Context-based detection of post-disaster damaged buildings in urban areas from satellite images

This paper presents a new context-based damage detection approach using high-resolution optical satellite images. To exploit rich information represented in such high-resolution images, the proposed approach is designed to learn the image context based on the combination of the edge-based textures, the multi-spectral gray tone, and the spatial relationship formulated through a morphological scale-space. While edge-based textures infer the debris of damaged buildings, multi-spectral gray tone and spatial relationship provide the cues to extract the retained buildings after a catastrophe. It is demonstrated through processing a QuickBird image acquired over the damage areas of Bam city, Iran, which were severely hit by the earthquake on 26 December 2003.