Computerising scanned engineering documents

The paper presents an overview of techniques and algorithms for recognising scanned engineering images, with an emphasis on those for graphic element vectorisation and interpretation. Some background knowledge and basic issues regarding computerising scanned engineering documents are presented. Raster-to-vector conversion approaches are classified as interactive, semi-automatic or automatic, and they are discussed, respectively, with references to some of commercialised software. Various thresholding, segmentation and vectorisation methodologies are also discussed. A specific effort is made to provide information for evaluating and comparing techniques in the recent literature. Both strengths and weaknesses of these algorithms have been discussed, and the future trend envisaged.