Urban block outlining in high-resolution SAR images based on detection of linear features

While analyzing remotely sensed images of urban areas, in many cases an at least approximate knowledge of its block boundaries is useful for specializing operations over areas, within which a certain degree of homogeneity can be assumed. In the case of use for emergency management, though, this information may not be accessible in a reasonable time or with a reasonable effort, while it would be useful as a basis for e.g. earthquake damage assessment at the block level [1]. Automatic extraction of block boundaries thus becomes an interesting tool to fill this possible gap. City blocks are usually separated by major urban roads, and a linear feature extractor, originally developed for road network extraction, can provide basic features upon which to partition a very high resolution SAR scene acquired over an urban area. After the first experiments described in [2], the method has been improved and new results produced on a series of COSMO/SkyMed images over geohazard-prone areas.