Proposal of a Comparative Framework for Face Super-Resolution Algorithms in Forensics

Antonio Salguero-Cruz, Pedro Latorre-Carmona, César Ignacio García-Osorio

Abstract: The number of images acquired by electronic devices is growing exponentially. This is partly due to an easier and more extended access to high-tech portable acquisition devices. These devices could record people of interest, either because they were involved in some type of criminal act, or they could be missing people, so their identification (and specifically the identification of their faces) is very relevant. On the other hand, image quality sometimes does not allow a positive identification of these individuals. This paper presents a framework for the comparative analysis of face super-resolution (hallucination) algorithms in forensics. The super-resolved images could be used to help identify a person of interest, and subsequently be admitted by the competent judicial authority in a criminal process, maintaining, like the rest of evidence, the so-called chain of custody, as required by the Spanish law and other similar legal systems.