A new method improving watermark robustness against both global and local geometrical distortions is presented in this article. The proposed approach is based on a self-reference concept and exploits special autocorrelation features of a spatial template. The template allows identifying both the transformation parameters and translation coordinates. Distortions are estimated and reversed on a level as global as possible to maximize watermark recovery effectiveness and minimize a time needed for that purpose. Experimental results showed that an inserted watermark could be successfully read even after Stirmark attack.