Discrimination between computer generated and natural human faces based on asymmetry information

The recent development of information and communication technology has made computer software able to create highly realistic multimedia contents that can be, for human, impossible to distinguish from the natural ones. This fact leads to the need for tools and techniques that can reliably discriminate between natural and computer generated multimedia data in forensics applications. In this paper, we focus on the specific class of images containing faces, since we consider critical to be able to discriminate between photographic faces and the photorealistic ones. To this aim, we present a new geometric-based approach relying on face asymmetry information. Experimental results show that asymmetry information could be used as a hint to tackle this problem without requiring classification tools and training or combined with state-of-the-art approaches to improve their performances.