Fingerprint Image Segmentation

based on Oriented Pattern Analysis

Raimundo Claudio da Silva Vasconcelos

1,2

and Helio Pedrini

Federal Institute of Brasília, Taguatinga-DF, 72146-050, Brazil

Institute of Computing, University of Campinas, Campinas-SP, 13083-852, Brazil

Keywords:

Fingerprint Segmentation, Oriented Pattern, Directional Information, Biometric Systems.

Abstract:

Segmentation is a crucial task in automatic ﬁngerprint identiﬁcation systems. This paper describes a novel seg-

mentation approach which takes into account the directional information inherent in ﬁngerprint ridges. The

method considers a directional operator to feed a k-means unsupervised clustering algorithm that labels the

image in non-overlapping regions. Morphological operations are performed to ﬁll holes and properly separate

foreground from background. Experiments conducted on Fingerprint Veriﬁcation Competition (FVC) data-

sets demonstrate that the proposed method, denoted as Oriented Pattern-based Segmentation (OPS), achieves

competitive results when compared to other well-known available ﬁngerprint segmentation approaches.

1 INTRODUCTION

There is currently a major concern regarding secu-

rity, privacy, identiﬁcation and recognition of people.

Simultaneously, automatic ﬁngerprint identiﬁcation

systems (AFIS) have become the most widely used

technology for this task (Arora et al., 2015; Ashbourn,

2014; Cao and Jain, 2015; Guesmi et al., 2015; Jain

and Hong, 1996; Kasban, 2016; Krish et al., 2018; Li

and Jain, 2015; Neumann et al., 2016), due to a num-

ber of desirable biometric characteristics: (i) univer-

satility (every person has the characteristic); (ii) per-

manence (the characteristic should be sufﬁciently in-

variant over a long period of time); (iii) collectability

(the characteristic should be easily collected and me-

asured quantitatively); (iv) distinctiveness (the cha-

racteristic is sufﬁciently different from one person to

another, even in case of identical twins).

Fingerprints are oriented texture patterns created

by interleaved ridge and valley information present

on the ﬁngertip surface. There are different possi-

ble ways to obtain a ﬁngerprint image. Rolling an

inked ﬁnger on a paper and then scanning this pa-

per was the usual technique. Due to the advances

in sensor technology, different ﬁngerprint devices can

be used on ﬁngerprint acquisition (Arjona and Batu-

rone, 2014; Cappelli et al., 2002; Hong et al., 1998;

Liu et al., 2013; Maltoni et al., 2009). Figure 1 illus-

trates some ﬁngerprint images acquired from different

sensor technologies.

(a) (b)

Figure 1: Fingerprint images acquired from different sensor

techniques: (a) electric; (b) optical; (c) thermal sweeping;

(d) capacitive. Source: Cappelli et al. (2007).

Fingerprint segmentation (Bazen and Gerez,

2001; Chen et al., 2004; Fahmy and Thabet, 2013; Liu

et al., 2016; Mehtre et al., 1987; Sankaran et al., 2017;

Yang et al., 2015) aims to distinguish foreground regi-

ons from the image background, corresponding to an

important stage in automatic ﬁngerprint recognition

systems. Since ﬁngerprint images can be affected by

diverse conditions (such as noise) and acquired by a

variety of sensors, segmentation is a very challenging

task.

Vasconcelos, R. and Pedrini, H.

Fingerprint Image Segmentation based on Oriented Pattern Analysis.

DOI: 10.5220/0007409104050412

In Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2019), pages 405-412

ISBN: 978-989-758-354-4

405

This work describes and evaluates a novel seg-

mentation approach, denoted as Oriented Pattern-

based Segmentation (OPS), which takes into account

the directional information present in ﬁngerprint rid-

ges, which is based on an operator used by an unsu-

pervised clustering algorithm to separate the image

into non-overlapping regions. Evaluation is perfor-

med on four Fingerprint Veriﬁcation Competition

(FVC) (FVC, 2018) datasets to demonstrate the ef-

fectiveness of the results.

