In this master thesis seminar I am going to describe the design and
implementation of a system for generating detailed face models by
controlled lighting. The system is mainly based on two scanning
methods: structured light and photometric stereo. Structured light
scanning is used to capture the geometry, in which the performance is
improved through a sub-pixel precision step based on interpolation of
codewords in the reconstruction phase. With the second method,
photometric stereo, the goal is to compute the surface normals. The
quality of photometric stereo highly depends upon the accurate control
of the illumination condition, and the main part of the thesis is
devoted to the construction a light stage for performing photometric
stereo. Our light stage consists of 150 LED sources, each of which can
be illuminated and controlled over a wide range of resolution, and
hence provides the required level of precision.