Encyclopedia of Computer Graphics and Games

Living Edition
| Editors: Newton Lee

Fingerprint Verification Based on Combining Minutiae Extraction and Statistical Features

  • Anwar Yahya EbrahimEmail author
  • Hoshang Kolivand
Living reference work entry
DOI: https://doi.org/10.1007/978-3-319-08234-9_362-1



One of the most popular forms of biometrics used for personal identification is fingerprints. The extraction of multi-features illustrates the diversity of data represented from fingerprint samples, allowing mitigation of the intrapersonal variable. This study uses multiple feature extraction based on statistical tests of co-occurrence matrices to overcome the drawbacks of previous methods and minutiae extraction to achieve high accuracy toward an efficient fingerprint verification system.


The most important type of human biometrics is fingerprints. Fingerprints have been used for personal recognition in forensic applications, such as criminal investigation tools, national identity card validation, and authentication processors. The uniqueness and immutability of fingerprint patterns as well as the low cost of associated biometric equipment make fingerprints more desirable than the other types of biometrics (Maltoni and Cappelli 2009). A characteristic attribute of false fingerprints in the large view of complexly refers to forgery, combined with the actuality that fingerprint samples are distinctive to each individual. In reality, fingerprints present a distinguished basis of entropy that is used for security implementations (Dodis et al. 2006; Dass and Jain 2007; Khan 2009). There are two classes of features utilized for the recognition system: global attributes and local attributes (Ebrahim 2017a). The global attribute pattern represents a distinctive modality of a singular point as the center point that is utilized for fingerprint recognition. The local feature represented from fingerprint’s ridges data is called minutiae. The fingerprint verification (FV) system includes three processes: the fingerprint obtaining apparatus, extracted features, and verify features (Igaki et al. 1992). The dataset of FVC2002 is utilized to obtain the finger imprint by an optical device, which has high capability (Maio and Maltoni 1997). The extracted minutiae and minutiae verification are further obtained in the next stage.

A fingerprint authentication scheme is a model matching scheme that distinguishes the individual based on their fingerprint attribute (Maltoni et al. 2003). A number of various fingerprint classification systems have been improved using myriad classification approaches and datasets. Each classification method has its own specific characteristics that researchers capitalize on to advance fingerprint classification research using a particular dataset (Jain et al. 1997; Jea and Govindaraju 2005). In order to design a more reliable automatic identification system, preprocessing of fingerprints has to take place to enhance and extract fingerprint features (Wu et al. 2006; Rajkumar and Hemachandran 2011). According to Maltoni et al. (2009), most current fingerprint classification approaches depend on global attributes, including ridge orientation areas and singularities. Bazen and Gerez (2002) found that accurate classification of fingerprints is highly dependent on the orientation fields’ estimation and singular points detection algorithms. Later, Arivazhagan et al. (2007) suggested a fingerprint verification system utilizing Gabor wavelets. Yazdi and Gheysari (2008) utilized co-occurrence matrices to verify the fingerprint image.

Preprocessing Fingerprint Enhancement

Image improvement of fingerprint samples is very important for FV to function correctly. The efficiency of the image of the imprint sample is influenced by noise in ridges produced by the under-inked region, varying the fingerprint attributes because of skin elasticity, where splits are from dry skin and wounds may cause ridge discontinuities. To preserve the high accuracy of the FV system, two procedures utilized in the unique mark recognition basis (STFT) analysis suggested by Chikkerur et al. (2007) are utilized here for image enhancement of fingerprints and procedures. The system can be shortened as follows: The image of the fingerprint is split into overlapping squares. Then STFT analysis is performed. The test productions are images of ridge orientation O(x,y) and ridge hesitation F(x,y). The next stage O(x,y) image represents smoothing of the orientation. For improvement, each overlapping square B(x,y) in the image has been rebuilt for the sample by creating improved blocks B0(x,y).

Feature Extraction Methods

The fingerprint sample illustrates a scheme of oriented texture and contains important data. This scheme utilizes an ensemble of dichotomizes to combine features through several measures and feature extraction methods, commanding lower cost and accurate FV. The approach utilizes multi-feature extraction at different measure techniques for classification of preexisting extracted feature systems, creating a large set of features (Ebrahim 2017b; Ebrahim and Sulong 2014). The performance of fingerprint images is realized by applying two methods based on singular point detection and fingerprint ridge thinning, as discussed in the next section.

