Influence of types tone value on character formation and accuracy of optical character recognition

Abstract

The characters, primary carriers of textual information, fulfil a significant role in print quality. Character formation is important not just for the human reader, but also to a machine based reading and optical character recognition (abbreviation OCR). In this paper we investigated the influence of the characters optical density on the accuracy of commercial and open source optical recognition performance. The intention was to investigate the possibility of toner saving and the thresholds for typeface sizes that will enable good readability. The test chart with three type sizes (24pt, 12pt and 6 pt) and two typefaces (Times New Roman as a serif and Arial as sans serif type) were printed by varying the optical density through tone values (from 10 % to 100 % by a 10 % increment). Test form was printed on 80 g m–2 uncoated office paper using Riso EZ 570E digital screen printing system. The prints were scanned using a calibrated HP Scanjet G3010. Scanning resolution was set to 300 spi, as a recommended value for this kind of documents. Character area and perimeter were characterized using image analysis method (open source ImageJ software). For optical character recognition we used several OCR software: OmniPage, Abbyy Fine Reader, ReadIris Pro, Presto Pro, CuneiForm. The results obtained from the study indicate that text tone value has an important role in OCR accuracy for both type faces.