We develop a self-consistent automatic procedure to restore informations from astronomical observations. It relies on both a new deconvolution algorithm called LBCA (Lower Bound Constraint Algorithm) and the use of the Wiener filter. In order to explore its scientific potential for strong and weak gravitational lensing, we process a CFHT image of the galaxies cluster Abell 370 whichexhibits spectacular strong gravitational lensing effects. A highquality restoration is here of particular interest to map the darkmatter within the cluster. We show that the LBCA turns out specially efficient to reduceringing effects introduced by classical deconvolution algorithms in images with a high background. The method allows us to make a blind detection of the radial arc and to recover morphological properties similar to those observed from HST data. We also show that the Wiener filter is suitable to stop the iterative process before noise amplification, using only the unrestored data.