The text is organized as follows. Section 2 in-

troduces an operator that extracts anisotropic quality

information from ﬁngerprint images. In Section 3,

the segmentation problem related to these images

is addressed. Experimental results are provided in

Section 4. Finally, concluding remarks and directions

for future work are presented in Section 5.

2 DIRECTIONAL INFORMATION

OPERATOR

Textural analysis (Jain et al., 2001; Joy and Azath,

2017; Marasco et al., 2018; Marasco and Sansone,

2010) constitutes an important technique for pro-

cessing images containing directional information,

whose magnitude of the corresponding anisotropy

should be measured.

This work is particularly interested in a measure

of the distance between ridge and valley information

in ﬁngerprints. A systematic way to compute such

distance is ﬁrstly considered within a given neighbor-

hood. Then, a speciﬁc ﬁngerprint quality can be set.

Some deﬁnitions related to the particular neig-

hborhood considered in this work are initially intro-

duced. Let Γ be a sliding window of size M × N (usu-

ally, M = N = (2l + 1), l ∈ Z) of an image f (x,y),

f : (x, y) ∈ D

⊂ Z

7→ Z. Moreover, let D be the

number of considered directions in Γ, and n the cor-

responding number of pixels in a given direction.

In this work, these pixels are referred to as test

points. It is worth noticing that, in order to represent

all D directions in a two-dimensional grid, the number

n of test points has a minimum bound, that is, for any

n ≥ 2, we can deﬁne up to (2n −2) directions.

Thus, given a discrete square grid with M = N = n

and the origin (0,0) located at its upper left corner,

the coordinates (x,y) of the n test points, in a given

direction α, are computed as:

x = x

center

+ p cos(α)

y = y

center

− p sin(α)

(1)

for all p such that −n/2 ≤ p ≤ n/2. Moreover, x

center

and y

center

are the coordinates of the point containing

the sliding window Γ centered in this location. Fi-

gure 2 shows an example of test points for α = 45

◦

and n = 9.

Figure 2: Test points for α = 45

◦

and n = 9.

Finally, this neighborhood can be deﬁned as a

set S

of D test points with length n and discrete

direction i, which can easily be computed by repe-

ating the above procedure for all D directions (i ∈

{0,1,...D −1}), by respectively changing the value of

α accordingly (α = 0, 1 · 180/D, 2 · 180/D,...,(D −

1) · 180/D).

In this approach, it is assumed that, in the

aforementioned neighborhood, the physics of the

image acquisition imposes certain arrangements on

the image gray levels. That is the case, for example,

of the image points associated with two distinct regi-

ons: one which is parallel and the other perpendicular

to the ﬂow orientation contained in an intensity pat-

tern created by some anisotropic process (Kass and

Witkin, 1987).

Under such conditions, it can be observed a strong

statistical relationship between the gray levels along

the ﬂow orientation and, by contrast, gradual changes

causing this relationship to weaken along the corre-

sponding perpendicular orientation. These aspects re-

veal a direct connection between the anisotropy and

particular combinations of distinct random variables

around of these regions.

For the sake of simpliﬁcation, this work borrows

and adapts the formalism presented by Oliveira and

Leite (2008), whose approach used oriented informa-

tion to reconnect broken ridges. Here, it is used to

measure quality. Therefore, the abstract idea behind

this quality index consists in analyzing samples drawn

from these two image regions in order to quantify the

difference that makes the anisotropy distinguishable.

The main steps of the proposed operator are des-

cribed as follows. Figure 3 shows some results pro-

duced through this process.

