–192.
Olivier D. Faugeras and Bernard Mourrain, On the Geometry and Algebra of the Point and Line Correspondences Between N Images, IEEE International Conference on Computer Vision, 1995, pp. 951–956.
M. Giusti and M. Merle, Singularités isolées et sections planes de variétés déterminantielles. II. Sections de variétés déterminantielles par les plans de coordonnées, Lecture Notes in Math., vol. 961, pp. 103–118, Springer, Berlin, 1982.
Daniel R. Grayson and Michael E. Stillman, Macaulay2, a software system for research in algebraic geometry, Available at https://math.uiuc.edu/Macaulay2/.
Richard Hartley and Andrew Zisserman, Multiple View Geometry in Computer Vision, Cambridge University Press, 2004.
Richard I Hartley, An investigation of the essential matrix, Tech. report, General Electric CRD, 1995.
Richard I Hartley, Lines and points in three views and the trifocal tensor, International Journal of Computer Vision 22 (1997), no. 2, 125–140.
Richard I Hartley and Peter Sturm, Triangulation, Computer Vision and Image Understanding 68 (1997), no. 2, 146–157.
Anders Heyden and Kalle Åström, Algebraic properties of multilinear constraints, Mathematical Methods in the Applied Sciences 20 (1997), no. 13, 1135–1162.
Fredrik Kahl and Anders Heyden, Affine structure and motion from points, lines and conics, International Journal of Computer Vision 33 (1999), no. 3, 163–180.
Joe Kileel, Minimal problems for the calibrated trifocal variety, SIAM Journal on Applied Algebra and Geometry 1 (2017), no. 1, 575–598.
Laurent Kneip, Roland Siegwart, and Marc Pollefeys, Finding the exact rotation between two images independently of the translation, European Conference on Computer Vision, vol. 7577, Springer, 2012, pp. 696–709.
Zuzana Kukelova, Martin Bujnak, and Tomas Pajdla, Polynomial eigenvalue solutions to minimal problems in computer vision, IEEE Transactions on Pattern Analysis and Machine Intelligence 34 (2011), no. 7, 1381–1393.
Viktor Larsson, Kalle Astrom, and Magnus Oskarsson, Efficient solvers for minimal problems by syzygy-based reduction, IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 820–829.
Vincent Lepetit, Francesc Moreno-Noguer, and Pascal Fua, EPnP: An accurate\({O}(n)\)solution to the PNP problem, International Journal of Computer Vision 81 (2009), no. 2, 155–166.
Max Lieblich and Lucas Van Meter, Two Hilbert schemes in computer vision, SIAM Journal on Applied Algebra and Geometry 4 (2020), no. 2, 297–321.
Peter Lindstrom, Triangulation made easy, 2010 IEEE Conference on Computer Vision and Pattern Recognition, 2010, pp. 1554–1561.
H Christopher Longuet-Higgins, A computer algorithm for reconstructing a scene from two projections, Nature 293 (1981), no. 5828, 133–135.
Quan-Tuan Luong and Olivier D Faugeras, The fundamental matrix: Theory, algorithms, and stability analysis, International Journal of Computer Vision 17 (1996), no. 1, 43–75.
Yi Ma, Stefano Soatto, Jana Košecká, and Shankar Sastry, An Invitation to 3-D Vision: from Images to Geometric Models, vol. 26, Springer, 2004.
Laurentiu G Maxim, Jose I Rodriguez, and Botong Wang, Euclidean distance degree of the multiview variety, SIAM Journal on Applied Algebra and Geometry 4 (2020), no. 1, 28–48.
Mateusz Michałek and Bernd Sturmfels, Invitation to Nonlinear Algebra, vol. 211, American Mathematical Society, 2021.
Behrooz Nasihatkon, Richard Hartley, and Jochen Trumpf, A generalized projective reconstruction theorem and depth constraints for projective factorization, International Journal of Computer Vision 115 (2015), no. 2, 87–114.
David Nistér, An efficient solution to the five-point relative pose problem, IEEE Transactions on Pattern Analysis and Machine Intelligence 26 (2004), no. 6, 756–770.
Luke Oeding, The quadrifocal variety, Linear Algebra and Its Applications 512 (2017), 306–330.
Patrik Persson and Kalle Åström, Global trifocal adjustment, Scandinavian Conference on Image Analysis, Springer, 2019, pp. 287–298.
