A closed form solution to natural image matting

DMCA

A closed form solution to natural image matting

Cached

Download Links

Other Repositories/Bibliography

DBLP

by Anat Levin , Dani Lischinski , Yair Weiss

Venue:	Computer Vision and Pattern Recognition, IEEE Computer Society Conference on
Citations:	110 - 3 self

BibTeX

@ARTICLE{Levin_aclosed,
    author = {Anat Levin and Dani Lischinski and Yair Weiss},
    title = {A closed form solution to natural image matting},
    journal = {Computer Vision and Pattern Recognition, IEEE Computer Society Conference on},
    year = {},
    pages = {2006}
}

OpenURL

Abstract

Interactive digital matting, the process of extracting a foreground object from an image based on limited user input, is an important task in image and video editing. From a computer vision perspective, this task is extremely challenging because it is massively ill-posed — at each pixel we must estimate the foreground and the background colors, as well as the foreground opacity (“alpha matte”) from a single color measurement. Current approaches either restrict the estimation to a small part of the image, estimating foreground and background colors based on nearby pixels where they are known, or perform iterative nonlinear estimation by alternating foreground and background color estimation with alpha estimation. In this paper we present a closed form solution to natural image matting. We derive a cost function from local smoothness assumptions on foreground and background colors, and show that in the resulting expression it is possible to analytically eliminate the foreground and background colors to obtain a quadratic cost function in alpha. This allows us to find the globally optimal alpha matte by solving a sparse linear system of equations. Furthermore, the closed form formula allows us to predict the properties of the solution by analyzing the eigenvectors of a sparse matrix, closely related to matrices used in spectral image segmentation algorithms. We show that high quality mattes can be obtained on natural images from a small amount of user input. 1.

Keyphrases

natural image matting background color closed form solution spectral image segmentation single color measurement sparse linear system current approach computer vision perspective background color estimation natural image quadratic cost function closed form formula foreground opacity important task foreground object iterative nonlinear estimation resulting expression alpha matte optimal alpha matte small amount user input limited user input high quality matte alpha estimation small part interactive digital matting nearby pixel local smoothness assumption cost function sparse matrix

DMCA