• Consider f as input image, S representing the

neighborhood and D as the number of considered

directions. Different amounts of test points and

directions can be set up in accordance with a cer-

tain scale and resolution for a given image. On the

other hand, several quality and information crite-

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

406

original directional difference quality

Figure 3: Fingerprint image operator process.

ria can be considered to express separability (or

contrast), variability, homogeneity, completeness,

entropy and so forth;

• Compute standard deviation (or other information

parameter as mean, moments of higher orders,

among others) on this neighborhood S for each of

the D directions;

• The information associated with each direction i

is compared to the one obtained from another di-

rection j, i 6= j. Once perpendicular direction

pairs are sufﬁcient to characterize orientated pat-

terns. Thus, the predominant orientation informa-

tion is obtained;

• The pair of directions i and j exhibiting the hig-

hest information contrast in a given pixel, deﬁnes

the local orientation (directional) image.

The problem of ﬁngerprint image segmentation

based on pixel-wise quality is discussed in the next

section.

3 PROPOSED FINGERPRINT

IMAGE SEGMENTATION

The segmentation method proposed in this work is

composed of the following steps: (i) ﬁngerprint qua-

lity analysis: this step estimates the local quality of

the input image; (ii) mathematical morphology trans-

formation: some morphological transformations are

applied to attenuate local discrepancies; (iii) unsuper-

vised classiﬁcation: the k-means clustering process

is performed on the attenuated image to ﬁnd mar-

kers (pixels) corresponding to regions with different

quality; (iv) image segmentation by watershed: a seg-

mented image is obtained through the application of

watershed inﬂuence zones.

This work considers ﬁngerprint pattern as a regu-

lar anisotropic texture. There is a certain regularity

on the ridge and valley information. The gray levels

in a perpendicular direction to the ridge-valley struc-

ture can be modeled as smoothed sinusoidal signals.

Similarly, despite the gradual changes on ridge and

valley gray levels, there is a certain homogeneity of

the pixels along their parallel orientations.

For directional ﬁeld estimation, this method uses

variance to express homogeneity of each S

. In such

a case, a pair of directions exhibits the highest con-

trast information and deﬁnes the directional image O

as follows:

O(x,y)=











i, if σ

(x,y)) < σ

(x,y))

i +

, if σ

(x,y)) ≥ σ

(x,y))

The descriptor expresses the strength of the in-

formation along certain oriented information. The

next step is the application of morphological transfor-

mations to attenuate discrepancies. Considering this

image, a k-means clustering algorithm is used to ﬁnd

non-overlapping regions with distinct quality. The pa-

rameter k deﬁnes the number of regions. Empirically,

value k = 3 showed the best response.

The centroid values are used as markers and the

region with lower value is considered as background.

Eventually, holes may exist in the image (background

surrounded by foreground) and a watershed transform

can be applied successfully. Figure 4 illustrates this

process. Original ﬁngerprint images are shown in Fi-

gures (a) and (d). There are three non-overlapping

Fingerprint Image Segmentation based on Oriented Pattern Analysis

407

regions in Figures (b) and (e), where the background

is represented in black color. It can be observed, in

(b) and (e), that there is a hole (in black color) in the

ﬁngerprint foreground, deﬁned by two regions, one

with gray color and the other in white color. This re-

gion will disappear after the application of a waters-

hed technique. Figures 4 (c) and (f) show the fore-

ground masks that encompass those regions.

4 EXPERIMENTAL EVALUATION

The effectiveness of the proposed segmentation algo-

rithm is veriﬁed through four public ﬁngerprint ve-

riﬁcation competition datasets (FVC2000, FVC2002,

FVC2004 and FVC2006) (Cappelli et al., 2007).

Related work on ﬁngerprint image segmentation

often validates the corresponding segmented images

via their own ground truth. Notwithstanding, there is

no public ground truth available for these FVC data-

bases. Ideally, a ground truth should be built by three

specialists in order to achieve a consensus.

The ﬁnal goal of the ﬁngerprint segmentation pro-

cess is to improve the ﬁngerprint recognition perfor-

mance. Thus, the assessment of the segmentation al-

gorithm effectiveness should be carried out through a

ﬁngerprint recognition test. Recognition performance

indicates whether the segmentation algorithm is ade-

quate or not.

