Infant Fingerprints: Quality Assessment & Recognition

João L. H. Dall Agnol; Dalcimar Casanova; Rodrigo Minetto; Luiz F. P. Southier; Marcelo Teixeira; Marcelo Filipak; Anil K. Jain

cs.CVOct 14, 2026

Infant Fingerprints: Quality Assessment & Recognition

Authors: João L. H. Dall Agnol, Dalcimar Casanova, Rodrigo Minetto, Luiz F. P. Southier, Marcelo Teixeira, Marcelo Filipak, Anil K. Jain

Organizations: Federal University of Technology - Paraná (UTFPR), Brazil · Michigan State University, East Lansing, MI, USA

Abstract

Infant fingerprint recognition is hindered by a combination of physiological characteristics, such as small ridge width and spacing, and acquisition challenges that include involuntary motion, skin residue, and uneven fingertip pressure. These factors undermine fingerprint quality standards originally designed for adults, which assume feature distributions that do not hold for infants. To address this gap, this paper evaluates ten fingerprint quality assessment methods on a longitudinal dataset that spans six age groups from newborns to three months, alongside a parameter tuning routine that uncovers age-specific fingerprint characteristics to recalibrate methods with domain-dependent assumptions. Both default and tuned configurations are compared through recognition experiments assessing the effectiveness of quality-guided gallery selection. Template-fusion strategies and latent fingerprint recognition technologies are further examined to identify the best possible matching rates achievable by the top-performing method. Results show that tuned parameter sets outperform default configurations for the majority of methods and age groups, indicating that targeted recalibration of existing frameworks improves gallery utility and contributes to more reliable infant fingerprint recognition.

Figures & tables

Panel of Figure 1 — Fig. 1 : Infant fingerprint acquisition process: top-left shows capture procedure used for the UTFPR-IFD dataset [ 5 ] ; top-right contains examples of fingerprint video frames obtained from a single newborn fingerprint; bottom shows the InfantID scanner used to acquire the images, alongside its key characteristics such as sampling rate and technology [ 6 ] .

Panel of Figure 2 — Fig. 2 : Fingerprint quality estimation taxonomy. Methods are classified according to their primary feature, resolution and the age they target. Features directly related to visible patterns are also showcased.

Panel of Figure 3 — Fig. 3 : Overview of the UTFPR-IFD dataset [ 5 ] . A total of 14 fingerprint videos are acquired per session from both hands, with thumbs and index fingers captured twice due to their higher recognition utility. Stacked images represent the video-based frames collected at each stage of life

Method	Parameters tuned
Gabor	$m$ , $f$ , $\sigma$ , Gaussian window ( $\sigma_{\theta}$ )
LCS	$W_{r}^{\min},W_{r}^{\Delta},W_{v}^{\min},W_{v}^{\Delta}$ , block size
RPS	$r_{\min},r_{\Delta},\Delta r$ , $\theta$
OCL	block size
RVU	block size
DFB	block size

TABLE I : Summary of methods subjected to parameter tuning and their corresponding search variables.

Age	Dom. Frequency	Rid. Width	Val. Width
Newborn	0.211 - 0.397	2-11	1-3
7-day	0.379 - 0.560	3-6	1-3
15-day	0.440 - 0.700	2-14	2-3
1-month	0.240 - 0.700	2-9	2-5
2-month	0.440 - 0.700	2-9	2-5
3-month	0.440 - 0.700	2-9	1-6

TABLE II : Tuned RPS and LCS parameter intervals for each age group (in pixels). Adult reference ranges reported in [ 7 ] are included for comparison.

Panel of Figure 6 — Fig. 4 : FNMR plots for default (def.) and tuned (opt.) parameters for all methods (one per panel), accumulated across all six age groups and all 30 folds. Lower values indicate better performance.

Panel of Figure 7 — Fig. 5 : LCS scoring of a newborn thumb (test only subset). The fingerprint core region is showcased for the best frames selected by the default and tuned configurations.

