Cue integration for urban area extraction in remote sensing images

Abstract

In this paper, we present a probabilistic framework for urban area extraction in remote sensing images using a conditional random field built over an adjacency graph of superpixels. Our discriminative model performs a multi-cue combination by incorporating efficiently color, texture and edge cues. Both local and pairwise feature functions are learned using sharing boosting to obtain a powerful classifier based on feature selection. Urban area are accurately extracted in highly heterogenous satellite images by applying a cluster sampling method, the Swendsen-Wang Cut algorithm. Example results are shown on high resolution SPOT-5 satellite images. © 2009 Springer Berlin Heidelberg.