A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science, School of Environment, Resources and Development

Thesis (M. Sc.) - Asian Institute of Technology, 1999

Nowadays, there is a growing need to conve11 new and existing paper documents into electronic documents for better archiving, retrieval and maintenance. In the field of GIS, as one of the most important methods of data input, image scanning can be faster, easier and more accurate, but the results are in the raster format which needs complicated raster - vector conversion and topological relationship reconstruction. This research applies the technologies of digital image processing, computer v1s1on, mathematical morphology etc to develop an image understanding software system for the analysis of color map images. A set of methods for extracting contour lines from color scanned map image has been presented in this thesis. The utilization of multiple information source and integration of various theories and methods are one of the main characters of this research. Color analysis, low-intermediate level visual algorithm, text - graphics separation and the development of object-orientated image analysis software etc. are also the topics of this thesis. Color Image Understanding system (named "CIU" by the author·) is developed to extract contour line from scanned color topographic map images. It is expected to be a useful PC software tool in the GIS application. In this research I propose methods for the analysis of color scanned map image. The main objective is to derive rcb11st and efficient methods that extract contour lines. The core of my methods consists of four components: color analysis, image analysis, contour line extraction, and vectorization. This Thesis consists of six chapters: introduction, literature review, methodology, color image understanding system, conclusions and recommendations, and references. Each of them is impo1tant and describes particular issues.