Method	Setting	Newborn		7-day		15-day		1-month		2-month		3-month
Method	Setting	0.1%	1.0%	0.1%	1.0%	0.1%	1.0%	0.1%	1.0%	0.1%	1.0%	0.1%	1.0%
Gabor	tuned	30.6 $\pm$ 6.6	46.8 $\pm$ 6.5	53.3 $\pm$ 5.5	75.8 $\pm$ 4.2	70.7 $\pm$ 6.7	86.3 $\pm$ 3.8	77.9 $\pm$ 6.4	87.3 $\pm$ 4.5	80.2 $\pm$ 5.3	88.4 $\pm$ 2.4	72.7 $\pm$ 7.5	84.7 $\pm$ 4.8
Gabor	default	29.4 $\pm$ 6.7	44.2 $\pm$ 8.3	54.1 $\pm$ 8.1	73.8 $\pm$ 5.9	68.2 $\pm$ 7.4	83.0 $\pm$ 4.9	78.4 $\pm$ 6.9	87.0 $\pm$ 4.9	80.0 $\pm$ 6.2	88.1 $\pm$ 3.6	73.6 $\pm$ 6.5	84.2 $\pm$ 4.4
LCS	tuned	34.5 $\pm$ 5.7	53.6 $\pm$ 6.9	57.1 $\pm$ 7.2	78.7 $\pm$ 5.4	73.1 $\pm$ 4.5	88.0 $\pm$ 3.1	75.8 $\pm$ 7.4	86.7 $\pm$ 5.0	79.9 $\pm$ 5.6	87.3 $\pm$ 3.6	71.8 $\pm$ 10.0	82.8 $\pm$ 6.3
LCS	default	27.5 $\pm$ 4.9	45.3 $\pm$ 6.3	50.5 $\pm$ 7.1	72.1 $\pm$ 4.9	65.3 $\pm$ 5.5	82.0 $\pm$ 4.1	76.5 $\pm$ 6.6	84.9 $\pm$ 4.9	75.8 $\pm$ 6.3	83.8 $\pm$ 5.3	68.1 $\pm$ 8.7	80.3 $\pm$ 6.6
RPS	tuned	30.2 $\pm$ 6.0	47.5 $\pm$ 6.2	53.8 $\pm$ 7.2	77.3 $\pm$ 5.0	73.8 $\pm$ 7.3	87.3 $\pm$ 4.1	80.1 $\pm$ 6.6	88.4 $\pm$ 3.7	83.2 $\pm$ 6.2	90.5 $\pm$ 2.9	78.4 $\pm$ 6.3	88.8 $\pm$ 3.7

TABLE III : Mean matching performance (TMR@FMR=0.1% and

1.0\%

) by age and method (90 digits per fold, 5 folds) with standard deviation. Bold values indicate best results per method, underlined values indicate best results for each age.

Method	Setting	Mean Rank
RPS	tuned	$2.8$
MQCDA	—	$3.6$
LCS	tuned	$4.4$
Gabor	tuned	$4.6$
SIG	—	$4.9$
DFB	tuned	$6.8$

TABLE IV : Friedman-Nemenyi ranking analysis of quality methods (lower rank = better performance) at FMR=1.0%.

Panel of Figure 10 — Fig. 6 : TMR@FMR curves by age group. Each curve corresponds to the top-performing quality assessment method at FMR=1.0% for its respective age group.

	VeriFinger	MegaMatcher
Age	TMR@FMR=1.0%	TMR@FMR=1.0%
Newborn	60.1 $\pm$ 7.8 ( $\uparrow 12.6$ )	77.2 $\pm$ 6.4 ( $\uparrow 29.7$ )
7-day	85.6 $\pm$ 2.6 ( $\uparrow 08.3$ )	93.2 $\pm$ 1.4 ( $\uparrow 15.9$ )
15-day	92.1 $\pm$ 3.1 ( $\uparrow 04.8$ )	96.4 $\pm$ 1.6 ( $\uparrow 09.1$ )
1-month	92.8 $\pm$ 2.5 ( $\uparrow 04.4$ )	96.4 $\pm$ 2.8 ( $\uparrow 08.0$ )
2-month	94.5 $\pm$ 1.0 ( $\uparrow 04.0$ )	97.6 $\pm$ 0.9 ( $\uparrow 07.1$ )

TABLE V : Matching performance with frame-fusion over VeriFinger and MegaMatcher (Latent) comparators using RPS (tuned) (90 digits per fold, 5 folds).

Panel of Figure 12 — Fig. 7 : Examples of genuine newborn sample pairs. The thumb pair (top) exceeded the MegaMatcher threshold, whereas the little-finger pair (bottom) did not.

Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

Age Group	Gabor	LCS	RVU
	( $m$ , $f$ , $\sigma$ , $\sigma_{g}$ )	$Wr$ — $Wv$ , B. Size	B. Size
Newborn	8, 0.500, 0.807, 1	2–11, 1–3, $64\times 64$	$128\times 128$
7-day	8, 0.500, 0.807, 5	3–6, 1–3, $64\times 64$	$64\times 64$
15-day	8, 0.050, 8.071, 5	2–14, 2–3, $64\times 64$	$64\times 64$
1-month	8, 0.500, 0.807, 1	2–9, 2–5, $64\times 64$	$64\times 64$
2-month	8, 0.500, 0.807, 2	2–9, 2–5, $64\times 64$	$64\times 64$

Appendix

TABLE VI : Tuned parameter sets (Part I).

Age Group	RPS	OCL	DFB
	( $r_{min}$ – $r_{max}$ , $\Delta r$ , $\theta$ )	B. Size	B. Size
Newborn	0.211–0.397, 10, 100	$128\times 128$	$64\times 64$
7-day	0.379–0.560, 22, 79	$128\times 128$	$64\times 64$
15-day	0.440–0.700, 17, 132	$128\times 128$	$64\times 64$
1-month	0.240–0.700, 57, 132	$128\times 128$	$64\times 64$
2-month	0.440–0.700, 81, 81	$128\times 128$	$64\times 64$

Appendix

TABLE VII : Tuned parameter sets (Part II).

Fingerprint Identification

See all 21 papers

Iterative Framework For Data Augmentation Of Segmented Fingerprints
May 29, 2026João Leonardo H. D. Agnol, Wesley Augusto de Bona, Erick Oliveira Rodrigues +4Synthetic Data Augmentation Biometric Identification
Federal University of Technology (UTFPR) Pato Branco, Parana Brazil · InfantID Curitiba, Parana, Brazil
Recursive Class Connectivity Classification (R3C) Applied to Binary Image Segmentation for Improved Infant Fingerprint Enhancement
May 25, 2026Joao Leonardo Harres Dall Agnol, Luiz Fernando Puttow Southier, Jefferson Tales 0liva +5Biometric Identification Image Enhancement
Graduate Program in lectrical and Computer Engineering (PPGEEC), Federal University of Technolozy-Parand (UTFPR), Pato Branco 85503-390, Brazil · Gradute Program in Electrical Engineering and Industrial Informatics (CPGEI), Federal University of Technology-Paraná (UTFPR), Curitiba 80230-901, Brazil · Graduate Program in Production and Systems Engineering (PPGEPS), Federal University of Technology-Paran (UTFPR), Pato Branco 85503-390, Brazil +1
AFID: A Unified Open Framework for Automated Fingermark Identification, Quality Assessment and Feature Extraction
Sep 7, 2026Tim Oblak, Rudolf Haraksim, Peter PeerBiometric Identification Fingerprint Identification
Joint Research Centre of the European Commission in Ispra, Italy

May 29, 2026J/K move · Enter open · S save

Iterative Framework For Data Augmentation Of Segmented Fingerprints

João Leonardo H. D. Agnol, Wesley Augusto de Bona, Erick Oliveira Rodrigues, Luiz Fernando Puttow Southier+3

Federal University of Technology (UTFPR) Pato Branco, Parana Brazil · InfantID Curitiba, Parana, Brazil

Synthetic Data Augmentation Biometric Identification Data Augmentation

Open paper AskPDF arXiv

Infant biometrics presents unique challenges due to the physiological differences between infants and adults, compounded by the scarcity of available data for research that limits the development of robust matching systems. This paper proposes a novel data augmentation method that uses iterative techniques to generate diverse variants of segmented fingerprints by inducing errors in a convolutional neural network trained to extract fingerprint ridges and valleys. Experiments on real infant fingerprints demonstrate the method's effectiveness in expanding fingerprint variability, with augmentations exhibiting significant fluctuations in minutiae counts while still retaining visual similarity to the originals. The study also highlights the method's customizable nature for applying varying levels of changes to fingerprint segmentations. Future research includes training segmentation and matching neural networks using datasets augmented by the proposed framework.

Abstract

Fingerprint Identification

Explore similar work

Iterative Framework For Data Augmentation Of Segmented Fingerprints

Recursive Class Connectivity Classification (R3C) Applied to Binary Image Segmentation for Improved Infant Fingerprint Enhancement

AFID: A Unified Open Framework for Automated Fingermark Identification, Quality Assessment and Feature Extraction