Previous Issues

This paper examines the randomness inherent within the classical digital method used in data reduction in Geographical Information Systems - Raster Approach; namely Pixel Thinning. Although, this method is straightforward and widely applied in most Geographic Information Systems, especially in data reduction of Boolean Maps, but it is by no means acceptable cartographically, as it does not allow for preservation of feature characters during generalisation, especialy, at high levels of generalisation. The paper introduces a novel method which takes this context into account. It is called: Pixel-Averaged Classification. The results, both perceptual and numerical, show effectivness of the proposed method, during application to two different types of spatial arrangement of features. The study recommends that further evaluations are needed of different algorithms and methods which are designed for generalisation, as well as for devising new methods that takes into account the cartographic context.

Key words: Cartographic generalisation, Pixel thinning, Pixel aggregation, Pixel averaged classification, Feature character.