Browse Prior Art Database

Character Recognition System

IP.com Disclosure Number: IPCOM000073607D
Original Publication Date: 1971-Jan-01
Included in the Prior Art Database: 2005-Feb-22
Document File: 2 page(s) / 29K

Publishing Venue

IBM

Related People

Gorbatenko, GG: AUTHOR

Abstract

The segmentation or separation of adjacent character images is one of the most difficult problems of optical character recognition. "Natural" segmentation schemes (i.e., those which find a continuous white path between characters) are desirable because they are easily implemented and because, with few exceptions, they are error-free.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 82% of the total text.

Page 1 of 2

Character Recognition System

The segmentation or separation of adjacent character images is one of the most difficult problems of optical character recognition. "Natural" segmentation schemes (i.e., those which find a continuous white path between characters) are desirable because they are easily implemented and because, with few exceptions, they are error-free.

The present system extends the range of character images which may be separated by the natural schemes. Specifically, it has been found that video thresholds chosen for optimum recognition are generally not optimum for segmentation. In the system shown, transducer 10 receives video information for amplification and filtering by analog video circuits 11. The processed video is then passed to digitizer 12 whose threshold is adjusted to be optimum for recognition logic 13. Video from analog circuits 11 is also coupled to a second digitizer 14, whose threshold is adjusted to be optimum for character segmentation. The two thresholds may be either independently adjustable, or fixed relative to each other. The segmentation threshold will generally be the higher, so as to make the characters appear lighter and more separated. As a practical matter, the high segmentation threshold may be capable of inserting a false blank in the middle of a single character. Therefore, switch 15 inputs the digitized video from digitizer 12 into segmentation logic 16 during an initial portion of each character. When scan counter 17...