An algorithm, which is based on the DCT domain, embedding a watermarking color image into a host color image is proposed to protect the copyrights of color images. The coordinates of R,G and B components are scrambled by logistic mapping before those components are embedded. Then the scrambled R,G and B components in color spaces are used as watermarking sequences. The host image blocks are classified by the luminance, contrast and textures. The algorithm uses some characteristics of human visual system, by which watermarking components with different strength can be embedded into some different DCT coefficients of image blocks according to the above classification. The experimental results demonstrate that the proposed algorithm has some good properties such as imperception and robustness, and it also can withstand some common attacks well.