pith. sign in

arxiv: 2604.05877 · v1 · submitted 2026-04-07 · 💻 cs.CV · cs.AI

Automatic dental superimposition of 3D intraorals and 2D photographs for human identification

Antonio D. Villegas-Yeguas , Xavier Abreau-Freire , Guillermo R-Garc\'ia , Andrea Valsecchi , Teresa Pinho , Daniel P\'erez-Mongiovi , Oscar Ib\'a\~nez , Oscar Cord\'on This is my paper

Pith reviewed 2026-05-10 19:28 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords dental identification3D-2D superimpositionforensic odontologycamera pose estimationmorphological comparisonhuman identificationcomputer vision
0
0 comments X

The pith

A 3D-2D superimposition method enables automatic objective dental identification by aligning post-mortem scans to ante-mortem photographs.

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

The paper develops two optimization-based techniques to superimpose 3D intraoral scans onto 2D photographs so that dental morphology can be compared automatically. This matters in identification work because many cases lack formal ante-mortem records, making social-media photos the only available images of teeth. One technique pairs landmarks while the other segments the teeth region to estimate camera parameters; both are tested on more than twenty thousand cross-comparisons drawn from 142 samples. The resulting rankings place the correct match at average positions of 1.6 and 1.5, beating chart-based filters and supplying a visual, quantitative score. A sympathetic reader cares because the approach turns an otherwise slow and subjective step into a fast, repeatable computation that can be inspected directly on the overlaid images.

Core claim

By estimating camera parameters through either paired landmarks or teeth-region segmentation, the 3D post-mortem model can be projected to match the viewpoint of an ante-mortem photograph, allowing direct morphological comparison that produces mean ranking positions of 1.6 and 1.5 across 20,164 trials and supplies an objective quantitative score together with the superimposed visualization.

What carries the argument

Camera-parameter optimization that aligns the 3D intraoral scan to the 2D photograph either by landmark pairs or by teeth segmentation so that morphological features can be compared directly.

Load-bearing premise

Real ante-mortem photographs and post-mortem 3D scans contain consistent dental features that the landmark or segmentation optimization can align despite differences in lighting, perspective, camera quality, and dental state.

What would settle it

A collection of real forensic cases in which the correct identity is never ranked among the top five candidates despite successful optimization on controlled data would show that the alignment does not generalize.

Figures

Figures reproduced from arXiv: 2604.05877 by Andrea Valsecchi, Antonio D. Villegas-Yeguas, Daniel P\'erez-Mongiovi, Guillermo R-Garc\'ia, Oscar Cord\'on, Oscar Ib\'a\~nez, Teresa Pinho, Xavier Abreau-Freire.

Figure 1
Figure 1. Figure 1: The three landmark sets proposed. A. First approach: Superimposition using landmarks This proposal is to use the Posest algorithm to solve the P nP + f problem [20], similar to the solution proposed in [15], although in our case we do not have the problem of soft tissue since we are comparing the same bone tissue. With regard to the landmarks to be used, we propose a whole set including landmarks along the… view at source ↗
Figure 2
Figure 2. Figure 2: Examples of occlusion of the upper part (top [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example of the segmentation for one case. In [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example of a 3D mesh segmentation. Above the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of the superimposition of the PT0230 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of the superimposition of the PT0123 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of the superimposition of the PT0046 [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Example of the PDF for the results of the regions [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: CMC curves for the complete dataset. Note that [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗
read the original abstract

Dental comparison is considered a primary identification method, at the level of fingerprints and DNA profiling. One crucial but time-consuming step of this method is the morphological comparison. One of the main challenges to apply this method is the lack of ante-mortem medical records, specially on scenarios such as migrant death at the border and/or in countries where there is no universal healthcare. The availability of photos on social media where teeth are visible has led many odontologists to consider morphological comparison using them. However, state-of-the-art proposals have significant limitations, including the lack of proper modeling of perspective distortion and the absence of objective approaches that quantify morphological differences. Our proposal involves a 3D (post-mortem scan) - 2D (ante-mortem photos) approach. Using computer vision and optimization techniques, we replicate the ante-mortem image with the 3D model to perform the morphological comparison. Two automatic approaches have been developed: i) using paired landmarks and ii) using a segmentation of the teeth region to estimate camera parameters. Both are capable of obtaining very promising results over 20,164 cross comparisons from 142 samples, obtaining mean ranking values of 1.6 and 1.5, respectively. These results clearly outperform filtering capabilities of automatic dental chart comparison approaches, while providing an automatic, objective and quantitative score of the morphological correspondence, easily to interpret and analyze by visualizing superimposed images.

