A Frame-based System for Modelling and Executing Visual Tasks

This paper describes a framework for model representation and control intended as the basis of a computer vision system capable of undertaking arbitrary but well defined image interpretation tasks. An inter-related structure of models represents both the visual task and image content. The user defines this structure but not the internal details of all models, so non vision experts can program the system. The approach is illustrated by considering two examples, one in interpreting cell images, the other in industrial inspection.