Palm Vein Image Quality Assessment through Natural Scene and Texture Statistics

Palm vein biometrics has emerged as a highly secure and contactless biometric technology. However, various factors reduce the image quality during the acquisition process, limiting the biometric recognition performance, such as noise, irregular contrast, and high levels of blurring. Usually, palm vein recognition methods include an image preprocessing stage to enhance the detail of vascular patterns and mitigate the impact of low-quality images. This article aims to evaluate the image quality of the most referenced palm vein databases in the state-of-the-art. The evaluation was performed using five quantitative metrics: NIQE, BRISQUE, PIQE, GLCM, and UMAP projections. The experimentation showed that NIQE presents a similar mean and negligible variance, while BRISQUE and PIQE exhibit notable variability in quality scores. The GLCM metric and UMAP projections allowed visualizing the variability among the databases. The conducted evaluation highlights the need to establish quality standards and image enhancement techniques for palm vein images to maximize the effectiveness of biometric systems.