This paper presents a novel image watermarking algorithm based on Discrete Cosine Transform (DCT). To imbed the watermark, we first divide the image into non-overlapping partition. Then, after applying DCT, named as one-level DCT, to the host images blocks, and DC coefficients of each block are extracted. Next, we perform DCT again, named as two-level DCT, on the matrix comprised of the selected DC coefficients of each image blocks. Finally, the watermark is embedded on the part of the coefficients after the two-level DCT. The host signal interference (HIS) is the persistent concern in the spread spectrum principle. This paper overcomes the drawback by substituting spread spectrum watermark (SSW) for the original values. However, in the traditional way, these original values are modified by adding with SSW. Moreover, fixed-number blocks for image segmentation plays an important role in resisting scaling attacks. Experimental results demonstrate that the proposed method is resilient to many kinds of image processing attacks and geometric distortions, such as filtering, noise adding, JPEG compression, scaling, aspect ratio change and some kinds of combined transforms.