Skip to Main Content
We propose a blind deconvolution method based on the minimization of a cost function and the projection of an image onto the image space satisfying nonnegativity constraint and/or support constraint. These minimization and projection procedures are used iteratively in this method. The basic concept of this method and the constructed algorithm are shown in this paper. In computer simulation results, it is shown that the cost function decreased monotonically. This stable property was seen even when the support constraint was not used. However, this algorithm needs a lot of iterations for the convergence to the true image to be retrieved, and is sometimes suffered from the stagnation problem. A method to overcome this stagnation problem is also shown.
Date of Conference: 14-17 Dec. 2003