Variational methods for imaging
Variational methods for imaging
One of the most difficult challenges in scientific computing is the development of algorithms and software for large scale ill-posed inverse problems, such as imaging denoising and deblurring. Such problems are extremely sensitive to perturbations (e.g. noise) in the data. To compute a physically reliable approximation from given noisy data, it is necessary to incorporate appropriate regularization into the mathematical model. Numerical methods to solve the regularized problem require effective numerical optimization strategies and efficient large scale matrix computations. In these lectures we describe first and second-order methods, dual or primal-dual approaches, and Bregman-type schemes and how to efficiently implement the ideas with iterative methods on realistic large scale imaging problems.
Docenti: Valeria Ruggiero - Luca Zanni
18-20 January 2016 (15 ore)
Lectures start Monday - January, 18, 2016, 9:00
Ferrara, Scientific-Technological Campus, Building B, Seminar room.
Preliminary material
Materials
Splitting methods for monotone operators (V. Ruggiero)
Lagrangian methods for convex optimization problems
I. Methods
II. Applications in Imaging Lab
First-order methods for differentiable optimization in imaging
I . Scaled projection methods
II. Applications in Imaging
III. Lab
