Computing with Stochastic Signals.
Abstract
The project had set out to create a computer vision system for the analysis of natural scenes. The system was to split the process into several time scales such that it could later be implemented in parallel hardware (e.g., pixel-parallel, feature-sequential). The PI has fully succeeded in reaching these goals. He has created, tested and published a system that is able to recognize objects (e.g., faces) from freely taken digital images. Recognition is based on stored sample images. Recognition of new object types is possible simply on the basis of new sample images. The system has been extensively tested and optimized, with face recognition from large galleries as test environment. Results have been published. The system is ready for implementation in parallel hardware, such as that developed in the VIGILANTE project by S. Suddarth of BMDO in collaboration with JPL. Due to reduction of funding in the third year the PI has not been able to integrate figure-ground separation in the system.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 25, 1996
- Accession Number
- ADA329802
Entities
People
- Christoph Von Der Malsburg
Organizations
- University of Southern California