Long Quan, Invariants of six points and projective reconstruction from three uncalibrated images, IEEE Transactions on Pattern Analysis and Machine Intelligence 17 (1995), no. 1, 34–46.
Michel Raynaud and Laurent Gruson, Criteres de platitude et de projectivité, Inventiones Mathematicae 13 (1971), no. 1, 1–89.
Antonio L Rodríguez, Pedro E López-de Teruel, and Alberto Ruiz, Reduced epipolar cost for accelerated incremental SfM, IEEE Conference on Computer Vision and Pattern Recognition, 2011, pp. 3097–3104.
Ewelina Rupnik and Marc Deseilligny, Towards structureless bundle adjustment with two-and three-view structure approximation, ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. 2, 2020, pp. 71–78.
Frederik Schaffalitzky, Andrew Zisserman, Richard I. Hartley, and Philip H. S. Torr, A six point solution for structure and motion, European Conference on Computer Vision (David Vernon, ed.), vol. 1842, Springer, 2000, pp. 632–648.
J. Schneider, C. Stachniss, and W. Förstner, On the quality and efficiency of approximate solutions to bundle adjustment with epipolar and trifocal constraints, ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences IV-2/W3 (2017), 81–88.
Amnon Shashua and Lior Wolf, On the structure and properties of the quadrifocal tensor, European Conference on Computer Vision, Springer, 2000, pp. 710–724.
Richard Steffen, Jan-Michael Frahm, and Wolfgang Förstner, Relative bundle adjustment based on trifocal constraints, European Conference on Computer Vision, Springer, 2010, pp. 282–295.
Henrik Stewénius, Gröbner basis methods for minimal problems in computer vision, PhD thesis 2005.
Henrik Stewénius, Frederik Schaffalitzky, and David Nistér, How hard is 3-view triangulation really?, IEEE International Conference on Computer Vision, vol. 1, 2005, pp. 686–693.
Bernd Sturmfels, Gröbner bases and Convex Polytopes, University Lecture Series, vol. 8, American Mathematical Society, Providence, RI, 1996.
Bernd Sturmfels and Andrei Zelevinsky, Maximal minors and their leading terms, Advances in Mathematics 98 (1993), no. 1, 65–112.
Camillo J Taylor and David J Kriegman, Structure and motion from line segments in multiple images, IEEE Transactions on Pattern Analysis and Machine Intelligence 17 (1995), no. 11, 1021–1032.
Sebastian Thrun, Wolfram Burgard, and Dieter Fox, Probabilistic robotics, MIT Press, 2005.
Matthew Trager, Martial Hebert, and Jean Ponce, The joint image handbook, IEEE International Conference on Computer Vision, 2015, pp. 909–917.
Matthew Trager, Martial Hebert, and Jean Ponce, Coordinate-free carlsson-weinshall duality and relative multi-view geometry, IEEE Conference on Computer Vision and Pattern Recognition, 2019, pp. 225–233.
Bill Triggs, Matching Constraints and the Joint Image, IEEE International Conference on Computer Vision, 1995, pp. 338–343.
Bill Triggs, Factorization methods for projective structure and motion, IEEE Conference on Computer Vision and Pattern Recognition, 1996, pp. 845–851.
Bill Triggs, Philip F McLauchlan, Richard I Hartley, and Andrew W Fitzgibbon, Bundle Adjustment—a Modern Synthesis, International Workshop on Vision Algorithms, Springer, 1999, pp. 298–372.
Cynthia Vinzant, Real radical initial ideals, J. Algebra 352 (2012), 392–407.
Acknowledgements
We thank Jessie Loucks Tavitas, Erin Connelly & Craig Citro for helpful discussions. Timothy Duff acknowledges support from the National Science Foundation Mathematical Sciences Postdoctoral Research Fellowship (DMS-2103310). Max Lieblich was partially supported by a National Science Foundation Grant (DMS - 1902251). Rekha Thomas was partially supported by a National Science Foundation grant (DMS - 1719538).
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Teresa Krick and Hans Munthe-Kaas.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Invited paper associated to the FoCM 2021 Online Seminar lecture Chirality in Vision presented by Rekha Thomas in July 2021.
Appendices
Appendix A. Dimension Counts
Proposition A.1
Below, \({\bar{\mathbf {A}}} \in (\mathbb {P}^{11})^m, {\bar{\mathbf {q}}} \in (\mathbb {P}^3)^n, {\bar{\mathbf {p}}} \in \Gamma ^{m,n}_{\mathbf {p}}\) are generic whenever they appear.
