Source-linked AI summary
Minutiae Extraction from Fingerprint Images - a Review
Roli Bansal, Priti Sehgal, Punam Bedi
TL;DR
Fingerprint images can be degraded by skin and impression conditions, complicating accurate minutiae extraction and rejection of false minutiae. This paper reviews extraction techniques and classifies them by image type and processing choices; it concludes that local minutia-validity criteria and statistical study of fingerprint uniqueness warrant further work.
Problem
Skin and impression conditions can corrupt ridge structures, while accurate minutiae extraction and rejection of false minutiae are important for fingerprint matching.
Source: §1 · p. 1Method
Results
The review distinguishes approaches by input image type, binarization and segmentation, thinning requirements, and post-processing effort.
Source: §5 · p. 9Takeaways & Limitations
The conclusion points to defining local criteria for minutia validity and studying degrees of freedom in fingerprint populations to understand minutiae uniqueness.
Source: §5 · p. 9The paper identifies defining local criteria for minutia validity and studying statistical uniqueness across fingerprint populations as areas needing further work.
Source: §5 · p. 9
Abstract
from arXiv · showhide
Fingerprints are the oldest and most widely used form of biometric identification. Everyone is known to have unique, immutable fingerprints. As most Automatic Fingerprint Recognition Systems are based on local ridge features known as minutiae, marking minutiae accurately and rejecting false ones is very important. However, fingerprint images get degraded and corrupted due to variations in skin and impression conditions. Thus, image enhancement techniques are employed prior to minutiae extraction. A critical step in automatic fingerprint matching is to reliably extract minutiae from the input fingerprint images. This paper presents a review of a large number of techniques present in the literature for extracting fingerprint minutiae. The techniques are broadly classified as those working on binarized images and those that work on gray scale images directly.
1. Introduction
Fingerprints are widely used for personal identification, and minutiae are local ridge discontinuities used to determine fingerprint uniqueness. Because skin and impression conditions can degrade images, the paper reviews extraction techniques and the enhancement steps used before extraction.
1. Introduction
- Minutiae are local ridge discontinuities—ridge endings and bifurcations—that are used to determine fingerprint uniqueness.A good-quality image has around 40 to 100 minutiae.Source: §1 · p. 1
- Noise from scars, humidity, dirt, and non-uniform contact can obscure ridge structures and reduce fingerprint image quality.Accurate fingerprint matching depends largely on those ridge structures.Source: §1 · p. 1
- Image enhancement is used to reduce noise and clarify ridges against valleys before minutiae extraction, while latent-print matching can combine manually marked and automatically extracted minutiae.The paper reviews enhancement approaches including normalization and Gabor filtering, directional Fourier filtering, and fuzzy logic and neural-network methods.Source: §1 · p. 2
2. Fingerprint Features
Fingerprint features range from global ridge-flow patterns and landmarks to local minutiae. The section describes how these feature levels differ and why minutiae are widely used despite extraction challenges in poor-quality images.
2. Fingerprint Features
- Level 1 features describe macro ridge-flow shape, Level 2 features are minutiae used for recognition, and Level 3 features are pores that complement Level 2 uniqueness.The paper groups fingerprint features into these three levels.Source: §2 · p. 2
2.1 Global Ridge Pattern
- Global ridge patterns include pseudo-parallel flows and high-curvature flows around cores or deltas, alongside singular points, orientation maps, and frequency maps.Singular points are orientation-field discontinuities; cores and deltas are landmarks used for registration and classification. Orientation maps represent local ridge-valley direction, while frequency maps represent reciprocal ridge spacing perpendicular to that direction.Source: §2.1 · p. 2, Figure 2 · p. 2, Figure 3 · p. 2
2.2 Local Ridge Pattern
- Minutiae are local ridge-structure discontinuities; ridge endings and bifurcations are the most commonly used types, and representations include their location and orientation.Other minutia types can be seen as combinations of endings and bifurcations; minutiae are described as relatively stable across contrast, resolution, and global distortion, but difficult to extract from poor-quality images.Source: §2.2 · pp. 2–3, Figure 4 · p. 2, Figure 5 · p. 3
3 What is Fingerprint Recognition?
Fingerprint recognition includes enrollment, verification, and identification, with verification comparing one-to-one and identification searching across a database. Matching techniques use minutiae, ridge patterns, image correlation, or whole-image features, each with distinct trade-offs.
