Nonlinear diffusion based image segmentation using two fast algorithms (Q2668566)

A general variational model for image segmentation based on a nonlinear diffusion process is proposed. An existing method, the alternating direction method of multipliers (ADMM), which transforms a minimization problem into three subproblems that can be solved more quickly with well-known techniques, is applied to avoid solving nonlinear partial differential equations. Models and algorithms introduced in the paper are presented in detail and validated by means of several experiments with real and synthetic images. Using these new techniques it is possible to preserve the edges of objects in the images even in images with a high level of noise. While the paper is very technical, it is carefully written and not difficult to understand. It is noteworthy that the proposed methodology works well with RGB images. Finally, the list of bibliographic references is extensive and up to date.