This paper proposes a methodology in designing a spatial watermark which is robust to geometrical attacks. The proposed watermarking methodology is based on self-registering watermark that tiles the watermark pattern over the entire image. Thus, the peaks in the autocorrelation domain reveals the information about the geometrical transformations which the image has undergone. However, due to the limited precision of the autocorrelation domain, the template search is not reliable enough. The proposed scheme is based on a novel methodology in designing a watermark that is robust to small geometrical attacks. The watermark pattern is designed such that when the synchronization is off by the small amount of geometrical transformations, it can be identified without any searching. This characteristic of the watermark eventually leads to the reduction in search space of the template and compensation for the limited precision of the autocorrelation domain when the synchronization is off by the large amount. The proposed watermark is generated as a filtered white pattern, and in order for the watermark to be robust against geometrical transformation and lossy compression the filter must be carefully designed. The watermark generated by the filter designed by the proposed method has shown improvement in detection reliability.