pith. sign in

arxiv: 2601.02318 · v2 · submitted 2026-01-05 · 💻 cs.CV

Fusion2Print: Deep Flash-Non-Flash Fusion for Contactless Fingerprint Matching

Roja Sahoo , Anoop Namboodiri This is my paper

Pith reviewed 2026-05-16 17:33 UTC · model grok-4.3

classification 💻 cs.CV
keywords contactless fingerprintsflash non-flash fusionridge enhancementbiometric matchingattention networkU-Netdeep embeddings
0
0 comments X

The pith

Fusing paired flash and non-flash contactless fingerprints with an attention network produces clearer ridges and achieves 0.999 AUC matching.

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

The paper introduces Fusion2Print to solve degraded ridge visibility in contactless fingerprint images caused by uneven lighting, skin discoloration, and reflections. It builds a custom paired dataset and fuses flash captures, which hold ridge detail but add noise, with non-flash captures, which reduce noise but lower contrast, through manual subtraction plus an attention-based network and U-Net module. The result is an enhanced grayscale image whose deep embeddings sit in a single space usable for both contactless and contact-based prints. If correct, this would let contactless systems reach recognition performance that matches or exceeds traditional contact methods while avoiding hygiene and pressure problems.

Core claim

Systematic capture and fusion of paired flash-non-flash contactless fingerprints isolates ridge-preserving signals; an attention-based fusion network emphasizes useful channels while a U-Net produces an optimally weighted grayscale image, after which a deep embedding model generates discriminative representations compatible with both contactless and contact-based fingerprints for verification.

What carries the argument

The attention-based fusion network that integrates flash and non-flash modalities by weighting informative channels and suppressing noise, followed by the U-Net enhancement module.

If this is right

  • Produces grayscale images with measurably higher ridge clarity than either flash or non-flash alone.
  • Achieves AUC of 0.999 and EER of 1.12 percent, outperforming Verifinger and DeepPrint on the FNF Database.
  • Generates embeddings that require no domain-specific retraining to work with both contactless and contact-based fingerprints.
  • Removes the need for physical contact, thereby eliminating pressure artifacts and hygiene risks.

Where Pith is reading between the lines

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

  • The same paired-capture fusion could be tested on other surface-detail biometrics such as palm or finger-vein imaging.
  • A lightweight version of the network might run on mobile hardware for on-device contactless enrollment.
  • Widespread use would lower disease transmission risk at high-traffic biometric checkpoints.

Load-bearing premise

Paired flash and non-flash captures can be fused to isolate ridge-preserving signals without introducing new artifacts that harm matching accuracy.

What would settle it

On a held-out set of paired flash-non-flash images, the fused embeddings yield EER above 1.12 percent or visibly lower ridge clarity than the better of the two single captures.

Figures

Figures reproduced from arXiv: 2601.02318 by Anoop Namboodiri, Roja Sahoo.

Figure 1
Figure 1. Figure 1: Overview of proposed fingerprint acquisition and enhancement framework: (a) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed Fusion2Print (F2P) framework. Aligned flash ( [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Channel-wise RGB local contrast of Iflash, Idiff, and Ifuse [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of ridge-quality scores across input types. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: ROC and FRRvsFAR curves for the best TripletDistilNet model trained on Ienh. (a) Contact-based Sample (b) Contactless Sample [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Activation maps of TripletDistilNet before and after fine-tuning, highlighting higher ridge- and core-focused attention (red) across contact and contactless domains. adaptation, especially in early layers, without embedding drift, supporting im￾proved cross-domain generalization (see [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: ROC and FRRvsFAR curves for DeepPrint baseline and Fusion2Print pipeline [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Comparison of DeepPrint Baseline and F2P representations. (a),(b): t￾SNE embedding space plots show DeepPrint’s lower intra-subject consistency (more red links), while F2P achieves more stable features and higher verification accuracy. (c),(d): Pairwise embedding distance heatmaps show F2P with fewer off-diagonal matches, indicating stronger identity discrimination. vated the U-NetEnhancer for channel-bala… view at source ↗
read the original abstract

