A Form Document Image Parser ?

Forms are used extensively as input to many information management systems. An automatic way of processing a form should be preferred over the traditional methods due to high volume and numerous kinds of form documents. A system that minimizes operator error, cost and user interaction has to be developed. This paper proposes a system for automatically parsing form documents that have textual and nontextual components. The parser involves two parts. In the rst part, the form document goes through scan and preprocessing steps; then line extraction , vertex identiication and cell representation are carried out for geometric layout extraction. In the next part, content labelling among textual components, preprinted and user lled-in, is performed based on an unsupervised clustering. The presented system is shown to be robust by the performed experiments.