Based on this scenario, this work opted for two ty-

pes of validation: a quantitative and a qualitative one.

Two approaches, proposed by Thai and Gottschlich

(2016) and Kovesi (2018), were used in the compari-

son of the results. The ﬁrst segmentation method de-

composes the image into three portions and considers

texture and oriented patterns present in the ﬁngerprint.

The second one partitions the image into blocks and

evaluates the standard deviation in each region; if this

value is above a threshold, it is deemed part of the

ﬁngerprint.

4.1 Fingerprint Databases

The Fingerprint Veriﬁcation Competition (FVC) took

place in 2000, 2002, 2004 and 2006, as an initiative

to compare ﬁngerprint matching algorithms. These

competitions were organized by the Biometric Sy-

stem Laboratory of the University of Bologna Cap-

pelli et al. (2007), as well as Pattern Recognition and

Image Processing Laboratory of the Michigan State

University, Biometric Test Center of San Jose State

University and, in the last year, Biometrics Research

Laboratory of the Universidad Autonoma de Madrid.

In this work, 2000, 2002, 2004 and 2006 datasets

were used in the experiments to validate the propo-

sed operator based on directional information.

Each FVC dataset contains 4 databases, namely

DB1A, DB2A, DB3A and DB4A. The 2000, 2002

and 2004 databases contain 800 ﬁngerprint images

(i.e., there are 100 ﬁngers with eight samples). The

2006 dataset contains 4 databases and each one has

1680 ﬁngerprint images (140 individuals have col-

lected 12 samples).

The image size of each dataset is different from

one another and the resolution is over 500 dpi. Each

database was acquired by different sensor modalities.

Rules have changed from one competition (2004)

to another (2006). For 2004, each sample in the

subset A is matched against the remaining samples

of the same ﬁnger to compute the False Non Ma-

tch Rate (FNMR) (also referred to as False Rejection

Rate (FRR)). If matching g against h is performed,

the symmetric one (i.e., h against g) is not executed to

avoid correlation.

The total number of genuine tests (if no enrol-

lment rejections occur) is: ((8*7)/2)*100 = 2,800.

The total number of false acceptance tests (False Ma-

tch Rate (FMR), also referred to as False Acceptance

Rate (FAR)) is calculated as follows: the ﬁrst sam-

ple of each ﬁnger in the subset A is matched against

the ﬁrst sample of the remaining ﬁngers in A. If ma-

tching g against h is performed, the symmetric one

(i.e., h against g) is not executed to avoid correla-

tion: ((100*99)/2) = 4,950. Therefore, it is possible

to compute the False Rejection Rate (FRR), which is

the likelihood of samples for the same ﬁnger being

considered as having different ﬁngers.

4.2 Griaule AFIS

In this study, Griaule AFIS Biometrics (2018) was

used to represent and match ﬁngerprints as minu-

tiae. The Griaule ﬁngerprint recognition framework

won the Open Category, section “average results over

all databases” of the Fingerprint Veriﬁcation Con-

test 2006 (Cappelli et al., 2007), achieving the best

average equal error rate (EER).

Minutiae matching is certainly the most well-

known and widely used method for ﬁngerprint cor-

respondence, as an analogy with the way forensic ex-

perts compare ﬁngerprint images and their acceptance

as a proof of identity in court (Maltoni et al., 2009).

4.3 Quantitative Analysis

In this work, we used the DB2A and DB3A datasets in

our experiments. Performance was measured through

Equal Error Rate (EER) and Area Under the Recei-

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

408

(a) original (b) three regions (c) mask (d) original (e) three regions (f) mask

Figure 4: Fingerprint segmentation process: (a) original image; (b) three clustered regions by k-means; (c) image mask after

watershed.

Table 1: Results of AUC and ERR metrics for ﬁngerprint veriﬁcation.

