Varational method using the Kuan filtering approach for the restoration of blurred images with multiplicative noise

Abstract

The main idea of the proposed restoration approach is the joint use of the Kuan filtering approach and a variational method of restoration. Three alternative formulations of the total mean quadratic error criterion are considered (stochastic, integral and differential). We show that the resulting integral and differential potential energies are well adapted for the purpose of image restoration as they correspond to regularization energies. The differential potential energy coincides with the regularization energy of Geman-McClure. A fidelity term to the data is introduced in the two integral and differential energies. The two methods are evaluated on different images blurred with different PSFs and degraded with multiplicative noise. The results are overall promising like those for most notconvex regularization energies.