Video object extraction using feature matching based on nonlocal matting

Video object extraction is used for extracting foreground and background object from still image or video sequences. We proposed a new method extract an object from image or video sequences with nonlocal matting. One of disadvantages in nonlocal matting is user need to determine manually scribbles to specify foreground and background. To reduce a cost in this process, we add one more step using feature matching based on SIFT algorithm. SIFT is used to identify scribbles automatically based on similarity in image template that provided before. We try to extract objects from video which consist 80 image sequences. Our experiment by using this proposed method show a good result with θ = 0,05.