Global optimization based video completion

YiMin Yu,Duanqing Xu,Chun Chen,Lei Zhao,Yijun Yu
2006-01-01
WSEAS TRANSACTIONS ON COMPUTERS
Abstract:Video completion is a computer vision technique to restore the holes in video sequences due to removal of the dynamic or static foreground object by filling the unknown region with the known information. In this paper, a novel method for completion of large holes in video sequences is proposed. We extend the concept of pixels in images to video sequences by introducing the time information. Benefit from the extension, we can apply the information in other frames besides the frame containing the hole to be filled. Contrary to existing greedy ways, we formulate video completion as a global optimization problem to enforce global consistency and solve it using belief propagation to minimize the energy defined on 3D label graph. Measuring the similarity between source and target patches, we find the optimal patches by iterative steps and copy its color information to each node in the unknown regions to complete the synthesis. Experiment results on natural videos are encouraging and indicate that the performance of the proposed approach is effective.
What problem does this paper attempt to address?