Matching vein patterns from color images for forensic investigation
Kong, Adams Wai-Kin
Date of Issue2012
IEEE International Conference On Biometrics: Theory, Applications And Systems (5th : 2012 : Arlington, Virginia, US)
School of Computer Engineering
Child sexual abuse is a serious global problem and has gained public attention in recent years. Due to the popularity of digital cameras, many perpetrators take images of their sexual activities with child victims. Traditionally, it was difficult to use cutaneous vascular patterns for forensic identification, because they were nearly invisible in color images. Recently, this limitation was overcome using a computational method based on an optical model to uncover vein patterns from color images for forensic verification. This optical-based vein uncovering (OBVU) method is sensitive to the power of the illuminant and does not utilize skin color in images to obtain training parameters to optimize the vein uncovering performance. Prior publications have not included an automatic vein matching algorithm for forensic identification. As a result, the OBVU method only supported manual verification. In this paper, we propose two new schemes to overcome limitations in the OBVU method. Specifically, a color optimization scheme is used to derive the range of biophysical parameters to obtain training parameters and an automatic intensity adjustment scheme is used to enhance the robustness of the vein uncovering algorithm. We also developed an automatic matching algorithm for vein identification. This algorithm can handle rigid and non-rigid deformations and has an explicit pruning function to remove outliers in vein patterns. The proposed algorithms were examined on a database with 300 pairs of color and near infrared (NIR) images collected from the forearms of 150 subjects. The experimental results are encouraging and indicate that the proposed vein uncovering algorithm performs better than the OBVU method and that the uncovered patterns can potentially be used for automatic criminal and victim identification.
DRNTU::Engineering::Computer science and engineering
© 2012 IEEE.