Non-cooperative iris recognition: Issues and trends

To date, no research effort has produced a machine able to covertly recognize human beings. Contrary to popular belief, such automata are confined to science fiction, although it's not hard to anticipate the potential impact that they would have in the security and safety of modern societies (forensics and surveillance). Among the research programs that pursuit such type of biometric recognition, previous initiatives sought to acquire data from moving subjects, at long distances and under uncontrolled lighting conditions. This real-world scenario brings many challenges to the Pattern Recognition process, essentially due to poor quality of the acquired data. Several programs now seek to increase the robustness to noise of each phase of the recognition process (detection, segmentation, normalization, encoding and matching). This paper addresses the feasibility of such extremely ambitious type of biometric recognition, discusses the major issues behind the development of this technology and points some directions for further improvements.