BORG: a knowledge-based system for the automation of image segmentation tasks

A knowledge-based system that automates the execution of image segmentation processes is presented. Given an image and a set of segmentation tasks, our system, named BORG, generates, performs and optimises the processing of classical segmentation operators, in order to produce new images that satisfy these tasks. Such a system must have capabilities for self-configuration to different segmentation tasks and to different image classes. The authors describe the design principles for building image segmentation applications based on the generation of a graph of operators that can deal with the whole set of images belonging to some given class. A model of the hierarchical and incremental planning process used for building this graph is presented and the system architecture is discussed. An application dealing with biomedical images is also described