Layout Understanding A Knowledge Based Approach

The layout of information is a complex task, and is usual-ly entrusted to a professional graphic designer. Research into automatic layout attempts to develop computational models of problem solving techniques used by graphic designers to solve layout problems. Graphic designers develop their problem solving skills through years of learning by following three approaches: i) learning by being told, ii) learning by example, and iii) learning by doing. Each form of learning involves the development of internal representations of that which is learned. This research seeks to develop computer-based knowledge representations of the result of the learning experiences from each of the three approaches, with the eventual goal of developing a methodology to build page layout understanding systems. In this research, understanding is defined as the task of automatically creating a high-level, structured representation of the layout from a structurally poor representation. As part of this research, we have built an experimental system which automatically converts a structurally poor representation of an existing layout an image obtained by optically scanning an existing layout into a high-level, structured representation the input layout is encoded as a set of constraint networks that encapsulate symbolic relationships between the elements in the layout. Thesis Advisor: Prof. Muriel Cooper Professor of Visual Studies, Program in Media Arts and Sciences This work was supported in part by Digital Equipment Corporation Layout Understanding A Knowledge Based Approach by Bagepalli C. Krishna The following people have served as readers for this thesis. urie ooper Professor of Visual Studies, Program in Media Arts and Sciences Thesis Advisor Henry Lieberman Rqseai'ch Scientist, Program in Media Arts and Sciences Joe Marks Member of Research Staff, Digital Equipment Corporation