Singular Point Detection Technique

The image of a fingerprint is made up of the design of ridges and valleys that is a copy of the individual imprints. The reference point is the point with extreme curvature on the convex ridge, and to define the single reference point reliably for all varieties of fingerprint samples, the orientation area, which characterizes the local orientations along the prevalent orientation, is utilized to distinguish the reference point. The reliability can also be calculated applying the method projected by (Ebrahim 2017a) and (Kaas and Witkin 1987). Figure 1 shows the block design for detecting the singular point and the implementation.
Fig. 1

Singular point block method diagram (Kaas and Witkin 1987)

Orientation Area

The first stage of finding the singular point is computing the orientation area. The orientation field is critical for the computation of the reliability. The accuracy of the reliability robustly depends on the accuracy of the orientation area values calculated. Below are the steps for orientation calculation:

The fingerprint sample is split into a non-overlapping block of size (W × W). In this paper, W is set to 16.

The horizontal and vertical gradients Gx(x, y) and Gy(x, y) at each pixel (x,y) respectively are calculated applying simple gradient factors (Gonzalez and Woods 2008).

The ridge orientation of each pixel (x,y) within a W × W window at points [xi,yj] is calculated as follows (Ratha et al. 1995):
$$ \theta (x.y)=\frac{1}{2}{\tan}^{-1}\frac{2{G}_{\mathrm{xy}}}{G_{\mathrm{yy}-{G}_{\mathrm{yy}}}}. $$
Because of noise, an inclined ridge, valley structures, and poor gray value, a low-pass filter can be applied to regulate the incorrect local ridge orientation. However, to achieve the low-pass filtering, the orientation image must be transformed into a continuous vector area, and the Gaussian low-pass filter can be used as follows:
$$ {\displaystyle \begin{array}{ll}{\Phi}_y^{.}(x.y)=& {\sum}_{u=-1}^1{\sum}_{v=-1}^1l(u.v)\\ {}& \times {\Phi}_y\left(x- ul.y- vl\right)\end{array}} $$
where l is a 2-directional low-pass filter with a complete unit.

The orientation area can be representing in this section.


Meanwhile the singular point has the maximum curvature. It can be found by calculating the strength of the reliability peak applying the following equation:
$$ Y=1-\frac{Y_{\mathrm{min}}}{Y_{\mathrm{min}}}. $$
Figure 2 displays the orientation area reliability and the singular point value in the center. The reliability rate of the singular point is = 0, but the value of background is also zero. However, there is a contour around the singular point of the reliability value of the contour between < 0.5 and > 0, and this is the area of interest.
Fig. 2

The detecting of singular point by (a) Orientation field on original fingerprint, (b) Both singular point and orientation field reliability within the contour, (c) Reliability image map to three singular points, (d) Region of interest after the segmentation, (e) Singular point contour after thinning, (f) Filled singular point contour, (g) Singular point location on the original fingerprint

Singular Point Location

After calculating the orientation area reliability, the next stage is to detect the singular point place. This can be achieved by the following stages:
  1. (i)

    The orientation field reliability needs to be segmented into two distinct regions. The region of interest contains the values > 0 and < 0.5.

    The result of the segmentation can be seen in Fig. 2.

  2. (ii)

    Thinning is the process of adjusting the width of contents of the image to one pixel while preserving the extent and connectivity of the original shape.

  3. (iii)

    After thinning, all pixels will be removed so that the contour shrinks to a connected ring halfway between each hole and the outer boundary, and the rest will shrink to a pixel that will be removed. Figure 2 shows the singular point contour without any noise.

  4. (iv)
    The singular point contour is well defined, and to determine the location of the singular point, the contour is filled using the morphological hole filling equation (Gonzalez and Woods 2008):
    $$ F\ \left(\mathrm{x}.\mathrm{y}\right)=\Big\{{}_0^{1-I(x.y)}, $$

    where “(x, y)” represents the border of I; otherwise, H is equal to the input image I, with holes filled.

  5. (v)

    Here, the singular point pixel can be found by performing the shrinking process for the singular point contour. Figure 2 shows the singular point pixel after applying the shrinking method, the position of the singular point on the original fingerprint, and Singular point pixel.


Fingerprint Thinning Minutiae Marking

Thinning is a procedure by which the sizes of the ridges are reduced. In each scan of the full imprint image of a fingerprint, every sample square represents (16 × 16) pixels (Maio and Maltoni 1997). This image along with other data will be recorded into the dataset (Jain et al. 1999). After obtaining this input file, it will undergo binarization. Usually, a vision of the distinctive mark will be obtained.


Normalization is a procedure for fingerprint verification. Samples of fingerprints do not come in the same dimensions. Because of this, the samples need to be aligned suitably to confirm an overlap of the common area in the two samples of imprint by the orientation of the image to zero at the reference point (Gonzalez and Woods 2008). Once all the features of each fragment have been extracted, then the feature values are normalized in the interval [0 1]. Normalization of features is very important because if the values of different features are in different ranges then the higher values dominate the lower values. Thus, the normalization technique makes the feature values in the same scales and ranges. The image of the fingerprint is split into a non-overlapping set of size W × W (for each 16 × 16) and the orientation that matches the most controlling orientation of the block (Ebrahim et al. 2018; Ebrahim and Ashoor 2018; Ebrahim 2018). After the process, the features of ridges in the fingerprint are represented in black and furrows are represented in white.

