An Interactive System for 3-D Modeling from Videotape

K2T, Inc. proposes to design and demonstrate an interactive system capable of reconstructing 3-D models from ordinary video image sequences. Although many people are currently working in the field of shape from motion, no one has yet developed a commercially viable system for reconstructing 3-D models from videotape. The objective of this Phase I research is to design such a system. The system we propose is based on recent advances in image understanding theories, in particular the factorization method and related modeling by videotaping work being done at Carnegie Mellon University. The result of this Phase I effort will be a complete object oriented design specifying tasks such as feature extraction and tracking, shape motion reconstruction, 3-D modeling, representation conversion, and attribute recovery. A small portion of this system will also be implemented and used to evaluate the feasibility of the approach. The feasibility study and the system design will provide a foundation for Phase II development.