Utjecaj predobrade ulaznih datoteka na točnost optičkog prepoznavanja znakova

master's thesis

• Title (english): The effects of the image pre-processing on the OCR accuracy
• Author: Mirela Majnarić
• Mentor: Hrvoje Stančić (mentor)
• Committee member: Tomislav Ivanjko (predsjednik povjerenstva)
• Committee member: Jozo Ivanović (član povjerenstva)
• Committee member: Hrvoje Stančić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Humanities and Social Sciences; (Department of information and Communication sciences); Zagreb
• Defense date and country: 2019-09-25, Croatia
• Scientific / art field, discipline and subdiscipline: SOCIAL SCIENCES; Information and Communication Sciences; Library Science

Abstract

Digitalizacija tekstualne građe danas je široko zastupljena u različitim domenama ljudskih djelatnosti. Najčešće se provodi skeniranjem ili fotografiranjem građe te upotrebom specijaliziranih programa za optičko prepoznavanje znakova. Na taj se način dobiva elektronička građa koju je moguće pregledavati, pretraživati i uređivati. Točnost dobivenih izlaznih podataka ovisi brojnim faktorima, a neki od njih su kvaliteta izvornika, razlučivost skeniranja, odabrani slikovni format, bitna dubina boje, ali i korišteni OCR softver. I postupak binarizacije, odnosno konverzije višebojne slike u crno-bijelu, utječe na rezultate optičkog prepoznavanja znakova. Binarizacija se obično provodi prilikom skeniranja ili kao jedan od koraka pri optičkom prepoznavanju znakova, a moguće ju je provesti i u nekom od programa za obradu slike prije korištenja OCR softvera. Cilj ovog rada jest istražiti utjecaj predobrade ulaznih datoteka na točnost optičkog prepoznavanja znakova s obzirom na navedene faktore. Točnost tekstova dobivenih primjenom OCR softvera ispitana je ISRI analitičkim alatima te je izražena postotkom ispravno prepoznatih znakova.

Abstract (english)

Digitization of textual materials is nowadays widely used in different domains of human activities. It is most commonly done by scanning or taking photos of the original source followed by using specialized software for optical character recognition. The result of this procedure is an electronic text which can be browsed, searched and edited. The OCR accuracy of the recognized text depends on numerous factors, including quality of the source materials, scanning resolution, chosen image file format, image bit depth as well as the OCR software used in the text recognition process. Image binarization or thresholding, a procedure in which a greyscale or colour image is converted into its bitonal variant, can also influence the OCR results. Binarization usually takes place during scanning or as a step in the OCR procedure, but it can also be accomplished by using a raster graphics editor, before the usage of the OCR software. The aim of this diploma thesis is to examine the effects of image pre-processing on OCR accuracy taking into account the above-mentioned factors. The OCR accuracy of the output texts was checked by ISRI Analytic Tools and is expressed as a percentage of correctly recognized characters.

• Keywords: optičko prepoznavanje znakova; OCR; točnost OCR-a; predobrada; razlučivost skeniranja; DPI; slikovni format; TIFF; JPG; kompresija slike; bitna dubina boje; binarizacija; ISRI analitički alati
• Keywords (english): optical character recognition; OCR; OCR accuracy; pre-processing; scanning resolution; DPI; image file format; TIFF; JPG; image compression; bit depth; binarization; thresholding; ISRI Analytic Tools
• Language: croatian
• URN:NBN: urn:nbn:hr:131:614062
• Study programme: Title: Information Sciences; specializations in: Library Science; Course: Library Science; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra bibliotekarstva (magistar/magistra bibliotekarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Public note: Izvanredni studij
• Created on: 2019-10-22 12:47:18