Multi-feature Extraction

There is a pressing need to develop a fingerprint verification system with software-based minutiae extraction and statistical features. The proposed approach is to extract the most suitable multi-features for classification. In this study, we combine a set of the strongest features. The effectiveness of the projected method has been evaluated through a comparison with several existing techniques for multi-features.

Tests have been executed depending on standard datasets. However, more accuracy lowers the popularization capabilities. In fact, the background minutiae extraction and statistical features have already managed to detect the differences in fingerprints, and statistical attributes are utilized by other FV systems, for example (Yazdi and Gheysari 2008). In this research, multi-features are extracted from different parts of the fingerprint, therefore increasing its selective power. There are also many outliers in each class, which are circled. The outlier is an observation that carries an irregular distance from its neighboring feature values that will result in misclassifications. Hence, it is observed that minutiae extraction or statistical features alone inherit some weaknesses, such as overlapping and outliers, which reduce their discriminating abilities.

Fingerprint Verification

Matching is an important part of fingerprint verification. It compares two features and returns a likeness score to indicate how comparable the two participating fingerprints are. The comparison depends on discovery of the Euclidean distance between the attribute of the conforming fingerprints.

The EER for experiment was computed by the FVC2002. In Table 1, the databases were split into four databases: DB1, DB2, DB3, and DB4. Each dataset contains 800 fingerprint samples collected from 100 persons, and each one is eight impressions. Sets of tests were carried out for each database, and the protocol is shown in Table 1. Yang et al. (Yang and Park 2008) used the tessellated invariant moment feature, Ross et al. (2003) used minutiae and ridge map features, Jin et al. (2004) used integrated wavelet and Fourier–Mellin invariant framework with four multiple training WFMT features, Amornraksa et al. (Amornraksa and Tachaphetpiboon 2006) used the DCT feature, Khalil et al. (2010) used statistical descriptors, and Flores et al. (2017) used Delaunay triangulations.
Table 1

EER (%) evaluation of the current techniques







Yang et al. (2008)






Ross et al. (2003)






Jin et al. (2004)






Amornraksa et al. (2006)






Khalil et al. (2010)






Flores et al. (2017)







Low-accuracy fingerprint images require multi-features to expand distinction. The process for FV comprises five sections: singular point detections, fingerprint ridge thinning, normalization, multi-feature extraction, and fingerprint verification. Moreover, minutiae extraction and the four statistical descriptors characterize the fingerprint texture. Each process plays an important role in fingerprint verification. Hence, to facilitate the accomplishment of such ambitious goals in the near future, researchers ought to perform statistical tests and minutiae extraction of fingerprint textures.



