We present in this paper an image segmentation approach that combines a fuzzysemantic region classification and a context based region-growing. Input imageis first over-segmented. Then, prior domain knowledge is used to perform afuzzy classification of these regions to provide a fuzzy semantic labeling.This allows the proposed approach to operate at high level instead of usinglow-level features and consequently to remedy to the problem of the semanticgap. Each over-segmented region is represented by a vector giving itscorresponding membership degrees to the different thematic labels and the wholeimage is therefore represented by a Regions Partition Matrix. The segmentationis achieved on this matrix instead of the image pixels through two main phases:focusing and propagation. The focusing aims at selecting seeds regions fromwhich information propagation will be performed. Thepropagation phase allows tospread toward others regions and using fuzzy contextual information the neededknowledge ensuring the semantic segmentation. An application of the proposedapproach on mammograms shows promising results
展开▼
