Convex non-convex image segmentation (Q2413469)

The paper considers the variational model for multiphase segmentation of images in which the minimized energy functional consists of a standard strictly convex quadratic fidelity term and a non-convex regularization term designed for penalizing simultaneously the non-smoothness of the inner segmented regions and the length of the boundaries. Sufficient conditions for convexity are derived for the model. An iterative minimization procedure based on the alternating direction method of multipliers is proposed. The convergence of the algorithm is established. The performance of the procedure is demonstrated by three numerical examples.