Location of specific items in a document, without using character recognition techniques

Abstract

In many fields of engineering, the scanning of various documents has become an essential input medium to provide digitized information for subsequent processing. A continuing problem, though, is the location of specific items in the document, without resorting to processor-intensive character-recognition techniques. This paper presents an automated system which locates the position of “amount due” in printed bills and checks. It starts by digitizing images. After a digital image has been acquired, a dynamic threshold value is computed, based on the results of an edge-detection algorithm. A binary image is thus obtained, by thresholding the gray scale image. Then the long horizontal and vertical line segments in the image are removed. This is followed by labeling the connected components, extracting features and classifying. Finally, an inference scheme is used to locate the position of “amount due”. An example is used to illustrate the process, which has considerable relevance in the wider fields of manufacturing and computer vision.