AUC

2000 2002 2004 2006

DB2_A DB3_A DB2_A DB3_A DB2_A DB3_A DB2_A DB3_A

Proposed OPS 0.98886 0.95783 0.99467 0.97470 0.92707 0.96974 0.99751 0.97251

Thai and Gottschlich (2016) 0.98982 0.95679 0.99418 0.78710 0.95089 0.94713 0.99565 0.97158

Kovesi (2018) 0.96708 0.94134 0.99480 0.95188 0.95090 0.95118 0.98200 0.97315

EER

2000 2002 2004 2006

DB2_A DB3_A DB2_A DB3_A DB2_A DB3_A DB2_A DB3_A

Proposed OPS 0.02524 0.07384 0.01279 0.05294 0.12369 0.05521 0.00857 0.04715

Thai and Gottschlich (2016) 0.02545 0.08116 0.00660 0.27725 0.09916 0.09359 0.01173 0.04904

Kovesi (2018) 0.04989 0.08209 0.01400 0.08138 0.09338 0.08236 0.03807 0.05826

ver Operating Characteristic Curve (ROC AUC) me-

trics (Toh et al., 2008; Vacca, 2007), considering the

protocol proposed and used in the FVC.

A quantitative comparison of the results for EER

and ROC AUC metrics for three segmentation met-

hods applied to FVC datasets is shown in Table 1.

It can be observed that the proposed method achie-

ved very competitive results, positioning slightly hig-

her or below the other evaluated approaches, when

AUC metric is considered. For EER metric, the pro-

posed method obtained better rates than the other ap-

proaches.

4.4 Qualitative Analysis

Table 2 shows the segmentation results of some ﬁn-

gerprint images based on different techniques. The

results indicate that the proposed segmentation met-

hod is highly competitive compared to the evaluated

approaches.

Image segmentation is a challenging problem and

many question remain open. Visual inspection con-

ducted by experts is still important. Issues related to

the size of the resulting segmented area, also known

as regions of interest (ROI), may be relevant, in addi-

tion to aspects derived from the quantitative analysis.

4.5 Discussion

Accurate segmentation is a complex, however, a criti-

cal task since it reduces the computational time of the

following processing steps and discards spurious mi-

nutiae. Most of the segmentation methods available

in the literature are highly dependent either on empi-

rical thresholds or a well-trained model. Furthermore,

many of these experiments are sensor dependent.

The proposed algorithm employs an unsupervised

clustering based only on oriented features inherent to

a ﬁngerprint image. The parameter k, which refers to

the number of regions to be created, can also be es-

tablished from a preliminary evaluation of ﬁngerprint

quality, making this step more adaptive. The morpho-

logical watershed operation was applied to obtain two

well-deﬁned regions.

The ﬁngerprint segmentation algorithm proposed

in this work is suitable for different sensors and does

not need empirical thresholds or a well-trained mo-

del. Experimental results using FVC2000, FVC2002,

FVC2004 and FVC2006 datasets, showed that our ap-

proach, in addition to its computational simplicity,

presents a lower classiﬁcation error when compared

to other segmentation methods.

Fingerprint Image Segmentation based on Oriented Pattern Analysis

409

Table 2: Comparative analysis among different ﬁngerprint image segmentation methods.

Dataset Proposed Thai and Gottschlich Kovesi

OPS (2016) (2018)

DB2A 2000

DB3A 2000

DB2A 2002

DB3A 2002

DB2A 2004

DB3A 2004

DB2A 2006

DB3A 2006

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

410

5 CONCLUSIONS AND FUTURE

WORK

This paper presented and evaluated a ﬁngerprint

image segmentation method. For each pixel, the al-

gorithm calculates the dominant direction within a gi-

ven neighborhood. By applying statistical measures,

it is possible to compute the strength of anisotropic

information. The proposed method also employed an

unsupervised clustering algorithm to deﬁne the inte-

rest regions. Followed by a set of morphological ope-

rations, the ﬁngerprint contour can be extracted.

