A method for scene structure construction with time constraint

For more efficiently organizing, browsing and retrieving video, it is important to segment video to extract description structure. Usually video is segmented into shots, but since the shots in a long video are too many to be presented to the user, and they do not capture the underlying semantic structure of the video, high-level structure, such as act or scene, needs to be constructed. This paper proposes a time-constraint shot clustering method for video scene structure construction, in which shots are grouped into clusters based on time-constraint visual similarity, and clusters are merged into semantic-related scenes. The experimental results show that our method provides a better way for video browsing and retrieval.