STRUCTURED DOCUMENT IMAGE ANALYSIS: PREFACE -- Henry S. Baird

Preface to Structured Document Image Analysis, H. S. Baird, H. Bunke, and K. Yamamoto (Eds.), Springer-Verlag, New York, 1992.

Document image analysis is concerned with the automatic interpretation of images of printed and handwritten documents, including text, engineering drawings, maps, music scores, etc. Research in this field descends in an unbroken tradition from the earliest experiments in computer vision, and remains distinguished by close and productive ties between the academic and industrial communities. While the difficulty of its characteristic problems continues to stimulate basic research, general agreement on quantifiable performance standards has encouraged the evolution of sound engineering methods. As a result, research in this area supports a rapidly-growing industry.

We are pleased to offer a collection of state-of-the-art papers touching on virtually every topic of current research interest: printed documents; character and symbol recognition; handwriting; graphics, maps, and technical drawings; music notation; and methodology. Among these are several authoritative critical surveys of the literature. We have singled out music notation for special emphasis since it offers an ideal vehicle for sharing basic research internationally: identical scores are available, and are unambiguously understood, in every country.

Document images offer computer vision researchers unique opportunities. The early stages of processing such as feature extraction are relatively tractable, allowing rapid access to extraordinarily challenging later stages requiring the construction of full interpretations of complex scenes. As a result, a major focus of research is the architecture of complete, integrated vision systems, often exhibiting extremely high competency. We have included several parallel studies of this kind.

All but four of these papers were first presented at the International Association for Pattern Recognition workshop on Syntactic and Structural Pattern Recognition (SSPR'90), held at Murray Hill, New Jersey, USA, June 13-15, 1990. At that time, they were reviewed by three referees, and in many cases revised; since then, most have been expanded or extensively reworked for publication here. For improved balance, we have also invited four new contributions.

In the closing section of the book, we report briefly on the SSPR'90 workshop and several of its working groups. These groups, consisting of experts interested in a common topic, drew up lists of open problems and proven methods; their debates provide a fascinating perspective on the field.