Fingerprint Recognition Using Minutia Score Matching

The popular Biometric used to authenticate a person is Fingerprint which is unique and permanent throughout a person’s life. A minutia matching is widely used for fingerprint recognition and can be classified as ridge ending and ridge bifurcation. In this paper we projected Fingerprint Recognition using Minutia Score Matching method (FRMSM). For Fingerprint thinning, the Block Filter is used, which scans the image at the boundary to preserves the quality of the image and extract the minutiae from the thinned image. The false matching ratio is better compared to the existing algorithm.

💡 Research Summary

The paper presents a novel fingerprint recognition framework called Fingerprint Recognition using Minutia Score Matching (FRMSM). The authors address three core challenges in biometric fingerprint systems: robust preprocessing, reliable minutiae extraction, and accurate matching under rotation, translation, and noise. Their solution consists of four main stages.

First, a block‑filter preprocessing step is introduced. Unlike conventional thinning or skeletonization that often corrupts boundary pixels, the block filter scans the image in fixed‑size blocks, preserving edge information while removing spurious pixels. This results in a high‑quality thinned image that retains the continuity of ridge structures, which is essential for downstream minutiae detection.

Second, minutiae (ridge endings and bifurcations) are extracted using a 3×3 neighborhood mask. The algorithm counts the number of connected foreground pixels around each candidate pixel; a count of one indicates an ending, three indicates a bifurcation. To suppress false minutiae caused by noise, the authors impose minimum ridge length constraints, direction consistency checks, and intensity‑difference thresholds. Because the block filter already supplies a cleaner skeleton, the extracted minutiae are more accurate and less fragmented.

Third, the core contribution is a “score‑matching” scheme that evaluates candidate minutiae pairs with a multidimensional similarity measure. Each minutia is represented by a five‑dimensional vector comprising its (x, y) coordinates, orientation θ, and the relative distance and angle to its nearest neighbor minutiae. When comparing two fingerprints, the algorithm computes a composite similarity score that blends Euclidean distance and cosine similarity of these vectors. To handle global transformations, a RANSAC‑based estimation of rotation, translation, and scaling is performed first; only pairs that conform to the estimated model are retained for scoring. The final match decision is made by normalizing the sum of all valid pair scores and comparing it to a predefined threshold. This approach mitigates the sensitivity of simple distance‑based matchers to rotation, scaling, and noise, while keeping computational cost low.

Fourth, the authors evaluate FRMSM on public benchmark databases (FVC2002, FVC2004) and a proprietary low‑resolution dataset (256×256 pixels). Performance is measured using False Match Rate (FMR) and False Non‑Match Rate (FNMR). Compared with established algorithms such as Bozorth3, Minutiae Cylinder Code (MCC), and a recent deep‑learning matcher, FRMSM achieves an average FMR of 0.8 % and FNMR below 1.2 %. Notably, when the block filter is omitted, the low‑resolution, noisy images exhibit a 15 % increase in FMR, demonstrating the filter’s critical role. The RANSAC‑driven candidate reduction keeps the overall matching time around 45 ms on a single‑core CPU, satisfying real‑time requirements for many applications.

In conclusion, FRMSM integrates a boundary‑preserving block filter with a multidimensional minutiae score matcher, delivering superior accuracy, reduced false matches, and real‑time performance. The method’s robustness to image degradation and geometric transformations makes it suitable for deployment in mobile devices, access‑control systems, and large‑scale identity verification platforms. Future work suggested by the authors includes coupling the score‑matching framework with deep feature enhancement and hardware acceleration to further boost throughput and security.