Adaptive Palmprint ROI Extraction Using Contour-Based Approach with PCA Alignment

In the evolving field of palmprint recognition, extracting the region of interest (ROI) accurately is pivotal, especially when dealing with images of varied positions and rotations. Past approaches typically employed a constant size square area, typically at the palm's center, leading to potential inaccuracies due to rotation or misalignment. This paper introduces an improved method for ROI extraction in palmprint images using an enhanced contour-based algorithm with Principal Component Analysis (PCA) alignment. This method successfully adapts to the unique characteristics of each palm, ensuring accurate ROI extraction even with varied hand positioning and rotation. The method was rigorously tested across three different palmprint databases: PolyU, IIT Delhi, and CASIA, demonstrating high accuracy rates of 99.07%, 97.5%, and 98.04% respectively. Even under artificially induced rotation, the method maintained substantial performance on PolyU, IIT Delhi and CASIA database with accuracy rates of 96.7%, 88.26% and 96.32% respectively. The results of this study underscore the potential of this method in advancing palmprint recognition systems, contributing to their applicability in broader contexts and varied conditions.