Sports video clustering

• Main Author: Ng, Brian Kim S.
• Format: text
• Language: English
• Published: Animo Repository 2005
• Subjects: Video; Sports; Cluster analysis > Computer programs; Cluster set theory
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/3307
• Institution: De La Salle University

Video browsing has become more and more challenging over the past couple of years due to increased digital video availability. Without efficient indexing and organization, searching for videos over a huge media library poses a real problem. The goal of this research is to use clustering algorithms to group similar sports videos based on color and camera motion features. Sports genre is suitable for clustering since most sports videos have color and camera motion features that serve as important visual cues and can be used to identify them. A statistical method for combining color and camera motion features for clustering will also be developed. Once the sports videos have been clustered according to the similarity of their features, video retrieval can be done directly on a particular cluster, thus improving retrieval performance.