A semi-automatic approach to detect highlights for home video annotation

This paper presents a semi-automatic highlight detection system for home video annotation. To automatically identify highlights from home videos is a challenging research issue in general. Currently home users mostly use video editing tools to manually find the highlight, which is very time consuming. To alleviate this hurdle and promote the reusability of home video material, we propose a semi-automatic user environment that aims at reducing the editing time required for users to find highlights. With a well designed user interface and using the localized visual similarity trail to estimate candidate highlight boundaries, we enable home users quickly and mostly accurately identify the highlight. The initial evaluation on a home video database demonstrates a 60% saving in editing time.