Categorical maps are a frequent data type in GIS applications and in thematic cartography. Examples include maps (or
databases) of soil, geology, vegetation, or classified remote sensing images. Networks of political or administrative
boundaries can be considered a special case of this maps type. Categorical maps are commonly modeled as either vector
data (i.e., as polygonal maps or polygonal subdivisions) or as raster data. Raster categorical data mainly originate from
grid samples, remote sensing imagery, or interpolated and classified point samples. Vector data are usually digitized
from the corresponding categorical maps. Although there are tools available in current commercial GIS and cartography
systems that allow processing raster and vector categorical data for purposes of analysis and display, specific methods
for automated generalization of such data are less well developed. They represent mere adaptations of methods developed
elsewhere to the problem of categorical map generalization. For vector categorical maps line generalization algorithms
are used instead of polygon-oriented methods. Techniques to “generalize” raster data sets are essentially equivalent
to simple pixel-based image processing operations, not respecting the object nature of “raster polygons” (regions,
connected components). More sophisticated methods of preliminary nature for both raster and vector categorical maps
have been proposed in the research literature, such as Schylberg (1993), Su et al. (1997) and Jaakkola (1998) for raster
data, or Muller and Wang (1992), de Berg et al. (1998) for vector models respectively. However, they need further improvement
and integration into a coherent framework and workflow if the generalization of categorical maps is to be
solved more comprehensively.