pith. sign in

arxiv: 2605.00934 · v2 · pith:4MVICZEYnew · submitted 2026-05-01 · 💻 cs.LG · cs.CV· stat.ML

Structured Analytic Coherent Point Drift for Non-Rigid Point Set Registration

Wei Feng , Haiyong Zheng This is my paper

Pith reviewed 2026-05-19 16:37 UTC · model grok-4.3

classification 💻 cs.LG cs.CVstat.ML
keywords point set registrationnon-rigid registrationcoherent point driftanalytic mappingsdeformation estimationGaussian mixture modelcontinuation strategyunsupervised registration
0
0 comments X

The pith

Coupling CPD posteriors with structured analytic mappings makes non-rigid registration scale with function order rather than point count.

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

The paper introduces Analytic-CPD, which keeps the Gaussian mixture model for finding point correspondences from standard Coherent Point Drift but replaces the deformation step with estimation of structured analytic mappings. This change makes the number of coefficients in the deformation model depend on the dimension of space and the chosen order of the analytic functions instead of the total number of points being moved. A sympathetic reader would care because large point sets from scans or images become feasible to register without heavy computation, and the method allows gradually increasing the complexity of the deformation as the matches become more certain.

Core claim

By coupling the Gaussian-mixture posterior mechanism of CPD with Structured Analytic Mappings (SAM), the method obtains a deformation model whose coefficient dimension is governed by the ambient dimension and analytic order rather than by the number of moving points. Deformation estimation is organized over an interpretable hierarchy of analytic function spaces so that the analytic order can be increased progressively as posterior correspondences become more reliable.

What carries the argument

Structured Analytic Mappings (SAM) lift the M-step from point-indexed kernel displacement estimation to structured analytic mapping estimation, organizing deformation over a hierarchy of analytic function spaces.

If this is right

  • Registration becomes computationally lighter for large point sets since the number of coefficients depends on ambient dimension and analytic order rather than point count.
  • Progressive increase in analytic order allows low-order maps to stabilize correspondences first and higher-order modes to refine nonlinear residuals later.
  • Complexity control during registration improves because the analytic order can be chosen or ramped explicitly.
  • The framework applies across controlled model-matched cases, smooth model-mismatch cases, and registered human-shape data with favorable accuracy-efficiency performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same posterior-plus-analytic-mapping pattern could be tested on other mixture-model registration algorithms to reduce their deformation cost.
  • Different analytic bases could be swapped in to match expected deformation classes, such as low-order polynomials for global bends.
  • The hierarchy might support incremental registration pipelines where early low-order stages run quickly to seed later stages.

Load-bearing premise

An increasing-degree continuation strategy with decreasing stage lengths will first stabilize the posterior correspondence structure before higher-order modes refine nonlinear residuals.

What would settle it

A direct comparison of registration error and runtime on a dataset with thousands of points, checking whether Analytic-CPD maintains or improves accuracy over standard CPD while using far fewer coefficients and showing progressive refinement with order increases.

Figures

Figures reproduced from arXiv: 2605.00934 by Haiyong Zheng, Wei Feng.

Figure 1
Figure 1. Figure 1: Summary of the proposed Analytic-CPD algorithm. 3.8 Algorithm Summary The proposed method follows the EM-style structure of CPD. The E-step computes the posterior correspondence matrix from the current moving configuration. The M-step condenses the posterior probabilities into weighted soft targets, estimates a structured analytic mapping, updates the moving point set compositionally, and re-estimates the … view at source ↗
Figure 2
Figure 2. Figure 2: Two-dimensional large-deformation registration examples using Analytic-CPD.Red points denote the fixed point set, and green points denote [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Initial configurations and error–time curves for two representative [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Three-dimensional large-deformation registration examples using Analytic-CPD. Red points denote the fixed point set, and green points [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Initial configurations for the multi-model and multi-seed experi [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: Three-dimensional registration under large smooth deformation [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Initial configurations of the nine FAUST registered human-shape [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
read the original abstract

