Facial image comparisons of morphed facial imagery

ABSTRACT

Facial image morphing is a technique whereby two or more faces are blended together to create a new face (the morph). Apps and open source software are now freely available to create morphs. While morph creation is predominantly for entertainment value, it is also a method being used to create facial imagery for use on fraudulent identity documents. This study sought to answer two questions. First, would participants accept a morph persona (an identity created using only morphed facial imagery) as a credible identity? Second, could participants match faces that had been used to create a morph (so-called constituent faces)? Results showed that morph personas were credible, being correctly matched more often than alternative images of real people. Further, results showed that propensity to declare a constituent face a match to its morph was dependent on the type of morph, with morphs made using two faces more likely to yield matches than those made using eight or 16 faces. A follow up similarity study on faces selected showed that even if a constituent face was selected it would more than likely have been ruled out as a match following a more thorough one-to-one comparison.

KEYWORDS:

• Facial comparison
• face matching
• identity fraud
• face morphs

Acknowledgements

The authors wish to acknowledge Ryan Dummin for the creation of the experimental application, Jessica O’Rielly for assistance with data collection, Veneta MacLeod for comments on this manuscript, and all participants in the research.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was conceptualized and funded by the Defence Science and Technology Group. Data were collected by the Defence Science and Technology Group and the University of Adelaide.