Facial image comparison

Introduction

In this chapter, the problems associated with the individualisation of people depicted in photographic forensic evidence such as closed circuit television (CCTV) images are described. Evidence of this type may be presented in court and, even with high-quality images, human identification of unfamiliar faces has been shown to be unreliable. Therefore, facial image comparison or mapping techniques have been developed. These have been used by expert witnesses providing opinion testimony as to whether two images depict the same person or not. With photographic video superimposition, one image is superimposed over a second so that a series of visual tests can detect differences or similarities in facial features. With morphological comparison analysis facial features are classified into discrete categories, providing an indication of whether these are similar across images. Finally, with photo-anthropometry the proportional distances and sometimes the angles between facial landmarks are calculated and compared. Recent research using each technique is described, and the difficulties associated with their application in forensic settings evaluated. At present, no method provides certainty of identification and great care should be taken if presented in court to obtain a conviction without substantiating alternative evidence.

Government and private sector investment in crime prevention initiatives has made CCTV systems common in many urban areas. Although there are no official records, the UK probably has the highest density in the world, with at least 3 million cameras nationwide (McCahill and Norris, 2003; Norris et al., 2004). There may be as many as 26 million cameras in the USA (Washington Post, 8 October 2005) and large-scale implementation seems inevitable elsewhere (Norris et al., 2004). Widespread deployment of CCTV raises many issues. Concerns have been raised about infringement of rights to privacy (Norris and Armstrong, 1999; Introna and Wood, 2004) and crime prevention efficacy (Brown, 1995; Gill et al., 2005). In this chapter, we focus on the reliability of CCTV for identification purposes.