Τίτλος : ¨On the contours of images in variational denoising”.

Περίληψη : “The idea of variational denoising of images is to find a minimizer of a suitable quantity, composed of a regularizing term and a fidelity term penalizing the distance from the noisy datum. One typically chooses the regularizing term of linear growth in the gradient, such as the total variation, in order to allow jump discontinuities (sharp contours) in the minimizer. It is then natural to ask where can the jumps appear. The focus of the talk will be a new approach to this question developed in collaboration with A. Chambolle (Paris), that significantly simplifies and generalizes previous results, in particular allowing the vector valued (color image) case”.