Remark A.2
For m and n sufficiently large, the formulas above involving \(\min \) and \(\max \) expressions can be simplified as follows:
Proof
(A.1): This follows at once from the birational equivalence \(\square \)
(A.2): Consider the projection
A fiber \(\pi _\mathbf {q}^{-1} ({\bar{\mathbf {A}}}, {\bar{\mathbf {p}}})\) can be identified with the projective linear space of all q satisfying \(A_i q_j \sim p_{i j}\). Equivalently, each of the 2mn matrices \(\left( \begin{matrix}A_i q_j&p_{i j}\end{matrix}\right) \) is of rank one. For generic \(({\bar{\mathbf {A}}}, {\bar{\mathbf {p}}}) \in \Gamma ^{m,n}_{\mathbf {A},\mathbf {p}},\) the \(2\times 2\) minors impose 2 linear conditions on \({\bar{\mathbf {q}}},\) so that \(\pi _\mathbf {q}^{-1} ({\bar{\mathbf {A}}}, {\bar{\mathbf {p}}})\) is a projective linear space of dimension \(\max (3n-2mn, 0).\) Hence, using (A.1) and the fiber-dimension theorem,
(A.3): Similar to (A.2), consider
For generic \(({\bar{\mathbf {q}}}, {\bar{\mathbf {p}}}) \in \Gamma ^{m,n}_{\mathbf {q}, \mathbf {p}},\) the fibers \(\pi _\mathbf {A}^{-1} ({\bar{\mathbf {q}}}, {\bar{\mathbf {p}}})\) are projective linear spaces of dimension \(\max (11m - 2mn, 0)\) which are defined by the \(\mathbf {A}\)-linear \(2\times 2\) minors of \(\left( \begin{matrix}A_i q_j&p_{i j}\end{matrix}\right) .\) Hence,
(A.4): For generic \(q_1, \ldots , q_5 \in \mathbb {P}^3,\) define \(\sigma = \{ \sigma _1, \ldots , \sigma _4 \} \subset [5],\)
and consider the projective change of basis matrix
where \(\bullet ^{-1}\) denotes matrix inversion. When defined, the projective transformation defined by \(C_{q_1, \ldots , q_5}\) maps \(q_1 \ldots q_5\) onto the standard projective basis:
Consider the projection
We observe a version of projective ambiguity [23, p 265], stating that the fibers of \(\pi _{\mathbf {A}, \mathbf {q}}\) are invariant under the action of \( {{\,\mathrm{\mathrm PGL}\,}}_4\) described in Sect. 2.2. Suppose first that \(n<6\). We need to show \(\dim \Gamma ^{m,n}_\mathbf {p}= 2mn.\) Let \({\bar{\mathbf {p}}} \in \Gamma ^{m,n}_\mathbf {p}\) be generic and suppose \(({\bar{\mathbf {A}}}, {\bar{\mathbf {q}}}, {\bar{\mathbf {p}}}) \in \pi _{\mathbf {A}, \mathbf {q}}^{-1} ({\bar{\mathbf {p}}}).\) Then for generic \({\bar{\mathbf {q}}} ' \in (\mathbb {P}^3)^n,\) we may find \(({\bar{\mathbf {A}}}', {\bar{\mathbf {q}}} ' , {\bar{\mathbf {p}}} ) \in \pi _{\mathbf {A}, \mathbf {q}}^{-1} ({\bar{\mathbf {p}}})\) by projective change of basis \(H = C_{{\tilde{\mathbf {q}}} '}^{-1} C_{{\tilde{\mathbf {q}}}},\) where \({\tilde{\mathbf {q}}}\) and \({\tilde{\mathbf {q}}}' \) extend \({\bar{\mathbf {q}}}\) and \({\bar{\mathbf {q}}} ' \) to projective bases when \(n<5.\) In other words, the generic fiber of \(\Gamma ^{m,n}_{\mathbf {q}, \mathbf {p}} \rightarrow \Gamma ^{m,n}_\mathbf {p}\) has dimension 3n. Applying (A.1),
Now suppose \(n\ge 6\). For \(m=1\) camera, (A.4) asserts that \(\dim \Gamma ^{m,n}_\mathbf {p}= 2 n ,\) which follows since there are no constraints on image points. Otherwise, observe that the quantity