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 two optimization-based computer vision methods for 3D-2D dental superimposition to support forensic human identification: one using paired landmarks and one using teeth-region segmentation to recover camera parameters from ante-mortem 2D photographs relative to post-mortem 3D intraoral scans. Both methods are evaluated on 20,164 cross-comparisons drawn from 142 samples, reporting mean ranking values of 1.6 and 1.5 respectively; these are claimed to outperform automatic dental-chart filtering while supplying an objective, visualizable quantitative score of morphological correspondence.

Significance. If the reported rankings prove robust under uncontrolled real-world conditions (lighting, viewpoint, dental state), the work would provide a practical, automated tool that addresses a documented bottleneck in odontological identification, particularly for cases lacking formal medical records. The scale of the cross-comparison experiment and the explicit production of superimposable images are concrete strengths that could be directly useful to practitioners.

major comments (2)
  1. [Abstract] Abstract: The central performance claims (mean ranks of 1.6 and 1.5) are presented without any description of the camera model, optimization objective, loss function, or ranking procedure. This absence prevents assessment of whether the optimizer is recovering accurate poses or merely exploiting dataset-specific regularities.
  2. [Abstract] Abstract / implied Results: The 142-sample test set is not characterized with respect to the very variations (uncontrolled social-media lighting, arbitrary viewpoints, changes in dental state) that the introduction identifies as the core challenge for prior work. Without this information, the low mean ranks demonstrate only that the methods function under matched conditions, not that they solve the stated identification problem.
minor comments (2)
  1. [Abstract] Abstract: 'specially' should be 'especially'; 'easily to interpret' should be 'easy to interpret'.
  2. [Abstract] Abstract: The phrase 'providing an automatic, objective and quantitative score of the morphological correspondence, easily to interpret and analyze by visualizing superimposed images' is grammatically awkward and should be rephrased for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. We address each major comment below and indicate the revisions we will make to improve the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central performance claims (mean ranks of 1.6 and 1.5) are presented without any description of the camera model, optimization objective, loss function, or ranking procedure. This absence prevents assessment of whether the optimizer is recovering accurate poses or merely exploiting dataset-specific regularities.

    Authors: We agree that the abstract is too concise and omits essential technical context. The manuscript fully specifies the pinhole camera model (including radial distortion), the optimization procedure (gradient-based minimization of either landmark reprojection error or teeth-region overlap), the loss functions, and the ranking method (ordering by final alignment similarity score). To address the concern, we will expand the abstract with a brief description of the optimization-based superimposition and ranking procedure while remaining within length limits. revision: yes

  2. Referee: [Abstract] Abstract / implied Results: The 142-sample test set is not characterized with respect to the very variations (uncontrolled social-media lighting, arbitrary viewpoints, changes in dental state) that the introduction identifies as the core challenge for prior work. Without this information, the low mean ranks demonstrate only that the methods function under matched conditions, not that they solve the stated identification problem.

    Authors: The referee correctly notes that the abstract and results section do not explicitly quantify dataset variation. The 142 samples were collected from real identification cases and include social-media photographs exhibiting natural differences in lighting, viewpoint, and dental state; however, we did not report summary statistics on these factors. We will add a dedicated paragraph in the Experiments section (and reference it from the abstract) that characterizes the observed ranges of viewpoint angles, lighting conditions, and dental variations to substantiate that the evaluation reflects the challenges outlined in the introduction. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical optimization evaluated on held-out comparisons

full rationale

The paper describes two optimization procedures (landmark-pair and teeth-segmentation) that estimate camera parameters to superimpose a post-mortem 3D intraoral scan onto an ante-mortem 2D photograph. Performance is then measured by ranking the correct match among 20,164 cross-comparisons drawn from 142 samples. No derivation step, equation, or claim reduces to a self-definition, a fitted parameter renamed as a prediction, or a load-bearing self-citation chain; the reported mean ranks (1.5–1.6) are direct experimental outcomes rather than algebraic consequences of the method’s own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the method implicitly relies on standard computer vision assumptions such as accurate landmark detection and segmentability of teeth regions, but none are detailed or justified here.

pith-pipeline@v0.9.0 · 5598 in / 1341 out tokens · 59307 ms · 2026-05-10T19:28:41.048378+00:00 · methodology

