Non-parametric Patch based Video Matting

In computer vision, matting is the process of extracting foreground objects while preserving their pixel-wise coverage in the scene. This coverage is referred to as opacity or alpha matte. Once an accurate alpha matte is estimated, a foreground object can be seamlessly composited onto a new background. In this paper we present a novel patch based non-parametric approach for video matting. The technique provides a strong mechanism to represent local image features, colours and textures which attempts to preserve the spatial information of a natural video sequence. This overcomes the limitation of parametric algorithms [1, 2, 3, 5, 7, 8] which only rely on strong colour correlation or affinities between the nearby pixels. The matting problem was first formulated in [4] as linear interpolation of distinct foreground and background images to form a composite image as