is increasing in n for fixed \(m\ge 2\) and increasing in m for fixed \(n\ge 6.\) Moreover, this quantity equals zero precisely in the minimal cases \((m,n)=(2,7), \, (3, 6).\) Thus, (A.4) asserts that \(\dim \Gamma ^{m,n}_{\mathbf {p}} = 11m + 3n - 15\) whenever either \(m\ge 2\) and \(n\ge 7\) or \(m\ge 3\) and \(n\ge 6.\) This leaves one exceptional case for \(n\ge 6,\) which is \((m,n) = (2,6)\); here, to show that \(\dim \Gamma ^{m,n}_{\mathbf {p}} = 2mn = 24\), it suffices to verify that the Jacobian of \(\pi _{\mathbf {A}, \mathbf {q}}\) evaluated at some point in local coordinates has rank 24. The same Jacobian check gives us \(\dim \Gamma ^{m,n}_{\mathbf {p}} = 11m + 3n - 15\) for the two minimal cases; equivalently, \(\dim \pi _{\mathbf {A}, \mathbf {q}}^{-1} ({\bar{\mathbf {p}}})= 15\) for generic \({\bar{\mathbf {p}}} \in \Gamma ^{m,n}_\mathbf {p}.\) Finally, if either \(m\ge 2\) and \(n\ge 8\) or \(m\ge 3\) and \(n\ge 7,\) note that the fiber \(\pi _{\mathbf {A},\mathbf {q}}^{-1} ({\bar{\mathbf {p}}}) \) for generic \({\bar{\mathbf {p}}} \in \Gamma ^{m,n}_\mathbf {p}\) is nonempty, and thus has dimension at least 15 by projective ambiguity. Since \(\pi _{\mathbf {A},\mathbf {q}}^{-1} ({\bar{\mathbf {p}}}) \) projects onto a fiber for one of the minimal cases, we also have \(\dim \pi _{\mathbf {A}, \mathbf {q}}^{-1} ({\bar{\mathbf {p}}}) \le 15.\) Thus,
(A.5)–(A.11) In all cases, \(\Gamma ^{m,n}_{{\bar{X}}, Y}\) is the generic fiber of \(\Gamma ^{m,n}_{X ,Y} \rightarrow \Gamma ^{m,n}_{X},\) so these formulas follow from the fiber dimension theorem and (A.1)–(A.4). \(\square \)
Appendix B. Miscellaneous Proofs
1.1 B.1. Generic Cameras
Proposition B.1
If a camera arrangement \({\bar{\mathbf {A}}} = ({\bar{A}}_1,\ldots {\bar{A}}_m)\) is ultra minor generic then it is minor generic, and if \({\bar{\mathbf {A}}}\) is minor generic, then it has pairwise distinct centers.
Proof
If \({\bar{\mathbf {A}}}\) is ultra minor generic, then all \(k \times k\) minors of \(\left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) \) are nonzero for any \(k \in [4]\). In particular, all \(4 \times 4\) minors are nonzero and \({\bar{\mathbf {A}}}\) is minor generic. If \({\bar{\mathbf {A}}}\) is minor generic, then for any \(1 \le i < j \le m\), the \(4\times 6\) matrix \(\left( \begin{array}{c|c}{{\bar{A}}_i}^\top&{{\bar{A}}_j}^\top \end{array} \right) \) has rank 4. This implies that \({\bar{A}}_i\) and \({\bar{A}}_j\) have district centers. \(\square \)
Theorem B.2
-
(1)
A camera arrangement \({\bar{\mathbf {A}}}\) has pairwise distinct centers if and only if it is equivalent to a minor generic camera arrangement under the group action (2.1).
-
(2)
A camera arrangement \({\bar{\mathbf {A}}}\) is minor generic if and only if it is equivalent to an ultra minor generic arrangement under the group action (2.2).
-
(3)
A camera arrangement \({\bar{\mathbf {A}}}\) has pairwise distinct centers if and only if it is equivalent to an ultra minor generic camera arrangement under the group action (2.3).
Proof
-
(1)
This statement was proved in [2, Lemma 3.6].