Coherent Point Drift (CPD) is a representative probabilistic framework for unsupervised non-rigid point set registration. Its standard non-rigid M-step, however, relies on a point-indexed Gaussian-kernel system whose size grows with the number of moving points, making deformation estimation computationally heavy for large point sets and difficult to control in complexity during registration. To address these limitations, we propose Analytic-CPD, a new unsupervised non-rigid registration framework that gives CPD a structured analytic reformulation. Analytic-CPD preserves the CPD posterior correspondence layer, but lifts the M-step from point-indexed kernel displacement estimation to structured analytic mapping estimation. By coupling the Gaussian-mixture posterior mechanism of CPD with Structured Analytic Mappings (SAM), the method obtains a deformation model whose coefficient dimension is governed by the ambient dimension and analytic order rather than by the number of moving points. More importantly, deformation estimation is organized over an interpretable hierarchy of analytic function spaces, so the analytic order can be increased progressively as posterior correspondences become more reliable. We implement this idea through an increasing-degree continuation strategy with decreasing stage lengths: low-order analytic maps first stabilize the posterior correspondence structure, while higher-order modes later refine nonlinear residual deformation. Experiments on controlled model-matched, smooth model-mismatch, and registered human-shape data demonstrate the effectiveness and favorable accuracy--efficiency performance of Analytic-CPD.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes Analytic-CPD, a reformulation of Coherent Point Drift (CPD) for unsupervised non-rigid point set registration. It preserves the CPD Gaussian-mixture posterior correspondence layer while replacing the standard point-indexed Gaussian-kernel M-step with estimation over Structured Analytic Mappings (SAM). This yields a deformation model whose coefficient dimension depends on ambient dimension and analytic order rather than the number of moving points. Registration proceeds via an increasing-degree continuation strategy with decreasing stage lengths, in which low-order maps first stabilize correspondences before higher-order terms refine residuals. Experiments on model-matched, model-mismatch, and human-shape data are reported to show favorable accuracy-efficiency trade-offs.

Significance. If the central claims hold, the work offers a principled route to scalable non-rigid registration by decoupling deformation complexity from point-set cardinality. The hierarchical analytic-function-space organization supplies an interpretable control mechanism absent from kernel-based CPD, and the dimension reduction could materially improve applicability to large point clouds in computer vision and medical imaging.

major comments (2)
  1. [Implementation approach / continuation strategy] Implementation section (increasing-degree continuation strategy with decreasing stage lengths): the central dimension-reduction claim is realized only if low-order analytic maps first produce a stable Gaussian-mixture posterior before higher-order coefficients are activated. The manuscript supplies neither a convergence argument for this ordering nor an ablation that isolates the effect of deliberately shortened later stages on posterior stability. If the ordering fails, the low-dimensional model can lock into poor correspondences, rendering the claimed independence from point-set size practically irrelevant.
  2. [Abstract and §3] Abstract and §3 (SAM substitution into CPD objective): the claim that the posterior layer is preserved while the M-step is lifted to structured analytic mapping estimation is stated without explicit derivation steps, error analysis, or the substituted objective function. Without these equations it is impossible to verify that the analytic mapping does not introduce self-referential parameters or alter the fixed-point properties of the original CPD EM procedure.
minor comments (2)
  1. [Notation and preliminaries] Notation for the analytic order and coefficient vector should be introduced once and used consistently; current usage mixes “analytic order” with “degree” without a clear mapping to the function-space hierarchy.
  2. [Experiments] The experimental section would benefit from an explicit statement of the maximum analytic order used in each dataset and the resulting coefficient dimension, to allow direct comparison with the point-set size.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive feedback and the positive evaluation of the potential impact of Analytic-CPD. We address the major comments point by point below, and we will incorporate revisions to strengthen the manuscript accordingly.

read point-by-point responses

  1. Referee: [Implementation approach / continuation strategy] Implementation section (increasing-degree continuation strategy with decreasing stage lengths): the central dimension-reduction claim is realized only if low-order analytic maps first produce a stable Gaussian-mixture posterior before higher-order coefficients are activated. The manuscript supplies neither a convergence argument for this ordering nor an ablation that isolates the effect of deliberately shortened later stages on posterior stability. If the ordering fails, the low-dimensional model can lock into poor correspondences, rendering the claimed independence from point-set size practically irrelevant.

