|
|
Video Inpainting |
Existing video inpainting schemes are slow as they typically perform a global search on similar patches to fill the missing area. We take advantage of the stationary cameras and fixed geometry used in a video surveillance system. Using background subtraction and visual tracking, we collect object templates as shown in the top right video. To fill the hole shown in the bottom left, we first use image inpainting techniques to fill the area occludded by stationary objects and then formulate a dynamic program using the object templates to optimally recreate the occluded object motion. More sample sequences using our inpainting scheme can be found in http://www.vis.uky.edu/~vijay/research/video.htm To learn more, check out our publications. |
