Minutiae-based palm photo recognition using deep neural networks
Issued Date
2026-02-01
Resource Type
ISSN
09521976
Scopus ID
2-s2.0-105024328861
Journal Title
Engineering Applications of Artificial Intelligence
Volume
165
Rights Holder(s)
SCOPUS
Bibliographic Citation
Engineering Applications of Artificial Intelligence Vol.165 (2026)
Suggested Citation
Khodadoust J., Monroy R., Medina-Pérez M.A., Marasco E., Kusakunniran W. Minutiae-based palm photo recognition using deep neural networks. Engineering Applications of Artificial Intelligence Vol.165 (2026). doi:10.1016/j.engappai.2025.113366 Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/114558
Title
Minutiae-based palm photo recognition using deep neural networks
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
Mobile biometrics has witnessed a surge to properly secure smartphones, offering functionalities such as device unlocking and access control to sensitive accounts. Contactless fingerprints and palmprints have gained increasing prominence due to their convenience and hygiene advantages. Although palmprints hold great potential for enhancing biometric authentication, recognizing palm photos captured by smartphone cameras poses several challenges, including photometric distortions and image blurring. Existing methods utilize the spatial domain for deblurring finger photos; however, their efficacy is limited in the case of palm photos due to the presence of creases. These methods often introduce new artifacts, such as new ridge lines. In this study, we introduce a generative adversarial network (GAN) model, FreqGAN, which leverages the frequency domain and deep neural networks (DNNs) to address blurring and enhance palm photos. This model significantly eliminates creases, leading to a reduction in false minutiae. We also discuss a deep convolutional neural network (DCNN) model, MinuExtNet, for minutiae extraction from each block, which is then mapped back to the original palm photo. Additionally, we enhance a minutiae feature extraction model originally designed for contactless fingerprints, adapting it for palm photos and naming it MinuFeExtNet. MinuFeExtNet ensures robustness against variations in the distance of palms from smartphone cameras. Our experiments, conducted on palm photos extracted from hand photos in four hand photo databases, demonstrate the superior performance of utilizing the frequency domain, GAN, and DNNs for palm photo deblurring and enhancement, minutiae extraction, and minutiae feature extraction in palm photo recognition, compared to existing methods.