Human Facial Illustrations: Creation and Evaluation using Behavioral Studies and Functional Magnetic Resonance Imaging
Author(s): Bruce Gooch.
PhD Thesis: University of Utah, July,
2003.
[BibTeX]
Abstract:
This dissertation presents: a method for creating black-and-white illustrations
and caricatures of human faces from source photographs; and series of perceptual
studies aimed at evaluating the effectiveness of the resulting images relative to
photographs. The illustrations are generated by superimposing two images: a
thresholded image of the output of a computational brightness model, and a thresholded
luminance image. In addition, a new interactive technique is demonstrated
for deforming images of faces to create caricatures that highlight and exaggerate
representative facial features. The photographs and black-and-white illustrations
are evaluated via psychophysical studies to assess speed and accuracy in learning
and recognition tasks. These studies show that the facial illustrations and
caricatures generated using these techniques are as eective as photographs in
the recognition tasks. In the learning studies, tasks involving illustrations or
caricatures were performed significantly faster than the same tasks were performed
with photographs. The recognition invariance effect is used as an experimental
probe in a functional magnetic resonance imaging (fMRI) experiment. The results
of this experiment indicate that viewers may process illustrations differently from
photographs.