Variational multiframe restoration of images degraded by noisy (stochastic) blur kernels

We wish to recover an original image u from several blurrynoisy versions fk, called frames. We assume a more severe degradation model, in which the image u has been blurred by a noisy (stochastic) point spread function. We consider the problem of restoring the degraded image in a variational framework. Since the recovery of u from one single frame f is a highly ill-posed problem, we propose two minimization problems based on the multiframe approach introduced for image superresolution by Marquina-Osher [28]. Several experimental results for greyscale and color image restoration are shown, together with binary image segmentation of noisy-blurry data and restoration of static videos, illustrating that the proposed models give visually satisfactory results.