A Model-Based Field Frame Detection for Handwritten Filled-in Forms

摘要

In this paper, a method for detection and model definition of field frames in forms intended for handwritten data input is presented. The pre-printed field frames are extracted and parameterized to contribute to several tasks in the context of high-volume data-entry form scanning: assistance to the operator in the process of form specification, fine location of the pre-printed field frames, frame dropout and character restoration in the frequent case that handwritten characters data cross the field boundaries. Examples on real forms with very different layouts and characteristics are presented.