A robust framework for automated digital video inpainting

In this paper, a novel method for filling-in the missing information in images and videos is proposed. The algorithm presented here is able to reconstruct the structure as well as the texture. The pixel values are chosen by interpolating and searching the similar patch in a small window in the input image, and are filled in an order depending on their confidence values. The approach for video inpainting involves the automated tracking of the object selected for removal, followed by filling-in the holes while enforcing the global spatio-temporal consistency. Experiments show the effectiveness of our algorithm in tracking and removing large occluding objects as well as thin scratches.