Threshold dynamics for shape reconstruction and disocclusion

We propose a very efficient numerical algorithm for minimizing certain curvature dependent functionals that appear in a variety of well known variational models of image processing and computer vision. It has many applications, such as image inpainting, image segmentation, and surface fairing in computer graphics. The proposed algorithm is generalized to continuous interface version that has better resolution while keeping the same formal complexity. As an example, we apply our technique to a shape reconstruction problem based on Euler´s elastica.