3D face shape adaptation across changes in viewpoint, lighting and reflectance
Four experiments are reported using face adaptation to look at encoding of three-dimensional (3D) face shape across changes in viewpoint, lighting direction and surface reflectance. Categorization as Japanese or Caucasian was used as a task. A face-space was defined in terms of a principal components analysis (PCA) of 3D feature point positions with Linear Discriminant Analysis (LDA) used to define a direction within that space that best captured ethnicity. Adaptation was measure as a change in the point of subjective equality (PSE) along that dimension. Experiment 1 showed a shift in the point of subjective equality in the direction of the adapting population that transferred between views (15° and 45°). Experiment 2 used exaggerated adapting faces and showed increased adaptation that transferred across a change in modelled surface reflectance as well as view (from 30° to 0° and 60°). Experiment 3 showed that adaptation was not affected by having the same or different lighting at adaptation and test. Experiment 4 tested adaptation from 0° to 90° and found a fall-off with increasing angle of rotation. The results are interpreted as evidence for the encoding of 3D face shape independent of image properties at least over a limited range of viewpoint.
Please refer to publisher version or contact your library.