PhD thesis

AuthorChristian Petersohn
TitleTemporal Video Segmentation
TutorProf. Dr.-Ing. Thomas Sikora
AbstractThe objective of this thesis is to present steps toward simple and effective video access and browsing, to work towards technologies that can simplify annotation, automatic analysis, or video editing. This is done by developing methods and algorithms for the extraction of structural units in video on different hierarchical levels. The first problem examined is the extraction of video shots.
Visually complex shots with significant object or camera motion and a large variance in visual content cannot sufficiently be presented by a single key-frame. A richer and adaptive representation is needed. This second problem is investigated and as a result a new level in the hierarchy of temporal video segments, named sub-shots, is proposed.
The third problem addressed is the extraction of scenes. Various types of known approaches to scene detection are investigated.
Key wordspattern recognition, scene detection, sub-shot detection, shot boundary detection, video segments,
NoteNr. 26