Spectral Matting

Anat Levin, Alex Rav-Acha, Dani Lischinski


We present 'spectral matting': a new approach to natural image matting that automatically computes a set of fundamental fuzzy matting components from the smallest eigenvectors of a suitably defined Laplacian matrix. Thus, our approach extends spectral segmentation techniques, whose goal is to extract hard segments, to the extraction of soft matting components. These components may then be used as building blocks to easily construct semantically meaningful foreground mattes, either in an unsupervised fashion, or based on a small amount of user input.

»The Paper (PDF)

»A Technical Report with some more details (PDF)

»Supplementary Material (PDF)

Ground Truth data

(open the compressed file using: tar -xvvzf GT.tar.gz)

Source Code

(open the compressed file using: tar -xvvzf code.tar.gz)

The spectral matting process:

Input Image

Smallest eigenvectors of the matting Laplacian matrix

Matting components - linear combinations of smallest eigenvectors

Foreground matte extracted by adding matting components

Comparison- hard segmentation