Extraction of different urban area categories from satellite images using Window Independent Context Segmentation

Per-pixel based methods for spectral classification of remote-sensed images in urban areas are problematic because of the rather high spectral variability in urban materials and the fact that a specific spectral signature might appear in a number of different contexts in the urban landscape. It is by and large the specific contextual arrangement that defines the urban features, not the individual pixel´s spectral characteristics. Consequently the classification reliability has been lower in urban areas than in rural settings. In this paper a novel method called Window Independent Context Segmentation (WICS) is used to show that it is possible to extract categories of urban areas that differ in both function and underlying planning ideology from a SPOT5 satellite image covering the central parts of Stockholm, Sweden.