An Object Removal Using Multi-view Sequence Inpainting Technique

A framework for object removal in multi-view video sequences is proposed in this work. After a mask cloning procedure, we arrange the sites at which source patches are pasted over mask areas. To each site, candidate blocks are assigned by a decimation scheme, which helps to reduce the computational complexity in the following procedures. Finally, we choose the best source block from the candidates by a global optimization technique. Simulation results illustrate that the proposed algorithm removes the objects in the mask area effectively, while preserving the spatial-temporal consistency.