Multigrid Method for the Chan-Vese Model in Variational Segmentation

Multigrid Method for the Chan-Vese Model in Variational Segmentation

Year:    2008

Communications in Computational Physics, Vol. 4 (2008), Iss. 2 : pp. 294–316

Abstract

The Chan-Vese method of active contours without edges [11] has been used successfully for segmentation of images. As a variational formulation, it involves the solution of a fully nonlinear partial differential equation which is usually solved by using time marching methods with semi-implicit schemes for a parabolic equation; the recent method of additive operator splitting [19,36] provides an effective acceleration of such schemes for images of moderate size. However to process images of large size, urgent need exists in developing fast multilevel methods. Here we present a multigrid method to solve the Chan-Vese nonlinear elliptic partial differential equation, and demonstrate the fast convergence. We also analyze the smoothing rates of the associated smoothers. Based on our numerical tests, a surprising observation is that our multigrid method is more likely to converge to the global minimizer of the particular non-convex problem than previously unilevel methods which may get stuck at local minimizers. Numerical examples are given to show the expected gain in CPU time and the added advantage of global solutions.

You do not have full access to this article.

Already a Subscriber? Sign in as an individual or via your institution

Journal Article Details

Publisher Name:    Global Science Press

Language:    English

DOI:    https://doi.org/2008-CiCP-7791

Communications in Computational Physics, Vol. 4 (2008), Iss. 2 : pp. 294–316

Published online:    2008-01

AMS Subject Headings:    Global Science Press

Copyright:    COPYRIGHT: © Global Science Press

Pages:    23

Keywords: