In this paper, a robust watermarking technique based on fractional cosine transform and singular value decomposition is presented to improve the protection of the images. A meaningful gray scale image is used as watermark instead of randomly generated Gaussian noise type watermark. First, host image is transformed by the means of fractional cosine transform. Now, the positions of all frequency coefficients are changed with respect to some rule and this rule is secret and only known to the owner/creator. Then inverse fractional cosine transform is performed to get the reference image. Watermark logo is embedded in the reference image by modifying its singular values. For embedding, the singular values of the reference image are found and then modify it by adding the singular values of the watermark image. A reliable watermark extraction algorithm is developed for extracting watermark from possibly attacked image. The experimental results show better visual imperceptibility and resiliency of the proposed scheme against intentional or un-intentional variety of attacks. © 2010 Copyright SPIE - The International Society for Optical Engineering.