Contactless fingerprint recognition offers a hygienic and convenient alternative to contact-based systems, enabling rapid acquisition without latent prints, pressure artifacts, or hygiene risks. However, contactless images often show degraded ridge clarity due to illumination variation, subcutaneous skin discoloration, and specular reflections. Flash captures preserve ridge detail but introduce noise, whereas non-flash captures reduce noise but lower ridge contrast. We propose Fusion2Print (F2P), the first framework to systematically capture and fuse paired flash-non-flash contactless fingerprints. We construct a custom paired dataset, FNF Database, and perform manual flash-non-flash subtraction to isolate ridge-preserving signals. A lightweight attention-based fusion network also integrates both modalities, emphasizing informative channels and suppressing noise, and then a U-Net enhancement module produces an optimally weighted grayscale image. Finally, a deep embedding model with cross-domain compatibility, generates discriminative and robust representations in a unified embedding space compatible with both contactless and contact-based fingerprints for verification. F2P enhances ridge clarity and achieves superior recognition performance (AUC=0.999, EER=1.12%) over single-capture baselines (Verifinger, DeepPrint).

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 / 1 minor

Summary. The paper proposes Fusion2Print (F2P), a framework for fusing paired flash and non-flash contactless fingerprint images. It constructs a custom FNF Database, applies manual flash-non-flash subtraction to isolate ridge signals, uses a lightweight attention-based fusion network to emphasize informative channels, and employs a U-Net enhancement module to produce grayscale images. A deep embedding model then generates representations in a unified space compatible with both contactless and contact-based fingerprints, reporting AUC=0.999 and EER=1.12% that outperform single-capture baselines such as Verifinger and DeepPrint.

Significance. If the cross-domain compatibility claim holds, the work offers a practical advance in contactless biometrics by addressing illumination variation and specular reflections through paired capture fusion, potentially enabling hygienic, high-accuracy systems that integrate with existing contact-based matchers without retraining. The empirical pipeline and custom dataset construction provide a concrete starting point for further fusion research in fingerprint imaging.

major comments (2)
  1. [Abstract] Abstract: The headline metrics (AUC=0.999, EER=1.12%) are presented without any statement of dataset size, number of subjects, capture pairs, or cross-validation protocol, which is load-bearing for assessing whether the reported performance reflects generalization rather than dataset-specific tuning.
  2. [Method (embedding model section)] Method (embedding model section): The assertion of a 'unified embedding space' compatible with contact-based fingerprints 'without domain-specific retraining' is unsupported by quantitative evidence such as cross-domain EER on a contact dataset, embedding overlap metrics, or an ablation that isolates the contribution of the fusion step; this directly underpins the central claim of cross-domain applicability.
minor comments (1)
  1. [Abstract] Abstract: The phrase 'manual flash-non-flash subtraction' is introduced without clarifying whether this is a fixed preprocessing operation or learned; a brief equation or pseudocode would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and have revised the manuscript to improve clarity and strengthen the supporting evidence for our claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The headline metrics (AUC=0.999, EER=1.12%) are presented without any statement of dataset size, number of subjects, capture pairs, or cross-validation protocol, which is load-bearing for assessing whether the reported performance reflects generalization rather than dataset-specific tuning.

    Authors: We agree that these details are important for contextualizing the headline metrics. In the revised manuscript, we have updated the abstract to include the size of the FNF Database, the number of subjects and paired captures, and the cross-validation protocol, which were already reported in the experimental section. revision: yes

  2. Referee: [Method (embedding model section)] Method (embedding model section): The assertion of a 'unified embedding space' compatible with contact-based fingerprints 'without domain-specific retraining' is unsupported by quantitative evidence such as cross-domain EER on a contact dataset, embedding overlap metrics, or an ablation that isolates the contribution of the fusion step; this directly underpins the central claim of cross-domain applicability.

    Authors: We acknowledge the need for more explicit quantitative support for the cross-domain compatibility claim. While the original manuscript shows that the same embedding model is applied to both fused contactless images and contact-based images without retraining, we agree that additional metrics would strengthen this. In the revision, we have added cross-domain EER results on a contact-based dataset and an ablation isolating the fusion step's contribution. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical pipeline with no self-referential derivations or load-bearing self-citations

full rationale

The paper describes an empirical ML construction: paired flash/non-flash dataset (FNF Database), manual subtraction, attention-based fusion network, U-Net enhancement, and a deep embedding model asserted to produce a unified space. No equations, derivations, or parameter-fitting steps are shown that reduce the reported AUC/EER to quantities defined by the inputs themselves. No self-citation chains, uniqueness theorems, or ansatzes imported from prior author work appear in the provided text. The performance numbers are presented as measured outcomes on the custom dataset rather than predictions forced by construction. Cross-domain compatibility is an empirical claim, not a mathematical reduction. This is a standard non-circular empirical paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claim rests on the unstated assumption that flash and non-flash images contain complementary ridge information that survives fusion without domain shift, plus the existence of a paired dataset whose collection protocol is not detailed.

