Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova

Brisinello, Matteo

master's thesis

Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova

Brisinello, Matteo (2017)

Josip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
Department of Computer Engineering and Automation
Chair of Automation and Robotics

Cite this document

APA 6th Edition

Brisinello, M. (2017). Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova (Master's thesis). Retrieved from https://urn.nsk.hr/urn:nbn:hr:200:802903

MLA 8th Edition

Brisinello, Matteo. "Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2017. https://urn.nsk.hr/urn:nbn:hr:200:802903

Chicago 17th Edition

Harvard

Brisinello, M. (2017). 'Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova', Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, accessed 12 August 2020, https://urn.nsk.hr/urn:nbn:hr:200:802903

Vancouver

Brisinello M. Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek; 2017 [cited 2020 August 12] Available at: https://urn.nsk.hr/urn:nbn:hr:200:802903

IEEE

M. Brisinello, "Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova", Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Osijek, 2017. Available at: https://urn.nsk.hr/urn:nbn:hr:200:802903

Cite this item: https://urn.nsk.hr/urn:nbn:hr:200:802903

Islandora Bookmark - Please login to use this feature

Metadata

Title	Unaprjeđenje algoritma za prepoznavanje tekstualnih znakova
Title (english)	Optical character recognition algorithm improvement
Author	Matteo Brisinello
Mentor	Ratko Grbić (mentor)
Committee member	Ratko Grbić (predsjednik povjerenstva)
Committee member	Mario Vranješ (član povjerenstva)
Committee member	Marijan Herceg (član povjerenstva)
Granter	Josip Juraj Strossmayer University of Osijek Faculty of Electrical Engineering, Computer Science and Information Technology Osijek (Department of Computer Engineering and Automation) (Chair of Automation and Robotics) Osijek
Defense date and country	2017-10-03, Croatia
Scientific / art field, discipline and subdiscipline	TECHNICAL SCIENCES Computing Information Systems
Abstract	Točnost algoritma za prepoznavanje tekstualnih znakova (engl. Optical character recognition - OCR) na slikama dohvaćenih sa set-top box (STB) uređaja igra vrlo važnu ulogu pri testiranju istih. Međutim, pokretanje algoritma za prepoznavanje tekstualnih znakova na takvim slikama često rezultira niskim postotkom točnosti. Uzrok tomu su navedene slike koje mogu biti niske rezolucije, loše kvalitete ili mogu sadržavat šarenu pozadinu. U ovom radu preporučene su četiri metode predobrade slike kako bi se poboljšala točnost prepoznavanja tekstualnih znakova. Ispitane su točnosti na slikama dohvaćenih sa STB uređaja korištenjem programa Tesseract 3.5 te relativno nove verzije Tesseract 4.0. Na tim slikama točnost prepoznavanja pomoću Tesseract-a 3.5 je 35.7%, dok Tesseract 4.0 ostvaruje točnost od 70.2%. Metode ispitane u ovom radu unaprjeđuju točnost algoritma za prepoznavanje tekstualnih znakova za 33.3% za Tesseract 3.5 te za 22.6% za Tesseract 4.0. Također je ispitana metoda koja odvaja tekst od pozadine kada su u pitanju slike sa šarenom pozadinom. Točnost prepoznavanja tekstualnih znakova nakon izvršavanja navedene metode je 94.4%. Također, izgrađen je dodatni korak automatizacije procesa prepoznavanja tekstualnih znakova na slikama sa šarenom pozadinom pristupom strojnog učenja. Navedeni pristup detektira sliku na kojoj je prikazan tekst nakon odvajanja teksta od pozadine. Testiranje istog rezultiralo je s točnošću od 95,5%.
Abstract (english)	Efficient Optical Character Recognition (OCR) in images grabbed from Set-Top Boxes (STBs) plays an important role in STB testing. However, running OCR software on such images usually ends with low OCR performance since images can have low resolution, low image quality or colorful background. In order to improve OCR performance, four different image preprocessing methods are proposed. In this paper OCR is performed with Tesseract 3.5 and the relatively new Tesseract 4.0 on the images grabbed from different STBs. On the original images Tesseract 3.5 provides a 35.7% accuracy while Tesseract 4.0 attains a 70.2% accuracy. The proposed preprocessing methods improve OCR performance by 33.3% for Tesseract 3.5 and 22.6% for Tesseract 4.0 on the available images. On images with colorful background, a method which separates text from background is proposed. OCR performance after separating text from background results with an accuracy of 94,4%. Aditionally, a step of automation in a process of reading text from images with colorful background was build with an approach of machine learning. This approach detects the image containing text after the separation from the background with the accuracy of 95,5%.
Keywords
Keywords (english)
Language	croatian
URN:NBN	urn:nbn:hr:200:802903
Study programme	Title: Graduate University Study Programme in Computer Engineering Study programme type: university Study level: graduate Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
Type of resource	Text
File origin	Born digital
Access conditions	Open access
Terms of use
Created on	2017-10-27 12:07:15