    Authors: The referee correctly notes that the current manuscript does not provide a formal convergence argument for the continuation strategy. We believe such an argument would be valuable but is challenging to derive for this EM-based procedure with changing parameter spaces. Instead, the strategy is justified by the progressive refinement property of analytic function spaces, where lower degrees provide a stable initialization for the posterior. The decreasing stage lengths prioritize stability in early stages. To empirically validate this, we will add an ablation study in the revised manuscript comparing the proposed decreasing stage lengths against constant or increasing lengths, measuring posterior stability (e.g., via entropy or correspondence accuracy) and final registration error. This will demonstrate that the ordering helps avoid poor local optima and supports the dimension reduction claim. revision: partial

  2. Referee: [Abstract and §3] Abstract and §3 (SAM substitution into CPD objective): the claim that the posterior layer is preserved while the M-step is lifted to structured analytic mapping estimation is stated without explicit derivation steps, error analysis, or the substituted objective function. Without these equations it is impossible to verify that the analytic mapping does not introduce self-referential parameters or alter the fixed-point properties of the original CPD EM procedure.

    Authors: We agree that the manuscript would benefit from more explicit mathematical details. In the revision, we will expand §3 to include the full derivation of the substituted objective function. The CPD objective is the expected log-likelihood under the posterior probabilities, and substituting the analytic mapping means parameterizing the displacement as a sum of analytic basis functions up to a given order, with coefficients solved via a linear system whose size depends on the order and dimension, not the number of points. We will show that the posterior computation remains identical to standard CPD (depending on current deformation), and the M-step is still a maximization of the same expected log-likelihood but in the coefficient space. This preserves the EM fixed-point properties, as the analytic mapping is a reparameterization that does not introduce self-reference. An error analysis bounding the approximation error for finite order will also be added. revision: yes

standing simulated objections not resolved
  • A rigorous convergence proof for the increasing-degree continuation strategy with decreasing stage lengths

Circularity Check

0 steps flagged

No circularity: dimension reduction follows by direct substitution of SAM for point-indexed kernels

full rationale

The paper's central derivation replaces the CPD M-step's point-indexed Gaussian kernel system with Structured Analytic Mappings (SAM) while preserving the Gaussian-mixture posterior. The resulting claim that coefficient dimension is governed by ambient dimension and analytic order (rather than number of moving points) follows immediately from the definition of SAM as a fixed-order analytic function space; it is not obtained by fitting parameters to data or by any self-referential equation that equates output to input. The increasing-degree continuation strategy with decreasing stage lengths is presented purely as an implementation heuristic for progressive stabilization, without any assertion that it is mathematically derived from the model or that it constitutes a prediction equivalent to its own assumptions. No self-citations are invoked to establish uniqueness or to smuggle in ansatzes, and the provided text contains no equations showing that a claimed result reduces to a fitted quantity or prior self-result by construction. The derivation therefore introduces independent structure rather than renaming or circularly re-deriving its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the existence of Structured Analytic Mappings that can be coupled to the CPD posterior while preserving its properties, plus the assumption that progressive analytic order increase improves stability. No explicit free parameters or invented entities beyond the new mapping class are named in the abstract.

axioms (1)
  • domain assumption The CPD posterior correspondence layer remains unchanged when the M-step is replaced by structured analytic mapping estimation.
    Stated directly in the abstract as the preserved component of the original CPD framework.
invented entities (1)
  • Structured Analytic Mappings (SAM) no independent evidence
    purpose: To provide a deformation model whose coefficient dimension depends on ambient dimension and analytic order rather than point count.
    Introduced as the core replacement for the point-indexed Gaussian-kernel system.

pith-pipeline@v0.9.0 · 5774 in / 1341 out tokens · 43489 ms · 2026-05-19T16:37:08.567016+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages

  1. [1]

    A Method for Registration of 3-D Shapes,

    P . Besl and N. Mckay, “A Method for Registration of 3-D Shapes,” IEEE T rans. Pattern Anal. Mach. Intell., vol. 1611, pp. 586–606, 1992

    work page 1992

  2. [2]

    Non-Rigid Point Set Registration: Coherent Point Drift,

    A. Myronenko, X. Song, and M. Carreira-Perpi ˜n´an, “Non-Rigid Point Set Registration: Coherent Point Drift,” inProc. Adv. Neural Inform. Process. Syst., vol. 19, 2006, pp. 1009–1016

    work page 2006

  3. [3]

    Point set registration: Coherent point drift,

    A. Myronenko and X. Song, “Point set registration: Coherent point drift,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 32, no. 12, pp. 2262–2275, 2010

    work page 2010

  4. [4]

    Structured analytic mappings for point set registration,

    W. Feng, T. Wei, and H. Zheng, “Structured analytic mappings for point set registration,”SIAM J. Imaging Sci., 2026, accepted; arXiv:2602.16753

    work page arXiv 2026

  5. [5]

    Iterative point matching for registration of free-form curves and surfaces,

    Z. Zhang, “Iterative point matching for registration of free-form curves and surfaces,”Int. J. Comput. Vis., vol. 13, no. 2, pp. 119– 152, 1994

    work page 1994

  6. [6]

    ICP Registration using Invariant Features,

    G. Sharp, S. Lee, and D. Wehe, “ICP Registration using Invariant Features,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 24, no. 1, pp. 90–102, 2002

    work page 2002

  7. [7]

    Robust Euclidean Alignment of 3D Point Sets: The Trimmed Iterative Closest Point Algorithm,

    D. Chetverikov, D. Stepanov, and P . Krsek, “Robust Euclidean Alignment of 3D Point Sets: The Trimmed Iterative Closest Point Algorithm,”Image Vision Comput., vol. 23, no. 3, pp. 299–309, 2005

    work page 2005

  8. [8]

    Generalized-ICP,

    A. Segal, D. Haehnel, and S. Thrun, “Generalized-ICP,” inProc. Robot.: Sci. Syst., vol. 2, 2009, p. 435

    work page 2009

  9. [9]

    Linear Least-Squares Optimization for Point-To-Plane ICP Surface Registration,

    K.-L. Low, “Linear Least-Squares Optimization for Point-To-Plane ICP Surface Registration,”Chapel Hill, Univ. North Carolina, vol. 4, no. 10, pp. 1–3, 2004

    work page 2004

  10. [10]

    Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration,

    J. Yang, H. Li, D. Campbell, and Y. Jia, “Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 38, no. 11, pp. 2241–2254, 2015

    work page 2015

  11. [11]

    Fast and Robust Iterative Closest Point,

    J. Zhang, Y. Yao, and B. Deng, “Fast and Robust Iterative Closest Point,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 44, no. 7, pp. 3450–3466, 2021

    work page 2021

  12. [12]

    KSS-ICP: Point Cloud Registration Based on Kendall Shape Space,

    C. Lv, W. Lin, and B. Zhao, “KSS-ICP: Point Cloud Registration Based on Kendall Shape Space,”IEEE T rans. Image Process., vol. 32, pp. 1681–1693, 2023

    work page 2023

  13. [13]

    Affine Iterative Closest Point Algorithm for Point Set Registration,

    D. Shaoyi, Z. Nanning, Y. Shihui, and L. Jianyi, “Affine Iterative Closest Point Algorithm for Point Set Registration,”Pattern Recog- nit. Lett., vol. 31, no. 9, pp. 791–799, 2010

    work page 2010

  14. [14]

