In this paper, an analysis of existing methods of iterative blind deconvolution and the possibility oftheir usage for image processing is shown.
A large number of existing methods cannot guarantee the convergence to a global minimum, since they are based on the minimization of multimodal objective functions using gradient descent.
To improve their convergence multistart optimization method in wavelet transform domain (MOM in WT domain) is applied, cause it has reduced sensitivity to local extrema and the best rate of convergence for functions like "ravine".
The experiments demonstrate the ability of the new algorithm to reach the global minimum of the objective function for different initial conditions. This suggests to possible usage of MOM in WT domain for solving the blind deconvolution problem, for example, when processing distorted and blurry images.