SYSTEM AND METHOD FOR MERGING OBJECTS INTO AN IMAGE SEQUENCE WITHOUT PRIOR KNOWLEDGE OF THE SCENE IN THE IMAGE SEQUENCE

Abstract

This invention describes a method for execution by a data processor that seamlessly merges objects into an image sequence to obtain a montage image sequence, without knowing i(a priori) the content of the scene in the input image sequence. Such montage image sequence can be displayed to a viewer or recorded for future display. The method includes the steps of structure recovery, camera parameter derivation, silhouette computation, clone construction, and augmentation. Most of the steps are automated or semi-automated with minimal user intervention. The operation of the method includes the steps of interactively computing, displaying and recording merged image sequence in response to user commands to, for example, relocate inserted objects, change lighting parameters etc.