-
(2)
We already saw in Proposition B.1 that ultra minor genericity implies minor genericity. For the other direction, fix a minor generic arrangement \(({\bar{A}}_1,\ldots {\bar{A}}_m)\). Let \(\sigma \in \left( {\begin{array}{c}[4]\\ k\end{array}}\right) \) and \(\tau \in \left( {\begin{array}{c}[3m]\\ k\end{array}}\right) \) be subsets indexing the rows and columns of some \(k\times k\) minor of \(\left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) \). Using the Cauchy-Binet theorem,
$$\begin{aligned} \begin{aligned}&\det \left( \left( \begin{array}{c|c|c}{({\bar{A}}_1 H)}^\top&\cdots&{({\bar{A}}_m H)}^{\top {}}\end{array}\right) [\sigma , \tau ] \right) \\&\quad = \displaystyle \sum _{\upsilon \in \left( {\begin{array}{c}[4]\\ k\end{array}}\right) } \det \left( H^\top [\sigma , \upsilon ] \right) \cdot \det \left( \left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) [\upsilon ,\tau ] \right) . \end{aligned} \end{aligned}$$(B.1)The minors \(\det (\left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) [\upsilon ,\tau ])\) which occur in this sum range over all \(k\times k\) minors of the \(4 \times k\) matrix \(\left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) [[4], \tau ].\) This \(4\times k\) matrix has full rank k since if it did not, we could add \(4-k\) additional columns from \(\left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) \) to get a rank-deficient \(4\times 4\) matrix, contradicting our assumption that \(\left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) \) is minor generic. Thus, \(\det \left( \begin{array}{c|c|c}{{\bar{A}}_1}^\top&\cdots&{{\bar{A}}_m}^{\top {}}\end{array}\right) [\upsilon , \tau ] \ne 0\) for some \(\upsilon \in \left( {\begin{array}{c}[4]\\ k\end{array}}\right) .\) Hence, the expressions in (B.1) are not zero, and setting (B.1) to 0 we obtain a hypersurface in \(\text {GL}_4\). Any choice of H lying outside the union of the finitely many hypersurfaces, obtained by varying over all \(k, \sigma , \tau \) yields an arrangement \((A_1 H, \ldots , A_m H)\) satisfying the conclusion.
-
(3)
This statement follows from the first two. \(\square \)
1.2 B.2. Proof of Proposition 2.2, Part 1
Proof
Let \(I = \langle g_1, \ldots , g_s \rangle .\) By definition, \(g_1, \ldots ,g_s\) forming a Gröbner basis with respect to < means that
First, we verify that this monomial ideal in \(R[x_1,\ldots , x_k]\) is radical. To see this, let \(in_<(g_s) = x_{i_1} \cdots x_{i_l}\) and note that
Iterating this argument, we obtain \(in_< (I)\) as an intersection of prime ideals generated by subsets of the variables.
Now, to show that \(I = \langle g_1, \ldots , g_s \rangle \) is radical, suppose that \(f \in \sqrt{I},\) so that \(f^n \in I\) for some positive integer n. We need to argue that \(f \in I\). We have
where the first equality of leading terms uses the fact that R is a domain and the implication holds since \(in_<(I)\) is radical. Thus, there exists \(f_0 \in I \subseteq \sqrt{I}\) such that \(in_< (f_0) = in_< (f).\) Now \(f-f_0 \in \sqrt{I}\) is an element whose leading term is strictly smaller than \(in_<(f).\) Replacing f with \(f-f_0\) and iterating the argument, we obtain \(f_0, \ldots , f_l \in I\) such that \(f = f_0 + \cdots + f_l \in I\). \(\square \)
1.3 B.3. Proof of the Recognition Criterion: Proposition 2.3
Proof
A point \(x \in \mathbb {P}^{n_1} \times \cdots \times \mathbb {P}^{n_k}\) may be represented in homogeneous coordinates by a point \({\hat{x}}\) in the affine space \(\mathbb {C}^{n_1+1} \times \cdots \times \mathbb {C}^{n_k+1}.\) Consider the affine cone
This is an affine variety whose vanishing ideal is precisely the vanishing ideal of X.
