Abstarct: Any of hiding data within a digital product, such as sound, image and video, in order to achieve goals such as copyright protection, detecting unauthorized copies and to prevent unauthorized copies, is called Watermarking. Different methods of Image Watermarking are defined on different domains. Discrete Cosine Transform and Wavelet Transform are the most important transform domains in image watermarking. We will introduce these two transform and discuss about some important watermarking techniques that are working in this domains. Also A new method of watermarking in discrete wavelet transform domain is introduced. In this method, the watermark is embedded in the third level wavelet coefficients of selected sub-images of host image, by using singular values decomposition technique and modifying singular values of host image with singular values of watermark image. In addition, we present a robust method against distortions caused by geometric attacks on image that using the feature points of host image for restoring the distorted image. This method is use Scale Invariant Feature Transform technique for detecting the feature points. By this method synchronization errors due to geometric attacks that retail of common errors, are reduced. The process of watermark embedding in discrete wavelet transform domain is performed on each sub-image. In watermark detection process, input image is restored baxsed on saved feature points in embedding watermark phase and the watermark is extracted. Also the final diagnosis watermark, is baxsed on the number of matched bits between extracted watermark and the original one. And at least, we evaluate the method with two Evaluation criteria, including the similarity between original image and watermarked image, and second normalized correlation to evaluate the resistance of watermarking method against geometrical attacks, evaluated the proposed method. Practical results obtained by this method showed high resistance against common geometric attacks and image processing.