Tracking based depth-guided video inpainting

Abstract

In this paper, we propose a novel technique of tracking based video inpainting using depth information. Depth information obtained from the structure of motion is refined by extended proposed voting based algorithm. The refined depth map is used to extract moving foreground object from tracked moving object then replaces it into other video frame using integrated color and depth information based video inpainting. We compared the color based video inpainting with integrated color and depth information based video inpainting. Our proposed method acquaints special effect by including tracking and depth information to video inpainting. Inclusion of depth information increases the quality of inpainted video. Finally, we present experimental results of depth refinement and video inpainting for molecular video sequences captured with static camera with moving objects.