Probabilistic reasoning in high-level vision

摘要

High-level vision is concerned with constructing a model of the visual world and making use of this knowledge for later recognition. In this paper, we develop a general framework for representing uncertain knowledge in high-level vision. Starting from a probabilistic network representation, we develop a structure for presenting visual knowledge, and techniques for probability propagation, parameter learning and structural improvement. This framework provides an adequate basis for representing uncertain knowledge in computer vision, especially in complex natural environments. It has been tested in a realistic problem in cndoscopy, performing image interpretation with good results. We consider that it can be applied in other domains, providing a coherent basis for developing knowledge-based vision systems.