  1. Amornraksa, T., Tachaphetpiboon, S.: Fingerprint recognition using DCT features. Electron. Lett. 42(9), 522–523 (2006)CrossRefGoogle Scholar
  2. Arivazhagan, S., Arul Flora, T.G., Ganesan, L.: Fingerprint verification using Gabor co-occurrence features. Int. Conf. Intell. Multimedia. Appl. Proceedings of ICCIMA, 2, 281–285 (2007)Google Scholar
  3. Bazen, A.M., Gerez, S.H.: Systematic methods for the computation of the directional fields and singular points of fingerprints. IEEE Trans. Pattern Anal. Mach. Intell. 24(7), 905–919 (2002)CrossRefGoogle Scholar
  4. Chikkerur, S., Cartwright, A.N., Govindaraju, V.: Fingerprint Enhancement using STFT Analysis. Pattern Recog. 40, 198–211 (2007)CrossRefGoogle Scholar
  5. Dass, S.C., Jain, A.K.: Fingerprint-based recognition. Technometrics. 49, 262–276 (2007)MathSciNetCrossRefGoogle Scholar
  6. Dodis, Y., Osrovsky, R., Reyzin, L., Smith, A.: Fuzzy extractors: how to generate strong keys from biometrics and other noisy data. LNCS. 3027, Springer, 523–540 (2006)MathSciNetzbMATHGoogle Scholar
  7. Ebrahim, A.Y.: Classification of Arabic autograph as genuine and forged through a combination of new attribute extraction techniques. J Univ Babylon. 25(5), 1873–1885 (2017a)Google Scholar
  8. Ebrahim, A.Y.: Detection of breast cancer in mammograms through a new features and decision tree based, classification framework. J. Theor. Appl. Inf. Technol. 95(12), 6256–6267, ISSN: 1992-8645 (2017b)Google Scholar
  9. Ebrahim, A.Y.: A new model of Arabic handwritten recognition using combination between DWT with data reduction method. J. Theor. Appl. Inf. Technol. 96, 6376–6387. (1992–8645)- (2018)Google Scholar
  10. Ebrahim, A.Y., Ashoor, A.S.: Tumor classification using enhanced hybrid classification methods and segmentation of MR brain images. ARPN J. Eng. Appl. Sci. 3(20), 8270–8278 (2018)Google Scholar
  11. Ebrahim, A.Y., Sulong, G.: Offline handwritten signature verification using back propagation artificial neural network matching technique. JATIT LLS. 65(3), 790–800 (2014)Google Scholar
  12. Ebrahim, A.Y., Kolivand, H., Rehman, A., Rahim, M.S.M., Saba, T.: Features selection for offline handwritten signature verification: state of the art. Int J Comput Vision Robot. 8(6), 606–622 (2018)CrossRefGoogle Scholar
  13. Flores, G.M., Torres, G., Garcia, M.L.: Fingerprint verification methods using delaunay triangulations. Int. Arab J. Inf. Technol. 14(3), 346–354 (2017)Google Scholar
  14. Gonzalez, R.C., Woods, R.: Digital Image Processing. Prentice Hall, Upper Saddle River (2008)Google Scholar
  15. Igaki, S., Shinzaki, T., Yamagishi, F., Ikeda, H. Yahagi, H.: Minutia Extraction in Fingerprint Identification. US Patent No. US5109428 A (1992)Google Scholar
  16. Jain, A., Hong, L., Bolle, R.: On-line fingerprint verification. Pattern Anal. Mach. Intell. 19, 302–313 (1997)CrossRefGoogle Scholar
  17. Jain, L.C., Halici, U., Hayashi, I., Lee, S.B., Tsutsui, S.: Intelligent Biometric Techniques in Fingerprint and Face Recognition. The CRC Press, Boca Raton (1999)Google Scholar
  18. Jea, T.-Y., Govindaraju, V.: A minutia-based partial fingerprint. Recognition system. Pattern Recogn. 38, 1672–1684 (2005)CrossRefGoogle Scholar
  19. Jin, A.T.B., Ling, D.N.C., Song, O.T.: An efficient fingerprint verification system using integrated wavelet and ourierMellin invariant transform. Image Vis. Comput. 22, 503–513 (2004)CrossRefGoogle Scholar
  20. Kaas, M., Witkin, A.: Analyzing oriented patterns. Comp: Vision Graphics Image Process. 37, 362–385 (1987)Google Scholar
  21. Khalil, M.S., Mohamad, D., Khan, M.K., Al-Nuzaili, Q.: Fingerprint verification using statistical descriptors. In: Journal Digital Signal Processing, Elsevier, Academic Press, Inc. Orlando, FL, USA 20(4), 1264–1273 (2010)Google Scholar
  22. Khan, M.K.: Fingerprint biometric-based self-authentication and deniable. Authentication schemes for the electronic world. IETE Tech. Rev. 26, 191–195 (2009)CrossRefGoogle Scholar
  23. Maio, D., Maltoni, D.: Direct gray-scale minutiae detection in finger- prints. IEEE Trans. Pattern Anal. Mach. Intell. 19, 27–40 (1997).  https://doi.org/10.1109/34.566808CrossRefGoogle Scholar
  24. Maltoni, D., Cappelli, R.: Advances in fingerprint modeling. Image Vis. Comput. 27(3), 258–268 (2009)CrossRefGoogle Scholar
  25. Maltoni, D., Maio, D., Jain, A.K., Prabhaka, S.: Handbook of Fingerprint. Recognition., Springer-Verlag New York, Inc., Secaucus, NJ (2003)Google Scholar
  26. Rajkumar, R., Hemachandran, K.: A review on image enhancement of fingerprint using directional filters. Assam Univ. J. Sci. Technol. 7(2), 52–57 (2011)Google Scholar
  27. Ross, A., Jain, A., Reisman, J.: A hybrid fingerprint matcher. Pattern Recogn. 36, 1661–1673 (2003)CrossRefGoogle Scholar
  28. Wu, C., Tulyakov, S., Govindaraju, V.: Image Quality Measures for Fingerprint Image Enhancement. Multimedia Content Representation, Classification and Security, pp. 215–222. Springer, Berlin (2006)CrossRefGoogle Scholar
  29. Yang, J.C., Park, D.S.: A fingerprint verification algorithm using tessellated invariant moment features. Neurocomputing. 71, 1939–1946 (2008)CrossRefGoogle Scholar
  30. Yazdi, M., Gheysari, K.: A new approach for the fingerprint classification based on gray-level co-occurrence matrix. World Acad. Sci. Eng. Technol. 47, 313–316 (2008)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  1. 1.University of BabylonBabylonIraq
  2. 2.Department of Computer ScienceLiverpool John Moores UniversityLiverpoolUK