Digitalizacija i obrada slike i teksta u sustavima za optičko prepoznavanje znakova u domeni bibliotekarstva

master's thesis

• Title (english): Digitization, image and text processing in optical character recognition systems in the field of librarianship
• Author: Stela Kmetič
• Mentor: Sanja Seljan (mentor)
• Committee member: Vjera Lopina (član povjerenstva)
• Committee member: Ivan Dunđer (predsjednik povjerenstva)
• Committee member: Sanja Seljan (član povjerenstva)
• Granter: University of Zagreb; Faculty of Humanities and Social Sciences; (Department of information and Communication sciences); Zagreb
• Defense date and country: 2023-07-11, Croatia
• Scientific / art field, discipline and subdiscipline: SOCIAL SCIENCES; Information and Communication Sciences; Library Science

Abstract

U svijetu koji ide prema digitalizaciji svega, pa tako i tekstualnih podataka koji nisu izvorno digitalni, optičko prepoznavanje znakova ima centralnu ulogu. Takvi sustavi danas još nisu usavršeni, te ne postoji sustav koji ima 100%-tnu točnost prepoznavanja znakova. Donja granica za isplativost korištenja sustava za optičko prepoznavanje znakova često se citira kao 99,95%, a kako bismo se približili toj vrijednosti potrebna je i ljudska intervencija u raznim fazama rada sustava. Najčešće se takva intervencija predlaže u fazi pretprocesiranja slika prije korištenja istih u sustavima za optičko prepoznavanje znakova. Ovaj rad daje pregled procesa digitalizacije tekstualnog gradiva, pregled povijesti razvitka sustava za optičko prepoznavanje znakova, te kratak opis danas najpoznatijih i najefikasnijih takvih sustava na tržištu. Također istražuje i sve faze rada sustava za optičko prepoznavanje znakova, s naglaskom na razne metode pretprocesiranja, te njihovu implementaciju i efikasnost u poboljšavanju rezultata očitanja znakova u sustavima Tesseract i Google Document AI.

Abstract (english)

In a world that seems to be going towards digitising everything, including non-digitally native textual data, optical character recognition plays a central role. Optical character recognition systems haven't yet been perfected - a system with an accuracy rate of 100% in all cases is, as of the writing of this thesis, non-existent. 99.95% is often cited as the bottom barrier for accuracy necessary for the usage of optical character recognition systems to be cost-effective. To get near that number, human intervention is often necessary in many phases of an optical character recognition system’s work cycle. In literature, interventions are most often recommended in the phase of preprocessing, e.g., before we even feed an image to an optical character recognition system.
This thesis gives a brief summary of the process of digitising textual materials, the history of optical character recognition and the most popular and effective optical character recognition systems available today. It also explores all phases of an optical character recognition system's work cycle, with emphasis on preprocessing implementation and efficiency in increasing the rate of recognition in Tesseract and Google Document AI.

• Keywords: optičko prepoznavanje znakova; pretprocesiranje; poboljšanje slike; binarizacija; siva skala
• Keywords (english): optical character recognition; preprocessing; image enhancement; binarization; greyscale
• Language: croatian
• URN:NBN: urn:nbn:hr:131:825916
• Study programme: Title: Information Sciences; specializations in: Archivistics Study, Library Science, Information Sciences (research), Information Sciences (teaching), Museology and heritage management, Informatology; 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
• Created on: 2023-07-17 09:43:36