Basic properties of synthetic aperture radar (SAR) images are determined by many different factors that include real radar antenna size and its pattern, synthetic aperture size, used windows and spatial sampling, number of looks and other imaging conditions, etc. Due to several factors, an obtained image is degraded by speckle and this speckle occurs to be spatially correlated. Below we propose a method to cope with spatially correlated noise based on discrete co-sine transform (DCT) filtering adapted to an image at hand. Simulation results demonstrating that due to exploiting the obtained estimates it is possible to improve filter performance by 1…2 dB in terms of PSNR are presented. Real life examples of SAR image processing are given as well.