Identification using face regions: application and assessment in forensic scenarios

Area of Science:

• Forensic Science
• Biometrics
• Computer Vision

Background:

• Forensic examiners conduct region-by-region facial comparisons.
• Quantitative analysis of facial region discriminative power is crucial for forensic identification.
• Existing methods lack standardized data on the identification efficacy of individual facial regions.

Purpose of the Study:

• To quantitatively assess the discriminative power of human facial regions in forensic contexts.
• To compare the effectiveness of different facial analysis methodologies.
• To understand how acquisition distance influences the identification power of facial regions.

Main Methods:

• Comparison of manual and automatic facial landmark extraction techniques.
• Evaluation of various facial region extraction algorithms.
• Analysis of face image data acquired at different distances.
• Utilized MORPH and SCface databases for mugshot and CCTV image comparisons.

Main Results:

• The discriminative power of facial regions varies significantly based on acquisition distance.
• In certain scenarios, specific facial regions demonstrated higher identification performance than the entire face.
• Methodologies for facial landmark and region extraction impact identification accuracy.

Conclusions:

• Facial region analysis offers valuable insights for forensic identification.
• Acquisition distance is a critical factor influencing the utility of specific facial regions.
• This research provides a quantitative basis for forensic examiners to prioritize facial features during comparisons.