An interactive document image description for OCR of handwritten forms

The design and implementation of a user-friendly graphical package for the interactive layout and content description of document images are discussed. The specific problems addressed are those associated with predefined forms, such as proposal forms used in the insurance business, which have been completed by hand. The document image description allows position and content of the handwritten fields to be defined and labeled and their interrelationships to be specified. This information assists the automatic processing of handwritten forms using OCR techniques by enabling appropriate OCR algorithms and local syntax constraints to be applied to each field to produce the best recognition performance.<>