## Nonlinear diffusion filtering of images using the topological gradient approach to edges detection

• In this thesis, the problem of nonlinear diffusion filtering of gray-scale images is theoretically and numerically investigated. In the first part of the thesis, we derive the topological asymptotic expansion of the Mumford-Shah like functional. We show that the dominant term of this expansion can be regarded as a criterion to edges detection in an image. In the numerical part, we propose the finite volume discretization for the Catté et al. and the Weickert diffusion filter models. The proposed discretization is based on the integro-interpolation method introduced by Samarskii. The numerical schemes are derived for the case of uniform and nonuniform cell-centered grids of the computational domain $$\Omega \subset \mathbb{R}^2$$. In order to generate a nonuniform grid, the adaptive coarsening technique is proposed.

$Rev: 13581$