Image Segmentation By Contextual Region Growing Based On Fuzzy Classification

This paper presents a semantic image segmentation approach that combines a fuzzy region classification and a contextual region-growing. First image is over-segmented and a domain knowledge based fuzzy classification is applied on obtained regions to provide a fuzzy semantic labeling. This allows the proposed approach to operate at high level instead of using low-level features and consequently to remedy to the problem of the semantic gap. Each region is represented by its fuzzy membership degrees to the different thematic labels. The whole image is therefore represented by a regions' Membership Degrees Matrix. The segmentation is achieved on this matrix instead of the image pixels through two main phases: focusing and propagation. The focusing aims at selecting seeds regions from which information propagation will be performed. The propagation phase uses fuzzy contextual information to spread toward others regions the needed knowledge ensuring the semantic segmentation. An application of the proposed approach on Synthetic Aperture Radar (SAR) images shows promising results.