On the Application of Voronoi Diagrams to Page Segmentation

Layout is a physical arrangement of document components such as text-blocks, text-lines, figures and tables in a page. The difficulty of layout analysis depends on a class of layout to be analyzed. An important and the most investigated class would be rectangular layout. Layout is rectangular if all document components are circumscribed by non-overlapping upright rectangles. In recent years, however, we often see pages outside this class in, for example, magazines and journals. Such pages include document components of arbitrary shape as well as tilted text-lines to make them attractive. Thus we need to develop methods which are capable of analyzing pages with non-rectangular layout.