pith-pipeline@v0.9.0 · 5499 in / 1070 out tokens · 107676 ms · 2026-05-16T17:33:11.515326+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

41 extracted references · 41 canonical work pages · 1 internal anchor

  1. [1]

    https://docs.expo.dev/versions/latest/sdk/ camera/

    Camera | expo documentation. https://docs.expo.dev/versions/latest/sdk/ camera/

    work page

  2. [2]

    Egyptian Informatics Journal14(1), 15–25 (2013)

    Bahgat, G., Khalil, A., Abdel Kader, N., Mashali, S.: Fast and accurate algorithm for core point detection in fingerprint images. Egyptian Informatics Journal14(1), 15–25 (2013). https://doi.org/https://doi.org/10.1016/j.eij.2013.01.002

    work page doi:10.1016/j.eij.2013.01.002 2013

  3. [3]

    (01 2011)

    Bausinger, O., Tabassi, E.: Fingerprint sample quality metric nfiq 2.0. (01 2011)

    work page 2011

  4. [4]

    In: DICTA 2007

    Chen, W., Gao, Y.: A minutiae-based fingerprint matching algorithm using phase correlation. In: DICTA 2007. IEEE (Dec 2007)

    work page 2007

  5. [5]

    In: CVPRW

    Chopra, S., Malhotra, A., Vatsa, M., Singh, R.: Unconstrained fingerphoto database. In: CVPRW. IEEE (Jun 2018)

    work page 2018

  6. [6]

    Journal of Cybersecurity and Privacy2(3), 714–730 (2022)

    Chowdhury, A.M.M., Imtiaz, M.H.: Contactless fingerprint recognition using deep learning-a systematic review. Journal of Cybersecurity and Privacy2(3), 714–730 (2022). https://doi.org/10.3390/jcp2030036

    work page doi:10.3390/jcp2030036 2022

  7. [7]

    Derawi, M., Yang, B., Busch, C.: Fingerprint recognition with embedded cameras on mobile phones. vol. 94, pp. 136–147 (01 2012)

    work page 2012

  8. [8]

    IEEE TPAMI47(9), 7812–7831 (Sep 2025)

    Dong, C., Kumar, A.: Bridging dimensions in fingerprints to advance distinctive- ness: Recovering 3D minutiae from a single contactless 2D fingerprint image. IEEE TPAMI47(9), 7812–7831 (Sep 2025)

    work page 2025

  9. [9]

    In: Selfie Biometrics, pp

    Donida Labati, R., Genovese, A., Piuri, V., Scotti, F.: A scheme for fingerphoto recognition in smartphones. In: Selfie Biometrics, pp. 49–66. Advances in computer vision and pattern recognition, Springer International Publishing, Cham (2019)

    work page 2019

  10. [10]

    IEEE TPAMI43, 1981–1997 (2019)

    Engelsma, J.J., Cao, K., Jain, A.K.: Learning a fixed-length fingerprint represen- tation. IEEE TPAMI43, 1981–1997 (2019)

    work page 1981

  11. [11]

    Erwin, Karo, N.N.B., Yulia Sari, A., Aziza, N., Kesuma Putra, H.: The enhance- ment of fingerprint images using gabor filter. J. Phys. Conf. Ser.1196, 012045 (Mar 2019). https://doi.org/10.1088/1742-6596/1196/1/012045

    work page doi:10.1088/1742-6596/1196/1/012045 2019

  12. [12]

    In: CONECCT 2024

    G, Y.K., Bhattacharya, S., Aishwarya, N.: Remote photoplethysmography (rPPG) for contactless blood oxygen saturation monitoring. In: CONECCT 2024. pp. 1–6. IEEE (Jul 2024). https://doi.org/10.1109/CONECCT62155.2024.10677187

    work page doi:10.1109/conecct62155.2024.10677187 2024

  13. [13]

