Constrained level-set method and its applications to image segmentation
Vladimír Klement, Tomáš Oberhuber, Daniel Ševčovič · arXiv (Cornell University) · 2011
We propose a new constrained level-set method for semi-automatic image segmentation. The method allows to specify which parts of the image lie inside respectively outside the segmented objects. Such an a-priori information can be expressed in terms of upper and lower constraints prescribed for the level-set function. Constraints have the same meaning as the initial seeds of the graph-cuts based methods for image segmentation. A numerical approximation scheme is based on the complementary-finite volumes method combined with the Projected successive over-relaxation method adopted for solving range bounds constrained linear complementarity problems. The advantage of the constrained level-set method is demonstrated on several artificial images as well as on cardiac MRI data.