Seeing and measuring the 2D face

Susan Hayes

    Research output: ThesisDoctoral Thesis

    163 Downloads (Pure)


    This is a study of the factors that affect face shapes, and the techniques that can be used to measure variations in two dimensional representations of faces. The materials included thirty photographs of people in natural poses and thirty portraits that were based on the pose photographs. Visual assessors were asked to score the photographs and portraits in terms of pose (cant, turn and pitch) and also to compare the portraits to the photographs and score them in terms of likeness in the depiction of the face and its component features. Anthropometric indices were derived and used to score the images for the pose variables as well as for aspects of individual variation in external face shape and the spatial arrangement of the features. Geometric morphometric analysis was also used to determine the shape variation occurring in the photographs, the variation within the portraits, and to specifically discern where the portraits differ from the photographs in the depiction of head pose and individual differences in facial morphology. For the analysis of pose it was found that visual assessors were best at discerning the extent of head turning and poorest at discerning head pitch. These tendencies occurred in the visual assessments of both the photographs and the portrait drawings. For the analysis of the individual variation in face shapes it was found that external face shape varies according to upper face dimensions and the shape of the chin, and that vertical featural configurations are strongly linked to external face shape. When the portrait and photograph data were placed in the same geometric morphometric analysis the inaccuracies in the portrait drawings became evident. When these findings were compared to the visual assessments it transpired that, on average, visual assessment was generally congruent with the geometric morphometric analysis, but were possibly confounded by patterns of dysmorphology in the portraits that were contrary to what this study suggests are normal patterns of face shape variation. Overall this study has demonstrated that while anthropometric and visual assessments of facial differences are quite good, both were comparatively poor at assessing head pitch and tended to be confounded by the dysmorphologies arising in the portrait drawings. Geometric morphometric analysis was found to be very powerful in discerning complex shape variations associated with head pose and individual differences in facial morphology, both within and between the photographs and portraits.
    Original languageEnglish
    QualificationDoctor of Philosophy
    Publication statusUnpublished - 2009


    Dive into the research topics of 'Seeing and measuring the 2D face'. Together they form a unique fingerprint.

    Cite this