Block based completion for video stabilization

Video stabilization helps eliminate unwanted camera motion, yet it also introduces missing regions near the edge of the frame. We outline a new method for the completion of stabilized video using a block based approach. Building upon macroblock recovery techniques, candidate blocks are chosen using the nearby motion vectors of blocks at the edge of the original frame. Then, the candidate blocks with the best boundary match are selected to fill in the missing region. Results on videos of dynamic scenes show realistic completions, with the potential to leverage existing block motion estimation hardware and significantly decrease processing time when compared to other state-of-the-art completion methods.