<![CDATA[Fingerprint recognition has been widely adopted as a biometric authentication method because it offers a unique and reliable means of verifying identity. However, the accuracy of fingerprint recognition systems depended on the quality of the acquired images, which were often affected by noise, low contrast, and unclear ridge–valley patterns that hindered minutiae extraction. This study addressed that problem by applying the Hong method, which enhanced fingerprint images through normalization, ridge orientation estimation, ridge frequency estimation, and Gabor filtering, prior to minutiae extraction and template matching. The method was implemented in MATLAB R2023b and tested on grayscale fingerprint images from the SOCOFing dataset. Testing was conducted on ten classes of fingerprint data, with recognition accuracy evaluated through identity matching and further assessed using a confusion matrix built from twenty test images, consisting of ten registered and ten unregistered users, across four threshold values. The results showed that the Hong method achieved a recognition accuracy of only 10%, even though eight of ten test images were classified as recognized based on similarity scores exceeding the threshold, indicating that the system was prone to false acceptance at lower thresholds. Analysis of the False Acceptance Rate and False Rejection Rate revealed a trade-off, with the Equal Error Rate occurring at a threshold of 0.5495 and a value of 0.4367. These findings indicated that although the Hong method improved fingerprint image quality, the overall identification performance remained low, and careful threshold selection was required to balance false acceptance and false rejection in fingerprint authentication systems.]]>
Copyrights © 2026