Suppose that Conditions 1–3 are satisfied; we must show that \(\langle f_1, \ldots , f_s \rangle \) is the vanishing ideal of X, or equivalently that of \({\hat{X}}.\) Condition 3 and the Nullstellensatz [15, Ch. 4, §2] imply that \(\langle f_1, \ldots , f_s \rangle \) is the vanishing ideal of the affine variety
Moreover, Conditions 2 and 3 together with standard properties of ideal quotients and saturation [15, Ch. 4, §4] imply \(\langle f_1, \ldots , f_s \rangle \) is the vanishing ideal of
Since affine varieties are uniquely determined by their vanishing ideals, it is now enough to observe the following equality, which holds whenever Condition 1 is satisfied:
Conversely, we verify Conditions 1–3 when \(\langle f_1, \ldots , f_s \rangle \) is the vanishing ideal of X:
-
(1)
\( X \subset \mathrm {V}(f_1, \ldots , f_s)\) since each \(f_i\) vanishes on all points of X. On the other hand, we have \(X = \mathrm {V}(g_1, \ldots , g_s)\) for certain homogeneous polynomials \(g_1, \ldots , g_s,\) all of which must be contained in \(\langle f_1, \ldots , f_s \rangle \). If \(f_1,\ldots , f_s\) vanish at a point, so must \(g_1, \ldots , g_s\), and thus \(\mathrm {V}(f_1, \ldots , f_s) \subset X.\)
-
(2)
Let \(f \in \langle f_1, \ldots , f_s \rangle : (\mathfrak {m}_{\mathbf {x}_1} \cap \cdots \cap \mathfrak {m}_{\mathbf {x}_k})^\infty .\) To show \(f \in \langle f_1, \ldots , f_s \rangle ,\) it is enough to show that each of the homogeneous components of f vanish on X, so suppose further that f is homogeneous. If X is empty, then f vanishing on X holds vacuously. Otherwise, for any point in X there exists some monomial of the form
$$\begin{aligned} m(x) = x_{1, i_1} \, \cdots \, x_{k, i_k} \in \mathfrak {m}_{\mathbf {x}_1} \cap \cdots \cap \mathfrak {m}_{\mathbf {x}_k} \end{aligned}$$which does not vanish at that point. Since \(m(x)^n f\) is in the vanishing ideal for some \(n\ge 1,\) we see that that f must vanish at this point.
-
(3)
If \(f^n (x) =0\) for some \(n\ge 1\) and all \(x\in X\), then \(f(x) =0\) for all \(x\in X.\)
\(\square \)
Appendix C. Generators of \(G_{{M^{m,1}_{\mathbf {A}, \mathbf {q}, \mathbf {p}}}}\)
The Gröbner basis \(G_{M_{\mathbf {A},\mathbf {q},\mathbf {p}}^{m,1}}\) of Proposition 5.2 contains elements in degree \(3,4,5,6,7,8,9.\) Here, for completeness, we give explicit formulas for all of them.
Degree 3 3m elements—for \(1 \le i \le m ,\) \(1\le i_1 < i_2 \le 3, \, \det \Big ( \left( \begin{matrix}A_i q&p_i \end{matrix}\right) [ \{i_1, i_2 \} , :] \Big ).\)
Degree 4 m elements—for \(1 \le i \le m , \, \det \left( \begin{matrix}A_i[:,1]&A_i q&p_i \end{matrix}\right) .\)
Degree 5 \(9\, {m \atopwithdelims ()2}\) elements—for \(1 \le i < j \le m ,\) \(1 \le i_1 < i_2 \le 3,\) \(1\le j_1 < j_2 \le 3,\)
Degree 6 \(6\, {m \atopwithdelims ()2}\) elements—for \(1 \le k_1 < k_2 \le 3,\) \(1\le i , j \le m,\) \(i\ne j,\)
Degree 7 \(\left( {\begin{array}{c}m\\ 2\end{array}}\right) \) elements of the form \(q_4\) times a 2-focal, plus an additional \(27 \left( {\begin{array}{c}m\\ 3\end{array}}\right) \) elements—for \(1\le i<j < k \le m,\) \(1\le i_1 < i_2\le 3,\) \(1\le j_1 < j_2 \le 3,\) \(1\le k_1 < k_2 \le 3,\)
Degrees 8 & 9 \(q_4\) times \(27 \left( {\begin{array}{c}m\\ 3\end{array}}\right) \) 3-focals and \(q_4\) times \(81 \left( {\begin{array}{c}m\\ 4\end{array}}\right) \) 4-focals, respectively.
Rights and permissions
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Agarwal, S., Duff, T., Lieblich, M. et al. An Atlas for the Pinhole Camera. Found Comput Math 24, 227–277 (2024). https://doi.org/10.1007/s10208-022-09592-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10208-022-09592-6