Recent advances in the complexity and manufacturability of digital video cameras coupled with the ubiquity of high speed computers and communication networks have led to burgeoning research in the fields of computer vision and image understanding. As the generated vision algorithms become increasingly complex, a need arises for robust communication between remote cameras on mobile units and their associated distributed vision algorithms.

A communication framework would provide a basis for modularization and abstraction of a collection of computer vision algorithms; the resulting system would allow for straightforward image capture, simplified communication between algorithms, and easy replacement or upgrade of existing component algorithms.

The objective of this thesis is to create such a communication framework and demonstrate its viability and applicability by implementing a relatively complex system of distributed computer vision algorithms. These multi-camera algorithms include body tracking, pose estimation and face recognition.

Although a plethora of research exists documenting individual algorithms which may utilize multiple networked cameras, this thesis aims to develop a novel way of sharing information between cameras and algorithms in a distributed computation system. In addition, this thesis strives to extend such an approach to using both stationary and mobile cameras. For this application, a mobile computer vision platform was developed that integrates seamlessly with the aforementioned communication framework, extending both its functionality and robustness.