Mosaic representation and video object manipulations for post-production applications
Henri Nicolas · 2002
This paper presents new techniques for object-based video manipulations using mosaic representations of the fixed background of the scene. For that purpose, a new efficient camera motion estimation technique based on temporal linking, full search method and descent gradient methods is proposed to create the mosaic image. This global information is therefore used for video editing (for example suppression of an object), or to artificially change the camera point of view, i.e., the camera location and its focal. This kind of automatic tool is of great interest for the video post-production world in which, with the existing techniques, long and tedious manual efforts are necessary to achieve such object-based video manipulations.