RestoPPXA_Lab
General software features
This software allows to restore images degraded by a convolution operator and a noise (Poisson or Gaussian). The method behind is based on convex criterion minimization. This criterion includes a data fidelity term (Kullback-Leibler divergence or l2 norm), an indicator function (e.g. pixel range constraint) and a regularization term that can be:
- a l1 norm applied on frame (DTT) coefficients
- a total variation term (TV)
- an hybrid regularization (l1 + TV)
PPXA (Parallel ProXimal Algorithm) is used to minimize the resulting criterion.