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

32 extracted references · 32 canonical work pages

  1. [1]

    Forensic odontology in DVI: Current prac- tice and recent advances,

    A. Forrest, “Forensic odontology in DVI: Current prac- tice and recent advances,”Forensic Science Research, vol. 4, no. 4, pp. 316–330, Oct. 2, 2019.DOI: 10.1080/ 20961790.2019.1678710

    work page arXiv 2019

  2. [2]

    Interpol disaster victim identification guide 2023: Annexure 8 - methods of identification,

    A. Cerritelli and R. Anderson, “Interpol disaster victim identification guide 2023: Annexure 8 - methods of identification,” 2023

    work page 2023

  3. [3]

    Swiss DVI at the tsunami disaster: Expect the un- expected,

    M. Perrier, M. Bollmann, A. Girod, and P. Mangin, “Swiss DVI at the tsunami disaster: Expect the un- expected,”Forensic Science International, vol. 159, S30–S32, 2006.DOI: 10.1016/j.forsciint.2006.02.007

    work page doi:10.1016/j.forsciint.2006.02.007 2006

  4. [4]

    Den- tal superimposition: A pilot study for standardising the method,

    D. De Angelis, C. Cattaneo, and M. Grandi, “Den- tal superimposition: A pilot study for standardising the method,”International Journal of Legal Medicine, vol. 121, no. 6, pp. 501–506, Nov. 1, 2007.DOI: 10. 1007/s00414-007-0198-y

    work page 2007

  5. [5]

    GrinLine identification using digital imaging and adobe photoshop,

    S. A. Bollinger, P. C. Brumit, B. A. Schrader, and D. R. Senn, “GrinLine identification using digital imaging and adobe photoshop,”Journal of Forensic Sciences, vol. 54, no. 2, pp. 422–427, 2009.DOI: 10.1111/j.1556-4029. 2008.00971.x

    work page doi:10.1111/j.1556-4029 2009

  6. [6]

    Personal identification through digital photo superim- position of dental profile: A pilot study,

    V . Santoro, F. Mele, F. Introna, and A. De Donno, “Personal identification through digital photo superim- position of dental profile: A pilot study,”The Journal of Forensic Odonto-stomatology, vol. 37, no. 3, pp. 21–26, Dec. 1, 2019

    work page 2019

  7. [7]

    Forensic dental identification using two-dimensional photographs of a smile and three- dimensional dental models: A 2d-3d superimposition method,

    G. V . Reesu, S. M ˆanica, G. F. Revie, N. L. Brown, and P. A. Mossey, “Forensic dental identification using two-dimensional photographs of a smile and three- dimensional dental models: A 2d-3d superimposition method,”Forensic Science International, vol. 313, p. 110 361, Aug. 1, 2020.DOI: 10 . 1016 / j . forsciint . 2020.110361

    work page arXiv 2020

  8. [8]

    Exploring the use of selfies in human identification,

    D. Naidu, A. Franco, and S. M ˆanica, “Exploring the use of selfies in human identification,”Journal of Forensic and Legal Medicine, vol. 85, p. 102 293, Jan. 1, 2022. DOI: 10.1016/j.jflm.2021.102293

    work page doi:10.1016/j.jflm.2021.102293 2022

  9. [9]

    Human identification through smile photographs: Comparison of two methods based on selfies,

    D. R. A. M. De Sousa et al., “Human identification through smile photographs: Comparison of two methods based on selfies,”Journal of Forensic Sciences, vol. 70, no. 3, pp. 1181–1187, 2025.DOI: 10.1111/1556-4029. 70020 10

    work page doi:10.1111/1556-4029 2025

  10. [10]

    A review of 3d/2d registration methods for image-guided interventions,

    P. Markelj, D. Toma ˇzeviˇc, B. Likar, and F. Pernu ˇs, “A review of 3d/2d registration methods for image-guided interventions,”Medical Image Analysis, Computer As- sisted Interventions, vol. 16, no. 3, pp. 642–661, Apr. 1, 2012.DOI: 10.1016/j.media.2010.03.005

    work page doi:10.1016/j.media.2010.03.005 2012

  11. [11]

    From specific-source feature-based to common-source score-based likelihood-ratio systems: Ranking the stars,

    P. Vergeer, “From specific-source feature-based to common-source score-based likelihood-ratio systems: Ranking the stars,”Law, Probability and Risk, vol. 22, no. 1, mgad005, Jan. 1, 2023.DOI: 10 . 1093 / lpr / mgad005

    work page 2023

  12. [12]

    An overview of log likelihood ratio cost in forensic science – where is it used and what values can we ex- pect?

