The paper explores the possibility of applying Markov random fields to restoration and segmentation at multiple abstraction levels within a multilevel network of virtual sensors. In particular, the feasibility of driving the segmentation process by using knowledge provided by an external information source is discussed. The interactions between the MRF levels are modelled by means of belief revision theory. As a case study, the segmentation (provided by a radiologist) of a magnetic resonance slice of the human head is considered as guiding knowledge. Quantitative evaluations of convergence times and of segmentation accuracy, together with visual results, prove the effectiveness of the proposed approach.
MULTILEVEL MRFS FOR INTELLIGENT IMAGE SEGMENTATION
C. S. Regazzoni;
1992-01-01
Abstract
The paper explores the possibility of applying Markov random fields to restoration and segmentation at multiple abstraction levels within a multilevel network of virtual sensors. In particular, the feasibility of driving the segmentation process by using knowledge provided by an external information source is discussed. The interactions between the MRF levels are modelled by means of belief revision theory. As a case study, the segmentation (provided by a radiologist) of a magnetic resonance slice of the human head is considered as guiding knowledge. Quantitative evaluations of convergence times and of segmentation accuracy, together with visual results, prove the effectiveness of the proposed approach.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
