Robusna registracija slika temeljena na značajkama izlučenim iz slika rubova

scientific master's theses

• Title (english): Robust image registration based on features extracted from edge images
• Author: Darko Mihalić; VIAF: 305693911
• Mentor: Slobodan Ribarić (mentor)
• Committee member: Slobodan Ribarić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Electronics, Microelectronics, Computer and Intelligent Systems); Zagreb
• Defense date and country: 2007, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing; Data Processing
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

Registracija slika je disciplina koja se intenzivno razvija i istražuje, a svoju primjenu pronašla je na različitim područjima ljudskog života i rada. Do danas još nije razvijena računalna aplikacija iz područja registracije slike koja bi bila sveobuhvatna i koja bi se uspješno nosila sa svim problemima iz tog područja. Aplikacija razvijena za potrebe ovog rada predviđena je za registraciju parova IR – RGB slika fasada zgrada na kojima prevladavaju četvrtasti oblici, a temelji se na složenim značajkama koje su detektirane u slici rubova. Prilikom traženja optimalnih transformacijskih parametara IR slike poštivana su stanovita ograničenja s ciljem redukcije parametarskog skupa koji analiziramo. Optimalna transformacija je pronađena uz pomoć specifičnog algoritma razvijenog u ovome radu. Nakon pronalaženja optimalne transformacije IR slike primjenjuje se fuzija podataka, pribavljenih iz obiju slika (IR i RGB), s ciljem utvrđivanja mogućih mjesta oslabljene toplinske izolacije fasade zgrade. Aplikacija je testirana na više od 50 parova IR – RGB slika i pokazalo se da prilikom odabira vrijednosti parametara obrade slike treba uzeti u obzir prirodu slika koje podvrgavamo postupku registracije. Odabirom ispravnih vrijednosti parametara obrade slike u 90-ak posto slučajeva postignuta je ispravna registracija slika.

Abstract (english)

Image registration is a discipline which has been intensively researched and developed. Its use has been found within different aspects of human life and work. Until today, within image registration domain it hasn't been developed a computer application which would be overall and which could resolve all the problems within the domain. An application developed for the need of this work is intended for the registration of pairs of IR – RGB facade images with mostly square elements, and it is based on compound features that are extracted from the image of edges. While searching for the optimal transformation parameters of IR image, certain limitations have been respected in order to reduce the analyzed parametric set. The optimal transformation has been found by the use of a specific algorithm developed in this work. After finding an optimal transformation of IR image, a data fusion is applied on dana collected from IR an RGB images, in order to determine the points of possible reduced facade thermal isolation. The application has been tested on more than 50 pairs of IR – RGB images. The results showed that while choosing the image processing parameters, the nature of the images must be regarded. With the correct choice of value of image processing parameters correct image registration is achieved in about 90 percent of the cases.

• Keywords: registracija slike; rubovi; kompatibilne značajke; karakteristične točke; korespondentne točke; kriterij kvalitete poklapanja; optimalna transformacija; fuzija
• Keywords (english): image registration; edges; compatible features; characteristic points; correspondent points; criteria of overlapping quality; optimal transformation and fusion
• Language: croatian
• URN:NBN: urn:nbn:hr:168:234349
• Study programme: Title: Postgraduate master programe in computing; Study programme type: university; Study level: postgraduate; Academic / professional title: magistar inženjer računarstva (magistar inženjer računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-05-06 14:26:10