A new concept for active fusion in image understanding applying fuzzy set theory

Image understanding applications must be capable of integrating uncertain information from a variety of sources. Fuzzy set theory is especially suited to provide methods to deal with and to fuse vague and ambiguous information arising in computer vision. We introduce a general framework, called 'active fusion', that actively selects and combines information in order to arrive at a reliable result at reasonable costs. An active fusion module is designed followed by an outline of how to implement such a framework using fuzzy set theoretic methods. The realization of a fusion/control unit of such an active fusion module for the efficient control of complicated processes in image understanding seems feasible. The presented framework will be implemented to carry out experiments in active object recognition.