Detection of unknown forms from document images
Andrew W. Busch, Wageeh Boles, Sridha Sridharan, Vinod Chandran · 2003
This paper presents a novel technique for distinguishing images of forms from other document images. The proposed algorithm detects regions which are likely to be used for text entry, such as lines, boxes, and character entry fields, and calculates a probability of the document being a form based on the presence of such structures. Experimental results from testing on both filled and unfilled forms, as well as a selection of non-form documents are presented. All document images are assumed to have been scanned at a known resolution.