DovCut: a draft based online video compositing system

Guangyu Nie, Yue Liu, Yongtian Wang

Abstract: The customization of the video endows users the ability to design the desired videos by themselves, however it is still a challenging task to achieve video composition up to now. In this paper, we design a draft based online video customization system (DovCut) that can convert the freehand draft into a realistic video. Based on a freehand draft in canvas with tags, the proposed system first searches the candidate videos online automatically to match the scene items and segments each candidate object from the background, then replaces the objects in the freehand draft into the realistic objects. By using a filtering scheme the proposed system excludes the undesirable video clips and selects the suitable videos automatically to generate a high quality composition video. Experimental results show the application potential of the proposed system.