Normalized Matting of Interest Region

In this paper, we present an improved method of current closed-form solution for digital image matting. This method, which we call ‘normalized matting of interest region’, adopt the normalized cut technique where the objective function is normalized with the total degree of color similarities of foreground region. Unlike the existing solution, our method measures both the total dissimilarity between the foreground and background regions as well as the total similarity within foreground regions, which leads to better separation results, especially in case of extracting a specific region, rather than the closed-form solution. In addition, we employ a quadratic programming approach to solve the objective function to obtain a globally near-optimal matting result. Our method is empirically verified through several sample images.

[1]  Jitendra Malik,et al.  Normalized Cuts and Image Segmentation , 2000, IEEE Trans. Pattern Anal. Mach. Intell..

[2]  Dani Lischinski,et al.  A Closed-Form Solution to Natural Image Matting , 2008 .

[3]  Michael F. Cohen,et al.  An iterative optimization approach for unified image segmentation and matting , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[4]  David Salesin,et al.  A Bayesian approach to digital matting , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[5]  Stephen P. Boyd,et al.  Convex Optimization , 2004, Algorithms and Theory of Computation Handbook.

[6]  Dana H. Ballard,et al.  Computer Vision , 1982 .

[7]  Dani Lischinski,et al.  Spectral Matting , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.