Variational methods in the field of image restoration are based on the minimization of a functional which is usually sum of a fidelity term and a regularization term. Assuming that the gradients in the pixels are distributed according to a Multivariate Generalized Gaussian distribution, we propose an improvement of the widely used TV (Total Variation) regularization term, adopting a space variant (SV) approach in order to distinguish regions of the same image with different properties.

This is poster number 24 in Poster Session

Authors:

Monica Pragliola (University of Bologna)

Keywords:

image deblurring, image reconstruction, inverse problems, variational methods