    A New Affine Registration Algorithm for Matching 2D Point Sets,

    J. Ho, M.-H. Yang, A. Rangarajan, and B. Vemuri, “A New Affine Registration Algorithm for Matching 2D Point Sets,” inProc. IEEE Winter Conf. Appl. Comput. Vision, 2007, pp. 25–25

    work page 2007

  15. [15]

    A Global Optimisation Method for Robust Affine Registration of Brain Images,

    M. Jenkinson and S. Smith, “A Global Optimisation Method for Robust Affine Registration of Brain Images,”Med. Image Anal., vol. 5, no. 2, pp. 143–156, 2001

    work page 2001

  16. [16]

    Affine Registration of Point Sets using ICP and ICA,

    S. Du, N. Zheng, G. Meng, and Z. Yuan, “Affine Registration of Point Sets using ICP and ICA,”IEEE Signal Process. Lett., vol. 15, pp. 689–692, 2008

    work page 2008

  17. [17]

    Optimal Step Nonrigid ICP Algorithms for Surface Registration,

    A. Brian, R. Sami, and V . Thomas, “Optimal Step Nonrigid ICP Algorithms for Surface Registration,” inProc. IEEE Conf. Comput. Vis. Pattern Recognit., 2007, pp. 1063–6919. 22

    work page 2007

  18. [18]

    Embedded deformation for shape manipulation,

    R. W. Sumner, J. Schmid, and M. Pauly, “Embedded deformation for shape manipulation,”ACM T ransactions on Graphics, vol. 26, no. 3, p. 80, 2007

    work page 2007

  19. [19]

    Global correspondence op- timization for non-rigid registration of depth scans,

    H. Li, R. W. Sumner, and M. Pauly, “Global correspondence op- timization for non-rigid registration of depth scans,” inComputer graphics forum, vol. 27, no. 5. Wiley Online Library, 2008, pp. 1421–1430

    work page 2008

  20. [20]

    A new point matching algorithm for non-rigid registration,

    H. Chui and A. Rangarajan, “A new point matching algorithm for non-rigid registration,”Comput. Vis. Image Underst., vol. 89, no. 2–3, pp. 114–141, 2003

    work page 2003

  21. [21]

    Principal Warps: Thin-plate splines and The Decomposition of Deformations,

    F. L. Bookstein, “Principal Warps: Thin-plate splines and The Decomposition of Deformations,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 11, no. 6, pp. 567–585, 1989

    work page 1989

  22. [22]

    A Feature Registration Framework using Mixture Models,

    H. Chui and A. Rangarajan, “A Feature Registration Framework using Mixture Models,” inProc. IEEE Workshop Math. Methods Biomed. Image Anal., 2000, pp. 190–197

    work page 2000

  23. [23]

    A Robust Algorithm for Point Set Registra- tion Using Mixture of Gaussians,

    B. Jian and V . Baba C, “A Robust Algorithm for Point Set Registra- tion Using Mixture of Gaussians,” inProc. IEEE Int. Conf. Comput. Vis., vol. 2, 2005, pp. 1246–1251

    work page 2005

  24. [24]

    Robust Point Set Registration using Gaussian Mixture Models,

    B. Jian and B. C. Vemuri, “Robust Point Set Registration using Gaussian Mixture Models,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 33, no. 8, pp. 1633–1645, 2010

    work page 2010

  25. [25]

    Rigid and articulated point registration with expectation conditional maximization,

    R. Horaud, F. Forbes, M. Yguel, G. Dewaele, and J. Zhang, “Rigid and articulated point registration with expectation conditional maximization,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 33, no. 3, pp. 587–602, 2011

    work page 2011

  26. [26]

    Joint alignment of multiple point sets with batch and incremental expectation-maximization,

