Morphometric facial analysis: a methodology to create lateral facial images

Area of Science:

• Computer Vision
• Medical Imaging
• Facial Analysis

Background:

• Current facial analysis research primarily uses frontal projections.
• Lateral facial morphing analysis is underexplored, with unknown critical landmark features.

Purpose of the Study:

• Determine the optimal number of registry points (RP) for realistic lateral face synthesis.
• Identify key facial landmarks for creating synthetic lateral faces.

Main Methods:

• Generated 36 synthetic lateral faces with 50 to 250 RP.
• Evaluated realism of facial features (ear, eyebrow, eye, nose, lips, hairline, outlines) by seven experts using a 5-point Likert scale.

Main Results:

• Significant differences in realism were found between 200/250 RP templates and 50/100 RP templates (p < 0.05).
• Landmark variations were significant from 50 to 200 RP, but not between 200 and 250 RP.

Conclusions:

• Approximately 200 registry points (RP) are ideal for generating realistic lateral faces.
• Defined lateral facial registry point landmarks successfully create realistic synthetic faces.