Correcting for the Effect of Orientation in Geometric Morphometric Studies of Side-View Images of Human Heads

Abstract

For all of the promise of laser scanners and three-dimensional (3D) digitizers, digitizing two-dimensional (2D) landmarks on photographic images is still the most convenient way of sampling the shapes of biological structures in geometric morphometric studies. Geometric morphometric methods extract shape variables from landmark configurations such that they are invariant to the configuration’s location, orientation and scale. Biological 3D structures can be variously rendered into 2D representations depending on, among other things, the choice of a point of view (i.e., object orientation with respect to the “camera”) that is employed in the process of imaging. In the case of side view imagesof human heads, a generalized Procrustes superimposition (GPA; Rohlf and Slice, 1990) of the 2D digitized landmarks removes the variability due to the size of the head in the image, its x and y position and its orientation around the imaginary z-axis (emanating orthogonally from the surface of the image). This means that geometric morphometrics methods control for variation among individuals in “nodding” (what would be called “pitch” in the aeronautical convention), but not for deviation from norma lateralis by way of head turning (rotation around the y-axis or “yaw”) and head tilting (x-rotation or “roll”).