<![CDATA[Palmprint recognition is a promising biometric method due to the stability and uniqueness of its texture patterns. This study proposes the Warkac method (Wavelet-Wiener-Gabor-KPCA-Cosine), a systematic integration of image processing and feature extraction techniques to improve the robustness and accuracy of palmprint recognition systems. The process starts with wavelet decomposition and Wiener filtering for noise reduction, followed by detail weighting to enhance dominant features. Feature extraction is carried out using a 7x5 Gabor filter, with dimensionality reduction by Kernel Principal Component Analysis (KPCA). Matching is performed using cosine similarity, which efficiently distinguishes low-dimensional biometric features. Evaluations conducted on three public databases (PolyU, IITD, CASIA) with various matching and dimensionality reduction methods show that KPCA–Cosine delivers the best performance, achieving a verification rate of 99.455% and EER of 0.00546, followed closely by LDA–Cosine. Hausdorff and Ndistance methods perform poorly, with verification rates below 55%. This study demonstrates that the proper integration of filtering and non-linear transformation techniques can significantly enhance palmprint recognition performance under diverse input conditions.]]>
Copyrights © 2025
