SCS Undergraduate Thesis Topics

Student Advisor(s) Thesis Topic
Ian Graham James Kuffner Surface Capture of the Human Hand

Capturing the details of a moving surface with a system of cameras is a complex problem which has been receiving a great deal of attention in recent literature. Current methods for surface capture of human actors ("skin capture") do not yield results which are sufficiently realistic for use in the depiction of characters for computer animation. At the same time, current techniques for character animation lead to large amounts of production effort and cost to achieve realism when the deformations of skin must be rendered close-up, particularly in regions such as the hands or face where deformation is elaborate and subtle inconsistencies with reality are easy for the human eye to spot. A skin capture system with the ability to produce an animated 3D model whose deformations are indistinguishable from the original actor's skin would enable animators to produce characters with realistic skin using less time and effort.

We propose a new system for surface capture of the human hand, a subject which has proven difficult to depict in a realistic manner for character animation in films and computer games. Recent surface capture systems attempt to build an arbitrary surface from a subject about which little or nothing is known, and have trouble when confronted with a hand in a pose which produces an elaborate shape with many occlusions. In contrast to these systems, our approach will use structural knowledge of the human hand to very quickly estimate the position and shape of the surface of the skin and rapidly converge to an accurate representation of the current state of the actor's hand. The system will allow a new and unknown actor to set up, calibrate for their hand, and begin capturing in the span of a few seconds, much more quickly than typical setup times of several minutes for traditional motion capture systems. Finally, the method used will be generalizable for use in capture of other articulated subjects whose structural knowledge is available.

Close this window