Image Restoration Using Gradient Iteration and Constraints for Band Extrapolation

Images obtained with optical microscopes are often blurred and corrupted by noise. Specially in wide-field microscopy, there is a strong out-of-focus blur in focal plane direction, caused mainly by a frequency loss in the acquisition process. In order to restore these images it is important to use non-linear methods and relevant prior knowledge as constraints. In this paper, we propose a projection onto convex sets framework using a sequence of prototype images inspired by the Richardson-Lucy iterative algorithm. Within this framework it is possible to use constraints while restoring the image, during the iterations. We prove that the iterative method can be implemented as a gradient iteration, followed by a sequence of projections onto constraint sets. Moreover, we performed experiments using the techniques in order to improve restoration. The results showed a better restoration of frequencies, an improved focal plane restoration and a better contrast when compared with the regular iterative restoration algorithm.