Segmentation of the Yellow Pages

We present a fully automated process to scan the Australian Telecom Yellow Pages and produce a text document consisting only of the business entries, while removing the advertisements, graphics, and notes about the Yellow Pages. The system contains four major components: digitisation and thresholding, skew detection, segmentation (removal of unwanted parts of the image), and finally the recognition engine utilising the principles of mathematical morphology. This paper presents the current research, which consists of the process described above up to image segmentation. All the algorithms are written in C on a 5000/20 DEC workstation. We have tested more than 30 images with extremely promising results. 1: Introduction Over the past few years much research has been devoted towards finding automated methods for document segmentation. The main motivation for this effort has come from the need to process documents containing mixed text, graphics, line segments and other forms which can be..