This paper presents the results obtained in the development of an adaptive architecture for automated event discovery in sub-sea recorded video footage. The Video Marking System that has been built is the first step of the development of a vision system for an autonomous underwater vehicle. The principal aim of our work is to build an adaptive architecture that provides the vision system with the intelligence and robustness required to deal with the great variability found in underwater video. Different image processing techniques are embedded within the adaptive architecture; all of them running in parallel, adaptively parameterized, and assigned a confidence level by a voting system. Utilizing a simplified world model, the various outputs from the active image processing approaches are combined into an improved event description.