Timescape Image Panorama Registration Techniques

It appears your Web browser is not configured to display PDF files. Download adobe Acrobat or click here to download the PDF file.

Click here to download the PDF file.

Creator: 

Ali, Heider K. Muhsin

Date: 

2017

Abstract: 

Panoramic image viewing in its common form simulates a comprehensive spatial or temporal view. The term panorama indicates the concepts of space (or place) and time. The space sense of panorama has dominated as the default concept since it has been scientifically and commercially applied for decades. However, the temporal concept of panorama, which has been recently investigated by researchers and historians, offers a new context of the definition of panorama. A novel application in the field of panorama display has been realized and presented by this work which aims to participate effectively in cultural heritage reservations around the world and to provide an easy access tool to demonstrate a timelined panorama view of landmarks. Consequently, a new concept of panorama has been accomplished through a timelined display of registered historic and modern images for many famous landmarks around the world. This is achieved by merging geographical and historical information into a single attractive temporal and spatial panoramic view; also known as a timescape. This comprehensive view is achieved from a collection of historic and modern images of such landmarks available on the Internet. Hundreds of thousands of landmark images covering more than one century of time have been collected using websites like Flickr and Google Images. We use Gabor filters and neural networks to select a subset of images that have roughly the same view for each landmark in order to keep the most suitable images that will be used to build and demonstrate the temporal panorama. The processing of the selected set of historic and modern images has to contend with radiometric differences, geometric distortions, and severely differing capture systems and environments in order to prepare them for alignment. Finally, the aligned images; warped or registered, were prepared for presentation in a timelined image display. The image registration is done using optimized optical flow of extracted SIFT features after applying a set of similarity measure standards to select the highest precision subset of aligned images and to develop the timescape display.

Subject: 

Engineering - Electronics and Electrical
Computer Science
Artificial Intelligence

Language: 

English

Publisher: 

Carleton University

Thesis Degree Name: 

Doctor of Philosophy: 
Ph.D.

Thesis Degree Level: 

Doctoral

Thesis Degree Discipline: 

Engineering, Electrical and Computer

Parent Collection: 

Theses and Dissertations

Items in CURVE are protected by copyright, with all rights reserved, unless otherwise indicated. They are made available with permission from the author(s).