    G. D. Evangelidis and R. Horaud, “Joint alignment of multiple point sets with batch and incremental expectation-maximization,” IEEE T rans. Pattern Anal. Mach. Intell., vol. 40, no. 6, pp. 1397–1410, 2017

    work page 2017

  27. [27]

    Gen- eralised coherent point drift for group-wise registration of multi- dimensional point sets,

    N. Ravikumar, A. Gooya, A. F. Frangi, and Z. A. Taylor, “Gen- eralised coherent point drift for group-wise registration of multi- dimensional point sets,” inMedical Image Computing and Computer Assisted Intervention – MICCAI 2017. Springer, 2017, pp. 309–316

    work page 2017

  28. [28]

    Generalized coherent point drift with multi-variate gaussian distribution and watson distribution,

    Z. Min, J. Liu, L. Liu, and M. Q.-H. Meng, “Generalized coherent point drift with multi-variate gaussian distribution and watson distribution,”IEEE Robotics and Automation Letters, vol. 6, no. 4, pp. 6749–6756, 2021

    work page 2021

  29. [29]

    A Bayesian Formulation of Coherent Point Drift,

    O. Hirose, “A Bayesian Formulation of Coherent Point Drift,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 43, pp. 2269–2286, 2021

    work page 2021

  30. [30]

    Geodesic-Based Bayesian Coherent Point Drift,

    O. Hirose, “Geodesic-Based Bayesian Coherent Point Drift,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 45, no. 5, pp. 5816–5832, 2022

    work page 2022

  31. [31]

    Acceleration of Non-Rigid Point Set Registration with Downsampling and Gaussian Process Regression,

    O. Hirose, “Acceleration of Non-Rigid Point Set Registration with Downsampling and Gaussian Process Regression,”IEEE T rans. Pattern Anal. Mach. Intell., vol. 43, no. 8, pp. 2858–2865, 2020

    work page 2020

  32. [32]

    Correspondence-free nonrigid point set registration using unsu- pervised clustering analysis,

    M. Zhao, J. Jiang, L. Ma, S. Xin, G. Meng, and D.-M. Yan, “Correspondence-free nonrigid point set registration using unsu- pervised clustering analysis,” inIEEE Conf. Comput. Vis. Pattern Recog., 2024

    work page 2024

  33. [33]

    Filterreg: Robust and efficient probabilis- tic point-set registration using gaussian filter and twist param- eterization,

    W. Gao and R. Tedrake, “Filterreg: Robust and efficient probabilis- tic point-set registration using gaussian filter and twist param- eterization,” inIEEE Conf. Comput. Vis. Pattern Recog., 2019, pp. 11 095–11 104

    work page 2019

  34. [34]

    Non-rigid point set registration by preserving global and local structures,

    J. Ma, J. Zhao, and A. L. Yuille, “Non-rigid point set registration by preserving global and local structures,”IEEE T rans. Image Process., vol. 25, no. 1, pp. 53–64, 2015

    work page 2015

  35. [35]

    Faust: Dataset and evaluation for 3d mesh registration,

    F. Bogo, J. Romero, M. Loper, and M. J. Black, “Faust: Dataset and evaluation for 3d mesh registration,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014, pp. 3794–3801

    work page 2014

  36. [36]

    Zippered polygon meshes from range images,

    G. Turk and M. Levoy, “Zippered polygon meshes from range images,” inProceedings of SIGGRAPH 1994, 1994, pp. 311–318

    work page 1994

  37. [37]

    Medshapenet–a large-scale dataset of 3d medical shapes for computer vision,

    J. Li, Z. Zhou, J. Yang, A. Pepe, C. Gsaxner, G. Luijten, C. Qu, T. Zhang, X. Chen, W. Liet al., “Medshapenet–a large-scale dataset of 3d medical shapes for computer vision,”Biomedical Engineer- ing/Biomedizinische T echnik, vol. 70, no. 1, pp. 71–90, 2025

    work page 2025