Infrared face recognition

Konuk, Uğur
Face recognition is a leading biometrics technique that fulfills the increasing need to identify a person in today’s world. Face recognition also has broad range of utilization, such as commercial and law enforcement applications. That is the reason why it still gathers a lot of attention and is an active research topic. Nevertheless visible spectrum face recognition algorithms are not free of challenges. Illumination, pose, expression variances and existence of facial disguises still degrade the performance of recognition systems that utilize visible spectrum face images. That fact paved the way for infrared face recognition to emerge as an alternative solution to those limitations. In this thesis work, first a review of the infrared face recognition algorithms is presented. Four different methods, one of which is proposed for the first time in this thesis work, are implemented and compared in terms of recognition performance. The proposed method investigates the utilization of vascular network method with blood perfusion transform of the thermal data. Experiments are performed on Terravic and IRIS database.