Towards robust object segmentation in video sequences and its applications [Elektronische Ressource] / vorgelegt von Mustafa Karaman

Sujets

• Informatik

Informations

• Publié par: technische_universitat_berlin
• Publié le: 01 janvier 2010
• Nombre de lectures: 9
• Langue: English
• Poids de l'ouvrage: 13 Mo

Extrait

Towards Robust Object Segmentation in
Video Sequences and its Applications
Von der Falkultat IV - Elektrotechnik und Informatik
der Technischen Universitat Berlin
zur Verleihung des akademischen Grades
Doktor der Ingenieurwissenschaften
-Dr.-Ing.-
genehmigte Dissertation
vorgelegt von
Dipl.-Ing. Mustafa Karaman
Promotionsausschuss
Vorsitzender: Prof.Dr.-Ing. Thomas Wiegand
1.Gutachter: Prof.Dr.-Ing. Thomas Sikora
2.Gutachter: Prof.Dr. Levent Onural
Tag der wissenschaftlichen Aussprache: 30.03.2010
Berlin 2010
D 83Just lousesi!.Acknowledgement
This visible work represents what I carried out in the last few years at the Com-
munication Systems Group of the Technical University Berlin. During this time I
received advice, support and help from many people. I am grateful to all of them
for the contribution they made to this work. In particular, I would like to express
my sincere gratitide to my supervisor Prof. Dr.-Ing. Thomas Sikora who gave me
the opportunity to join his group and the chance to work on the challenging and
intersting research topic. I appreciate his support and motivation during my work.
I also want to thank Prof. Levent Onural for the reviewing of my work and for the
unforgettable research meetings on behalf of 3DTV, an European research network.
Special thanks go to my colleagues at the institute for sharing time with me inside
and outside the o ce. In particular, I would like to thank my roommates Lutz Gold-
mann and Amjad Samour for the unforgettable discussions, motivating words, times
before specic deadlines, and their support. Furthermore, I am grateful to Birgit
Boldin, the kind soul of the institute, for her support in any important organizational
a airs. I would like to thank Sebastian Knorr and Matthias Kunter for their partial
proofreading of my thesis and their support.
Due to my participation in several European projects I had the opportunity to
meet and exchange ideas with numereous researchers working on the same research
domain. In particular, I am grateful to Elisa Drelie, Alexandra Koutsia, and Thomasz
Adamek. Furthermore, I would also like to thank the students for their contributions
I had the opportunity to supervise them.
Finally, my warmest thanks go to my wife Filiz and my children Selina and Sinan,
my parents and family. Without their love and continous support I would not be able
to carry out this work. - tak tak - Tatlim.
vEidesstattliche Erklarung
Ich versichere an Eides statt, dass ich die von mir vorgelegte Dissertation selbststandig
ange- fertigt und alle benutzten Quellen und Hilfsmittel vollstandig angegeben habe.
Eine Anmeldung der Promotionsabsicht habe ich an keiner anderen Fakultat oder
Hochschule beantragt.
viiAbstract
Object segmentation is one of the fundamental challenges in image processing and
computer vision since it is the preliminary step for many object analysis tasks and
applications. Object analysis such as object recognition often requires that objects of
interests are well segmented. Dierent visual cues such as color, texture, and motion
can be exploited to achieve an accurate object segmentation. The primary goal of
this thesis is to develop a robust object segmentation method for videos acquired
from static cameras typically used in surveillance applications. Therefore, it explores
and improves background modeling and subtraction techniques.
In an object segmentation framework, a video goes through dierent processing
stages before being used for the