The validity of the proposed method is demon-

strated through a comparison against two other ap-

proaches available in the literature. No training or

prior information about thresholding level is neces-

sary, which makes the evaluation more independent.

The proposed method is suitable for different sensors.

Directions for future work include the evaluation

of the directional operator as a ﬁngerprint image qua-

lity indicator. It could be integrated into a quality as-

sessment framework along with other features. In ad-

dition, an accurate estimation of ﬁngerprint orienta-

tion image is essential in ﬁngerprint classiﬁcation and

this directional operator can also be used for this task.

ACKNOWLEDGMENTS

The authors thank FAPESP (grant #2017/12646-3),

CNPq (grant #305169/2015-7) and CAPES for the ﬁ-

nancial support.

REFERENCES

Arjona, R. and Baturone, I. (2014). A Hardware Solu-

tion for Real-Time Intelligent Fingerprint Acquisition.

Journal of Real-Time Image Processing, 9(1):95–109.

Arora, S. S., Cao, K., Jain, A. K., and Michaud, G. (2015).

Crowd Powered Latent Fingerprint Identiﬁcation: Fu-

sing AFIS with Examiner Markups. In International

Conference on Biometrics, pages 363–370. IEEE.

Ashbourn, J. (2014). Biometrics: Advanced Identity Veriﬁ-

cation: The Complete Guide. Springer.

Bazen, A. M. and Gerez, S. H. (2001). Segmentation of

Fingerprint Images. In Workshop on Circuits, Systems

and Signal Processing, pages 276–280. Citeseer.

Cao, K. and Jain, A. K. (2015). Learning Fingerprint

Reconstruction: From Minutiae to Image. IEEE

Transactions on Information Forensics and Security,

10(1):104–117.

Cappelli, R., Ferrara, M., Franco, A., and Maltoni, D.

(2007). Fingerprint Veriﬁcation Competition 2006.

Biometric Technology Today, 15(7-8).

Cappelli, R., Maio, D., and Maltoni, D. (2002). Synthetic

Fingerprint-Database Generation. In 16th Internati-

onal Conference on Pattern Recognition, volume 3,

pages 744–747. IEEE.

Chen, X., Tian, J., Cheng, J., and Yang, X. (2004). Segmen-

tation of Fingerprint Images using Linear Classiﬁer.

EURASIP Journal on Advances in Signal Processing,

2004(4).

Fahmy, M. F. and Thabet, M. A. (2013). A Fingerprint Seg-

mentation Technique based on Morphological Proces-

sing. In IEEE International Symposium on Signal Pro-

cessing and Information Technology, pages 215–220.

FVC (2018). Fingerprint Veriﬁcation Competitions.

https://biolab.csr.unibo.it/FVCOnGoing/UI/Form/

Home.aspx.

Griaule AFIS Biometrics (2018). Griaule Big Data Biome-

trics. http://www.griaulebiometrics.com/new/.

Guesmi, H., Trichili, H., Alimi, A. M., and Solaiman, B.

(2015). Fingerprint Veriﬁcation System based on Cur-

velet Transform and Possibility Theory. Multimedia

Tools and Applications, 74(9):3253–3272.

Hong, L., Wan, Y., and Jain, A. (1998). Fingerprint Image

Enhancement: Algorithm and Performance Evalua-

tion. IEEE Transactions on Pattern Analysis and Ma-

chine Intelligence, 20:777–789.

Jain, A. and Hong, L. (1996). On-line Fingerprint Veriﬁ-

cation. In 13th International Conference on Pattern

Recognition, volume 3, pages 596–600. IEEE.

Jain, A., Ross, A., and Prabhakar, S. (2001). Fingerprint

Matching using Minutiae and Texture Features. In

IEEE International Conference on Image Processing,,

volume 3, pages 282–285. IEEE.