    S. van Lierop, D. Ramos, M. Sjerps, and R. Ypma, “An overview of log likelihood ratio cost in forensic science – where is it used and what values can we ex- pect?”Forensic Science International: Synergy, vol. 8, p. 100 466, Jan. 1, 2024.DOI: 10.1016/j.fsisyn.2024. 100466

    work page doi:10.1016/j.fsisyn.2024 2024

  13. [13]

    ENFSI (european network of forensic science institutes) guideline for evaluative reporting in forensic science,

    C. Champod, A. Biedermann, J. Vuille, S. Willis, and J. De Kinder, “ENFSI (european network of forensic science institutes) guideline for evaluative reporting in forensic science,”Criminal Law and Justice Weekly, vol. 10, no. 180, pp. I–I, Apr. 1, 2016

    work page 2016

  14. [14]

    Smile pho- tograph analysis and its connection with focal length as one of identification methods in forensic anthropol- ogy and odontology,

    M. Mazur, K. G ´orka, and I. A. Aguilera, “Smile pho- tograph analysis and its connection with focal length as one of identification methods in forensic anthropol- ogy and odontology,”Forensic Science International, vol. 335, p. 111 285, Jun. 1, 2022.DOI: 10 . 1016 / j . forsciint.2022.111285

    work page arXiv 2022

  15. [15]

    A robust and efficient method for skull-face overlay in computerized craniofacial superimposition,

    A. Valsecchi, S. Damas, and O. Cord ´on, “A robust and efficient method for skull-face overlay in computerized craniofacial superimposition,”IEEE Transactions on Information Forensics and Security, vol. 13, no. 8, pp. 1960–1974, Aug. 2018.DOI: 10.1109/TIFS.2018. 2806939

    work page doi:10.1109/tifs.2018 1960

  16. [16]

    3d-2d silhouette-based image registration for comparative radiography-based forensic identification,

    O. G ´omez, O. Ib ´a˜nez, A. Valsecchi, O. Cord ´on, and T. Kahana, “3d-2d silhouette-based image registration for comparative radiography-based forensic identification,” Pattern Recognition, vol. 83, pp. 469–480, Nov. 1, 2018. DOI: 10.1016/j.patcog.2018.06.011

    work page doi:10.1016/j.patcog.2018.06.011 2018

  17. [17]

    Damas, O

    S. Damas, O. Cord ´on, and O. Ib ´a˜nez,Handbook on Craniofacial Superimposition: The MEPROCS Project. Cham: Springer International Publishing, 2020.DOI: 10. 1007/978-3-319-11137-7

    work page 2020

  18. [18]

    Zanatta, A

    M. Xu et al., “A critical analysis of image-based camera pose estimation techniques,”Neurocomputing, vol. 570, p. 127 125, Feb. 14, 2024.DOI: 10.1016/j.neucom.2023. 127125

    work page doi:10.1016/j.neucom.2023 2024

  19. [19]

    EPnP: An accurate o(n) solution to the PnP problem,

    V . Lepetit, F. Moreno-Noguer, and P. Fua, “EPnP: An accurate o(n) solution to the PnP problem,”Interna- tional Journal of Computer Vision, vol. 81, no. 2, pp. 155–166, Feb. 1, 2009.DOI: 10.1007/s11263-008- 0152-6

    work page doi:10.1007/s11263-008- 2009

  20. [20]

    Model-based pose esti- mation for rigid objects,

    M. Lourakis and X. Zabulis, “Model-based pose esti- mation for rigid objects,” inComputer Vision Systems, M. Chen, B. Leibe, and B. Neumann, Eds., Berlin, Heidelberg: Springer, 2013, pp. 83–92.DOI: 10.1007/ 978-3-642-39402-7 9

    work page 2013

  21. [21]

    Segmentation-driven 6d object pose estimation,

    Y . Hu, J. Hugonot, P. Fua, and M. Salzmann, “Segmentation-driven 6d object pose estimation,” in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2019, pp. 3380–3389. DOI: 10.1109/CVPR.2019.00350

    work page doi:10.1109/cvpr.2019.00350 2019

  22. [22]

