We present a new method for watermarking which is based on redundant wavelet transform (RDWT) and independent component analysis (ICA) for authenticity protection of multimedia products. RDWT gives an over complete presentation of the input sequences and functions as a better approximation to the wavelet transform, and therefore, is well-suitable for watermarking. An advantage of using a large logo is the ability to carry redundant information about copyright, increasing the robustness. By exploiting RDWT, we can embed a large binary logo watermark, which has the same size as the host image. The watermark is detected by a new intelligent ICA-based detector, with which we can directly extract the watermark in spatial domain without transformation steps. Our findings are that RDWT-based watermarking out-performs against image compression, and produces less image distortion than conventional DWT. The robustness of the proposed method is demonstrated and tested under various attacks.