Joy, R. C. and Azath, M. (2017). Fingerprint Image Seg-

mentation Using Textural Features. In International

Conference on Computer Vision and Image Proces-

sing, pages 1–12. Springer.

Kasban, H. (2016). Fingerprints Veriﬁcation based on their

Spectrum. Neurocomputing, 171:910–920.

Kass, M. and Witkin, A. (1987). Analyzing Oriented Pat-

terns. Computer Vision, Graphics, and Image Proces-

sing, 37(3):362–385.

Kovesi, P. (2018). Example of Fingerprint En-

hancement. http://www.peterkovesi.com/matlabfns/

FingerPrints/Docs/index.html.

Krish, R. P., Fierrez, J., Ramos, D., Alonso-Fernandez, F.,

and Bigun, J. (2018). Improving Automated Latent

Fingerprint Identiﬁcation using Extended Minutia Ty-

pes. Information Fusion.

Li, S. Z. and Jain, A. (2015). Encyclopedia of Biometrics.

Springer Publishing Company, Incorporated.

Liu, F., Zhang, D., Song, C., and Lu, G. (2013). Touchless

Multiview Fingerprint Acquisition and Mosaicking.

IEEE Transactions on Instrumentation and Measure-

ment, 62(9):2492–2502.

Liu, S., Liu, M., and Yang, Z. (2016). Latent Fingerprint

Segmentation based on Linear Density. In Internatio-

nal Conference on Biometrics, pages 1–6. IEEE.

Maltoni, D., Maio, D., Jain, A. K., and Prabhakar, S. (2009).

Handbook of Fingerprint Recognition. Springer, 2nd

edition.

Fingerprint Image Segmentation based on Oriented Pattern Analysis

411

Marasco, E., Feldman, A., and Romine, K. R. (2018). En-

hancing Optical Cross-Sensor Fingerprint Matching

Using Local Textural Features. In IEEE Winter Appli-

cations of Computer Vision Workshops, pages 37–43.

IEEE.

Marasco, E. and Sansone, C. (2010). An Anti-Spooﬁng

Technique using Multiple Textural Features in Fin-

gerprint Scanners. In IEEE Workshop on Biometric

Measurements and Systems for Security and Medical

Applications, pages 8–14. IEEE.

Mehtre, B., Murthy, N., Kapoor, S., and Chatterjee, B.

(1987). Segmentation of Fingerprint Images using the

Directional Image. Pattern Recognition, 20(4):429–

435.

Neumann, C., Armstrong, D. E., and Wu, T. (2016). Deter-

mination of AFIS "Sufﬁciency" in Friction ridge Ex-

amination. Forensic Science International, 263:114–

125.

Oliveira, M. A. and Leite, N. J. (2008). A Multiscale Di-

rectional Operator and Morphological Tools for Re-

connecting Broken Ridges in Fingerprint Images. Pat-

tern Recognition, 41(1):367–377.

Sankaran, A., Jain, A., Vashisth, T., Vatsa, M., and Singh,

R. (2017). Adaptive Latent Fingerprint Segmentation

using Feature Selection and Random Decision Forest

Classiﬁcation. Information Fusion, 34:1–15.

Thai, D. and Gottschlich, C. (2016). Global Variational

Method for Fingerprint Segmentation by Three-part

Decomposition. IET Biometrics, 5:120–130.

Toh, K.-A., Kim, J., and Lee, S. (2008). Maximizing Area

under ROC Curve for Biometric Scores Fusion. Pat-

tern Recognition, 41(11):3373–3392.

Vacca, J. R. (2007). Biometric Technologies and Veriﬁca-

tion Systems. Elsevier.

Yang, X., Feng, J., Zhou, J., and Xia, S. (2015). Detection

and Segmentation of Latent Fingerprints. In IEEE In-

ternational Workshop on Information Forensics and

Security, pages 1–6. IEEE.

VISAPP 2019 - 14th International Conference on Computer Vision Theory and Applications

412