Computergestützte Inhaltsanalyse von digitalen Videoarchiven

The change from analog to digital videos in recent years has led to significant improvements in film archives. New possibilities for the archives arise due to the digitalization of films and videos. Wear-out and aging of film reels can be eliminated and a long-term preservation of the quality will be guaranteed. Additionally, the net-based access is much easier and faster than the manual transport of film reels. New services for archivists and users are available which enable new search possibilities and facilitate fast and efficient navigation during the playback of videos. Metadata provide additional information about the content of videos and support the search within the archives. In spite of the time exposure and high costs, a large part of the metadata is manually added by the archivists. The automatic analysis of digital video archives reduces the effort to create metadata significantly. Presented in the first part of this dissertation are new algorithms and techniques to identify and extract relevant semantic content in videos. In particular, new algorithms were developed to detect shot boundaries in videos, to analyze the camera motion, to segment and classify moving objects in videos, to perform optical character recognition, and to detect and recognize faces in videos. Automatically extracted semantic information is very valuable due to the fact that this information supports the work with digital archives. The additional information not only enables the search of videos within an archive but also leads to new applications, which are presented in the second part of this dissertation. Two sample applications are examined: automatically generated video summaries and video adaptation algorithms which enable the playback of videos on arbitrary devices. Another focal point of this dissertation is the analysis of historical films. Four European film archives provided a vast number of historical video documentaries stemming from the beginning to the middle of the last century. The storage and wear-out of the film reels over several decades led to noisy videos and a large number of errors in the images. The quality of the hi-