3 What is Fingerprint Recognition?
- Verification compares a fingerprint one-to-one, while identification searches a database of user templates for a match.The paper also identifies enrollment as a recognition sub-domain.Source: §3 · p. 3
3.1 Techniques for Fingerprint Matching
- Minutiae-based matching compares local terminations and bifurcations, while pattern matching uses ridge series and is less affected by wear but sensitive to finger placement and storage demands.The paper describes minutiae-based matching as the backbone of current fingerprint recognition products.Source: §3.1 · p. 3
- Correlation compares a rotated and shifted input image with a template, but nonlinear distortion, changing skin conditions, and computational cost limit direct application.Local or block-wise correlation, more sophisticated measures, and Fourier-domain correlation are noted as ways to address these issues.Source: §3.1 · p. 3
- Image-based techniques match global features of the whole fingerprint and are described as an emerging approach useful for some intractable problems of minutiae-based matching.The supplied passage does not specify those problems further.Source: §3.1 · p. 3
4. Minutiae Extraction
The review organizes minutiae extraction into methods for binarized images and methods operating directly on gray-scale images. It surveys alternatives within both groups, including methods that avoid thinning and approaches that validate or filter candidate minutiae.
4. Minutiae Extraction
- Accurate minutiae extraction is needed for feature-based fingerprint matching, and the reviewed methods divide broadly into binarized-image and direct gray-scale approaches.Binary methods are further divided into unthinned and thinned image methods.Source: §4 · p. 4
4.1.1 Unthinned Binarized images
- Unthinned binary methods avoid skeletonization, which is noise-sensitive and can produce spurious minutiae, but still require post-processing to reject false candidates.The review covers chaincode, run-based, and ridge-flow or local-pixel approaches.Source: §4.1.1 · p. 4
- Chaincode methods trace ridge contours and detect minutiae from significant turns, using a threshold to distinguish turns and their direction.The contour is traced counterclockwise; significant left and right turns correspond to ridge endings and bifurcations, respectively.Source: §4.1.1.1 · p. 5
- Run-based extraction identifies termination and bifurcation candidates from adjacent scan-line runs, then applies geometric checks because characteristic runs can be false minutiae.Runs with one adjacent run on either neighboring scan line indicate termination candidates; two adjacent runs indicate bifurcation candidates.Source: §4.1.1.1 · p. 5, §4.1.1.2 · p. 6
4.1.2.1 Crossing Number
- Crossing Number classifies a thinned ridge pixel from its eight-neighbourhood, with values of one and three indicating endings and bifurcations; noisy images and thinning artifacts motivate post-processing.False structures include spurs, holes, triangles, and spikes, which can generate false endings or bifurcations.Source: §4.1.2.1 · pp. 6–7
4.1.2.2 Morphology based
- Morphology-based methods apply structuring elements and the Hit-or-Miss transform to a thinned image to detect minutiae, with the reviewed algorithm reported to reduce post-processing effort.Ridge endings have one neighbor in a 3×3 neighbourhood; bifurcations have three nonadjacent neighbors.Source: §4.1.2.2 · pp. 7–8
4.2 Minutiae Extraction from Gray-Level images
- Gray-level methods can avoid information loss and time costs from binarization and thinning, which may also introduce spurious minutiae or perform poorly on low-quality images.The review describes ridge-flow tracing, including dynamic step selection, and simultaneous tracking of a central ridge and neighboring valleys.Source: §4.2 · p. 8, §4.1.2.2 · p. 8
- Other direct gray-level approaches extract minutiae from Gabor phase, use fuzzy DARK and BRIGHT levels, or select points with responses above a symmetry-filter threshold.The symmetry approach uses a 9 × 9 window; the fuzzy method uses pixel averages in a 5 × 5 test window and its surrounding border.Source: §4.2 · p. 8, §4.1.2.2 · p. 8, §4.2.2 · p. 9
5. Conclusions
The review compares minutiae-extraction techniques by image type and processing choices, while emphasizing image quality and local criteria for validating minutiae. It also identifies broader minutiae definitions and statistical uniqueness as directions for further study.
5. Conclusions
- Low-quality images need preprocessing to increase contrast and reduce noise, but enhancement can also make noisy pixels produce spurious minutiae.The review calls for stronger local criteria to establish minutia validity.Source: §5 · p. 9
- The reviewed approaches differ by input type, binarization and segmentation, thinning requirements, and post-processing effort.Source: §5 · p. 9
- Further work could extend minutiae definitions to trifurcations, islands, bridges, and spurs for matching and more sophisticated identification models.Source: §5 · p. 9
- Further study of fingerprint minutiae statistics could determine population degrees of freedom and clarify their statistical uniqueness.Source: §5 · p. 9