    IEEE TIFS17, 196–210 (2022)

    Grosz, S.A., Engelsma, J.J., Liu, E., Jain, A.K.: C2CL: Contact to contactless fingerprint matching. IEEE TIFS17, 196–210 (2022). https://doi.org/10.48550/ arXiv.2104.02811

    work page arXiv 2022

  14. [14]

    Deep residual learning for image recognition,

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR. pp. 770–778 (2016). https://doi.org/10.1109/CVPR.2016.90

    work page doi:10.1109/cvpr.2016.90 2016

  15. [15]

    Jawade, B., Mohan, D.D., Setlur, S., Ratha, N., Govindaraju, V.: RidgeBase: A Cross-Sensor Multi-Finger contactless fingerprint dataset (2023)

    work page 2023

  16. [16]

    Multimed

    Kaplesh, P., Gupta, A., Bansal, D., Sofat, S., Mittal, A.: Vision transformer for contactless fingerprint classification. Multimed. Tools Appl.84(26), 31239–31259 (Nov 2024). https://doi.org/10.1007/s11042-024-20396-4

    work page doi:10.1007/s11042-024-20396-4 2024

  17. [17]

    In: Contactless 3D Fingerprint Identification, pp

    Kumar, A.: Introduction to trends in fingerprint identification. In: Contactless 3D Fingerprint Identification, pp. 1–15. Springer, Berlin/Heidelberg, Germany (2018)

    work page 2018

  18. [18]

    In: CVPRW

    Kumar,A.,Zhou,Y.:Contactlessfingerprintidentificationusinglevelzerofeatures. In: CVPRW. Colorado Springs, CO, USA

    work page

  19. [19]

    In: RMKMATE 2025

    Kumaran, Agrawal, O., Paul, J., Rajesh: Contactless fingerprint recognition using SIFT enhanced minutiae matching. In: RMKMATE 2025. pp. 1–8. IEEE (May 2025). https://doi.org/10.1109/RMKMATE64874.2025.11042365

    work page doi:10.1109/rmkmate64874.2025.11042365 2025

  20. [20]

    In: Lecture Notes in Computer Science, pp

    Lee, C., Lee, S., Kim, J.: A study of touchless fingerprint recognition system. In: Lecture Notes in Computer Science, pp. 358–365. Springer, Berlin, Heidelberg (2006). https://doi.org/https://doi.org/10.1007/11815921_39 Fusion2Print 15

    work page doi:10.1007/11815921_39 2006

  21. [21]

    Multimedia Tools Appl.76(1), 333–354 (Jan 2017), https://doi.org/10.1007/s11042-015-3058-7

    Leng, L., Li, M., Kim, C., Bi, X.: Dual-source discrimination power analysis for multi-instance contactless palmprint recognition. Multimedia Tools Appl.76(1), 333–354 (Jan 2017), https://doi.org/10.1007/s11042-015-3058-7

    work page doi:10.1007/s11042-015-3058-7 2017

  22. [22]

    In: MACE 2011

    Li, T., Chen, J.: A fingerprint image enhancement method based on gabor filter. In: MACE 2011. IEEE (Jul 2011). https://doi.org/10.1109/MACE.2011.5987818

    work page doi:10.1109/mace.2011.5987818 2011

  23. [23]

    In: IJCB

    Lin, C., Kumar, A.: Multi-siamese networks to accurately match contactless to contact-based fingerprint images. In: IJCB. pp. 277–285 (2017)

    work page 2017

  24. [24]

    IEEE TIP27, 2008–2021 (2018)

    Lin, C.T., Kumar, A.: Matching contactless and contact-based conventional fin- gerprint images for biometrics identification. IEEE TIP27, 2008–2021 (2018)

    work page 2008

  25. [25]

    https://www.neurotechnology.com/ verifinger.html (2025)

    Neurotechnology: Verifinger sdk 2025.1. https://www.neurotechnology.com/ verifinger.html (2025)

    work page 2025

  26. [26]

    In: Handbook of Remote Biometrics, pp

    Parziale, G., Chen, Y.: Advanced technologies for touchless fingerprint recognition. In: Handbook of Remote Biometrics, pp. 83–109. Springer, London, UK (2009)

    work page 2009

  27. [27]

    https://fvs.sourceforge.net/ (2002)

