Raster and vector processing for scanned linework
David D. Greenlee · 1987
Recent advances in scanning technology have made it possible for linework to be scanned on relatively inexpensive systems and output as raster images. These images can be edited, automatically thinned to connected lines, and converted to vector representations using conventional image processing techniques. An investigation of raster editing techniques, including thinning, skeletonizing, filling, and node detecting was performed by using software tests implemented on a microprocessor. The technique was based on encoding a three-by-three neighborhood surrounding each pixel into a single byte. This byte has a range of 0-255, and stores any possible surround condition for a pixel. It was found that various methods for thinning, filling, and node detection could be quickly implemented and tested using surrounding topology. It was possible to efficiently develop decision tables for raster editing, and to assure symmetry and consistency in the decisions. A prototypical method was also developed for converting the edited raster linework into vectors. Once vector representations of the lines were formed, they could be formatted as a Digital Line Graph, and further refined by deletion of nonessential vertices and by smoothing with a curve-fitting technique.