User-oriented Content Labeling in Remote Sensing Image Archives

In this paper we present an intuitive way of semantic labeling of image content suited for query by image content from remote sensing image databases. The space of image content understanding is separated into a syntactic and a semantic space. The syntactic space contains objective, signaloriented classes of certain types of structures and the semantic space contains subjective, user-oriented content labels. We expand on the Bayesian inference between the two spaces and provide examples for semantic content labeling in both optical (Landsat TM) and synthetic aperture radar (XSAR) data.