In this study, a method for restoration of synthetic aperture radar (SAR) images is presented. We implement an iterative restoration algorithm on ERS-2 images of Istanbul, Turkey. We consider restoration to be a process that attemps to reconstruct or recover an image that is degraded by using some a priori knowledge of the degradation phenomenon. A degraded image often has a noise term. Speckle noise exists in all types of coherent imaging systems and its presence reduces the resolution of the image and detectibility of the target. The method handled in the study is based on a total sum preserving regularization ( TSPR) concept. The basic property of the TSPR method is that it preserves the mean values of local homogeneous regions and reduces speckle noise effectively. We obtain the original image from a degraded observed image. At each step of the technique the current pixel value depends on its neighbour ones, according to some regularizing parameters.