Due to the increasing amount of digital videos the segmentation and classification of videos is manually no longer feasible. Hence there is a need for algorithms, which are able to filter out relevant information from the video material by using suitable and significant descriptions. This diploma thesis presents a system for classification of videos through analyses of audiovisual features. This is a complex problem on arbitrary video material because the features should be able to gather the semantic meaning of pictures and audio signals from the videos. Therefore, this thesis is limited on the scope of the video classification using scenes of one particular type of video, the Muppet Show. First, the fundamental approaches and methods for video analysis are explained in a detailed research. After a short overview over the development of the Muppet Show, a subsequently analysis of video material shows the characteristic attributes. Based on the gained knowledge significant audiovisual features and suitable classification models are presented, which are employed for the development of the prototype. Finally the quality of the classification results are evaluated using different tests. The intention is to show that visual features such as the distribution of colours as well as the segmentation of audio signals in speech, music and environmental sounds are able to capture the semantic meaning of video scenes of the Muppet Show.
C. Fuchs: "Videosegmentierung durch Analyse audiovisueller Merkmale"; Supervisor: H. Eidenberger; Softwaretechnik und Interaktive Systeme, 2013.
Click into the text area and press Ctrl+A/Ctrl+C or ⌘+A/⌘+C to copy the BibTeX into your clipboard… or download the BibTeX.