    Periodontal health and its relation- ship with psychological stress: A cross-sectional study,

    M. Macr `ı et al., “Periodontal health and its relation- ship with psychological stress: A cross-sectional study,” Journal of Clinical Medicine, vol. 13, no. 10, May 16, 2024.DOI: 10.3390/jcm13102942

    work page doi:10.3390/jcm13102942 2024

  23. [23]

    In: Proc

    M. Gendreau and J.-Y . Potvin, Eds.,Handbook of Metaheuristics, vol. 272, International Series in Opera- tions Research & Management Science, Cham: Springer International Publishing, 2019.DOI: 10.1007/978- 3- 319-91086-4

    work page doi:10.1007/978- 2019

  24. [24]

    Hybrid mean-variance mapping optimization for solving the IEEE-CEC 2013 competition problems,

    J. L. Rueda and I. Erlich, “Hybrid mean-variance mapping optimization for solving the IEEE-CEC 2013 competition problems,” in2013 IEEE Congress on Evolutionary Computation, Jun. 2013, pp. 1664–1671. DOI: 10.1109/CEC.2013.6557761

    work page doi:10.1109/cec.2013.6557761 2013

  25. [25]

    Comparison of emerging metaheuristic algorithms for optimal hydrothermal system operation,

    M. P. Camargo, J. L. Rueda, I. Erlich, and O. A ˜n´o, “Comparison of emerging metaheuristic algorithms for optimal hydrothermal system operation,”Swarm and Evolutionary Computation, vol. 18, pp. 83–96, Oct. 1, 2014.DOI: 10.1016/j.swevo.2014.04.001

    work page doi:10.1016/j.swevo.2014.04.001 2014

  26. [26]

    Performance analysis of real- coded evolutionary algorithms under a computationally expensive optimization scenario: 3d–2d comparative ra- diography,

    O. G ´omez, O. Ib ´a˜nez, A. Valsecchi, E. Bermejo, D. Molina, and O. Cord ´on, “Performance analysis of real- coded evolutionary algorithms under a computationally expensive optimization scenario: 3d–2d comparative ra- diography,”Applied Soft Computing, vol. 97, p. 106 793, Dec. 1, 2020.DOI: 10.1016/j.asoc.2020.106793

    work page doi:10.1016/j.asoc.2020.106793 2020

  27. [27]

    Skeleton-ID: AI-driven human identification,

    A. Valsecchi et al., “Skeleton-ID: AI-driven human identification,” in2023 IEEE Conference on Artificial Intelligence (CAI), Jun. 2023, pp. 278–279.DOI: 10. 1109/CAI54212.2023.00124

    work page arXiv 2023

  28. [28]

    Algorithms for coplanar camera calibration,

    C. Chatterjee and V . P. Roychowdhury, “Algorithms for coplanar camera calibration,”Machine Vision and Applications, vol. 12, no. 2, pp. 84–97, Aug. 1, 2000. DOI: 10.1007/s001380050127

    work page doi:10.1007/s001380050127 2000

  29. [29]

    Computerized dental comparison: A critical review of dental coding and ranking algorithms used in victim identification,

    B. J. Adams and K. W. Aschheim, “Computerized dental comparison: A critical review of dental coding and ranking algorithms used in victim identification,” Journal of Forensic Sciences, vol. 61, no. 1, pp. 76–86, Jan. 1, 2016, MAG ID: 1943029493.DOI: 10 . 1111 / 1556-4029.12909

    work page arXiv 2016

  30. [30]

    A. D. Villegas-Yeguas et al.,On the use of aggregation operators to improve human identification using dental records, Mar. 24, 2026.DOI: 10 . 48550 / arXiv. 2603 . 23003 arXiv: 2603.23003[cs]

    work page arXiv 2026

  31. [31]

    Evidence evaluation in craniofacial superimposition using likelihood ratios,

    P. Mart ´ınez-Moreno, A. Valsecchi, P. Mesejo,´O. Ib´a˜nez, and S. Damas, “Evidence evaluation in craniofacial superimposition using likelihood ratios,”Information Fusion, vol. 111, p. 102 489, Nov. 1, 2024.DOI: 10. 1016/j.inffus.2024.102489

    work page arXiv 2024

  32. [32]

    S. Z. Li and A. K. Jain, Eds.,Handbook of Face Recognition, London: Springer, 2011.DOI: 10 . 1007 / 978-0-85729-932-1

    work page 2011