Extracting raster GIS data from scanned thematic maps

One of the problems which continues to trouble practical applications of GIS is the time and cost of capturing data from paper maps. Even where the data are required in raster form the complexity of most maps means that the data are first digitised in vector form, and then rasterised, a slow and error-prone process. This paper describes an alternative approach based upon readily available hardware and software in which the map is scanned using Spot Colour Scanning, and then processed using a specially designed majority filter and editing with a painting package. Tests on a small scale soil map produce results which are about 80% as accurate as digitising, but take far less time to produce. An analysis of the sources of error indicates that many of these are due to the difficulty of scanning such a complex document as a map, and suggests that more sophisticated image processing algorithms may be needed to classify the original scanned image before the application of the filter algorithm.