    Patel, S.: Fingerprint verification system (fvs). https://fvs.sourceforge.net/ (2002)

    work page 2002

  28. [28]

    In: 1990 Proc

    Pizer, S.M., Johnston, R.E., Ericksen, J.P., Yankaskas, B.C., Muller, K.E.: Contrast-limited adaptive histogram equalization: speed and effectiveness. In: 1990 Proc. 1st Conf. Vis. Biomed. Comput. IEEE Comput. Soc. Press (2002)

    work page 1990

  29. [29]

    Sensors (Basel)22(3), 792 (Jan 2022)

    Priesnitz, J., Huesmann, R., Rathgeb, C., Buchmann, N., Busch, C.: Mobile con- tactless fingerprint recognition: Implementation, performance and usability as- pects. Sensors (Basel)22(3), 792 (Jan 2022)

    work page 2022

  30. [30]

    EURASIP J

    Priesnitz, J., Rathgeb, C., Buchmann, N., Busch, C., Margraf, M.: An overview of touchless 2D fingerprint recognition. EURASIP J. Image Video Process.2021(1) (Dec 2021). https://doi.org/10.1186/s13640-021-00548-4

    work page doi:10.1186/s13640-021-00548-4 2021

  31. [31]

    In: BTAS

    Raghavendra, R., Busch, C., Yang, B.: Scaling-robust fingerprint verification with smartphone camera in real-life scenarios. In: BTAS. IEEE (Sep 2013)

    work page 2013

  32. [32]

    In: BTAS

    Sankaran, A., Malhotra, A., Mittal, A., Vatsa, M., Singh, R.: On smartphone camera based fingerphoto authentication. In: BTAS. IEEE (Sep 2015)

    work page 2015

  33. [33]

    Sankaran, A., Vatsa, M., Singh, R.: Learning representations for matching fin- gerprint variants. Ph.D. thesis, IIIT-Delhi (2017), https://repository.iiitd.edu.in/ xmlui/handle/123456789/512

    work page arXiv 2017

  34. [34]

    Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition (2015), https://arxiv.org/abs/1409.1556

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  35. [35]

    In: ISPACS 2004

    Song, Y., Lee, C., Kim, J.: A new scheme for touchless fingerprint recognition system. In: ISPACS 2004. Seoul, Korea

    work page 2004

  36. [36]

    Sreehari, S., Anzar, S.M.: Touchless fingerprint recognition: A survey of recent developments and challenges. Comput. Electr. Eng.122(109894), 109894 (Mar 2025). https://doi.org/10.1016/j.compeleceng.2024.109894

    work page doi:10.1016/j.compeleceng.2024.109894 2025

  37. [37]

    Stanton, B., Stanton, B., Theofanos, M., Furman, S., Grother, P.: Usability testing of a contactless fingerprint device: Part 2. Tech. rep., US Department of Commerce, National Institute of Standards and Technology, Gaithersburg, MD, USA (2016)

    work page 2016

  38. [38]

    IEEE TIFS pp

    Tan, H., Kumar, A.: Towards more accurate contactless fingerprint minutiae ex- traction and pose-invariant matching. IEEE TIFS pp. 1–1 (2020)

    work page 2020

  39. [39]

    IEEE Transactions on In- formation Forensics and Security16, 3299–3311 (2021)

    Tan, H., Kumar, A.: Minutiae attention network with reciprocal distance loss for contactless to contact-based fingerprint identification. IEEE Transactions on In- formation Forensics and Security16, 3299–3311 (2021)

    work page 2021

  40. [40]

    IEEE TIFS5, 750–760 (2010)

    Wang, Y., Hassebrook, L.G., Lau, D.L.: Data acquisition and processing of 3-d fingerprints. IEEE TIFS5, 750–760 (2010)

    work page 2010

  41. [41]

    Micromachines (Basel)14(6), 1253 (Jun 2023)

    Yu, Y., Niu, Q., Li, X., Xue, J., Liu, W., Lin, D.: A review of fingerprint sensors: Mechanism, characteristics, and applications. Micromachines (Basel)14(6), 1253 (Jun 2023). https://doi.org/10.3390/mi14061253

    work page doi:10.3390/mi14061253 2023